The NCAR Atmospheric Simulation Chamber consists of a 10 m3 FEP Teflon (0.005 in. thick) bag that is housed in a cubic enclosure with UV reflective surfaces and a bank of 128 wall-mounted blacklight tubes (32 W, type F32T8/BL). To characterize photolytic conditions in the chamber, irradiance spectra were collected in the wavelength range of 180–600 nm at ∼0.8 nm resolution by a custom-built spectroradiometer, as shown in Fig. S1 in the Supplement (Petropavlovskikh et al., 2007). Photolysis frequencies were calculated based on the measured downwelling spectral actinic fluxes. The computed photolysis rate of NO2 (JNO2∼1.27×10-3s-1) agrees within 3 % with that measured by photolyzing 18.6 ppb NO2 in the chamber and monitoring the NO production rate. The chamber is equipped with a standard set of measurements, including an integrated temperature and humidity probe (model 50U, VAISALA, CO) and a Magnehelic differential pressure indicating transmitter (model 605-11, Dwyer Instruments, IN). The chamber temperature is controlled at 295 K by the building's air conditioning system and increases to 305–306 K under maximum irradiation conditions. The relative humidity of the chamber air is below 10 % under dry conditions (the remaining water vapor is generated from methane combustion during the air purification process) and can be varied in the range of ∼10–50 % by flowing a portion of the purified dry flushing air into the chamber through a temperature-controlled water reservoir. Typical temperature and relative humidity profiles across the chamber under maximum irradiation conditions are given in Fig. S1 in the Supplement. The chamber internal pressure is maintained slightly above the ambient pressure to minimize the enclosure air contamination via penetration through the Teflon film.

Prior to each experiment, the chamber was flushed with purified dry air from an ultra high purity zero air generator (model 737, Aadco Instruments, OH) for >12 h until ozone and NOx levels were below 1 ppb. During the operation of the steady-state continuous flow mode, the chamber was constantly flushed with purified dry air at 40 Lmin-1, which gives an average chamber residence time of 4.17 h. The incoming and outgoing flows were balanced by a feedback control system that maintains a constant internal pressure of 1.2–4.9×10-4 atm above the ambient. The chamber is actively mixed by the turbulence created by the 40 Lmin-1 flushing air. The characteristic mixing time is defined as the time it takes for the measurement signal of a tracer compound (e.g., CO2 and NO) to stabilize following a pulse injection. The average mixing time in the NCAR chamber was determined to be ∼9 min, which is ∼4 % of the residence time. Under such conditions, the gas- and particle-phase composition in the outflow can be assumed identical to that in the well-mixed core of the chamber.

To mimic daytime photochemistry in the continuous flow mode, steady-state OH mixing ratio was created by photolyzing hydrogen peroxide (H2O2) vapor that was continuously flowing into the chamber (H2O2+hv→2OH, JH2O2∼4.33×10-7s-1). Specifically, a 20 mL syringe (NORM-JECT, Henke-Sass Wolf, MA) held on a syringe pump (model 100, KD Scientific, MA) kept at ∼4 ∘C was used to deliver H2O2 solution (1–30 wt %, Sigma Aldrich, MO) into a glass bulb that was gently warmed at ∼32 ∘C. The liquid delivery rate was sufficiently slow (100–300 µLmin-1) that all the H2O2 vapor was released into the glass bulb through evaporation of a small droplet hanging on the needle tip. An air stream (5 Lmin-1) swept the H2O2 vapor into the chamber, resulting in an H2O2 mixing ratio in the range of 600 ppb to 16.22 ppm in the injection flow as a function of the concentration of H2O2 aqueous solution used. A spreadsheet (Table S2) for calculating the inflow H2O2 mixing ratio using the above input method is provided in the Supplement. As H2O2-laden air was continuously entering the chamber, it took approximately three turnover times (∼12.5 h) for the desired H2O2 vapor mixing ratio to reach steady state in the chamber. The H2O2 vapor concentration in the chamber, though not measured, can be estimated from the steady-state OH mixing ratio derived from the observed exponential decay of a given parent hydrocarbon. Constant NO injection flow was achieved by diluting the gas flow from a concentrated NO cylinder (NO =133.16 ppm, balance N2) to a desired mixing ratio (0.1–100 ppb) using a set of mass flow controllers (Tylan FC260 and FC262, Mykrolis Corp., MA). The lowest steady-state NO level that can be achieved in the chamber is around 30 ppt (unpublished, NCAR). Note that for experiments performed in the absence of any VOC precursor, H2O2 and NO were the only two species that were continuously input into the chamber for the establishment of a combination of different photochemical conditions as denoted by the O3 and NOx measurements. For the isoprene photooxidation experiments, an isoprene standard (C5H8=531 ppm, balance N2) was constantly injected into the chamber and diluted with the flushing air to achieve an inflow concentration of ∼20 ppb.

To mimic the nighttime chemistry in the continuous flow mode, steady-state NO3 mixing ratio was created by constantly flowing diluted O3 and NO air into the chamber (NO+O3→NO2+O2; NO2+O3→NO3+O2). The NO source can be replaced by NO2, although the absolute absence of NO does not necessarily represent the actual atmospheric conditions. O3 was produced by photolyzing O2 in air at 185 nm using a mercury Pen-Ray lamp (UVP LLC, CA). Ozone concentration in the injection flow can be controlled from 3.5 to 457 ppb automatically by adjusting the mercury lamp duty cycle. To study the RO2+HO2 pathway, formaldehyde (CH2O) was input into the chamber along with NO and O3 to initiate HO2 production (NO3+CH2O+O2→HNO3+HO2+CO). Formaldehyde aqueous solution (37 wt %, Sigma Aldrich, MO) was diluted with ultrapure water (Milli-Q, Merck Millipore, MA) to 0.2–1.0 wt % and continuously input into the chamber using the same method used for H2O2 input described above. It is worth noting that the formaldehyde aqueous solution contains 10–15 % methanol as stabilizer to prevent polymerization. The presence of methanol in the chamber does not significantly impact the nocturnal chemistry as it consumes OH and NO3 radicals to generate formaldehyde and additional HO2 (CH3OH+NO3→HNO3+CH2O+HO2, CH3OH+OH→H2O+CH2O+HO2) (Atkinson et al., 2006). The use of formaldehyde as an HO2 source mimics closely the atmospheric nighttime conditions in forest environments (Schwantes et al., 2015). To study the NO3-initiated oxidation of isoprene, an injection flow of diluted isoprene (∼ 10 ppb) was achieved using the procedure described above.

A suite of instruments was used to monitor gas-phase concentrations in the chamber outflow. O3 was monitored by absorption spectroscopy with 0.5 ppb detection limit (model 49, Thermo Scientific, CO). The O3 monitor was calibrated using an Ozone Primary Standard in the range of 0–200 ppb (model 49i-PS, Thermo Scientific, CO). The O3 monitor used for chamber experiments was periodically checked with the primary standard and was shown to be stable over long periods of time (less than 1 ppb drift in over 1 year). NO was monitored by chemiluminescence with 0.5 ppb detection limit (model CLD 88Y, Eco Physics, MI). Zero-point and span calibrations of the NOx monitor were performed prior to each experiment by supplying the instrument with pure N2 gas and diluted NO, respectively. Multi-point calibration was performed on a weekly basis and has shown a good stability and linearity in the NO measurement ranging from 1 to 200 ppb. NOx (NO+NO2) measurements were performed using a photolytic converter that selectively converts NO2 to NO upstream of the photo-multiplier tube in the CLD 88Y NO monitor. This converter uses two opposing arrays of UV LEDs shining into a cylindrical quartz mixing tube to achieve approximately 50 % conversion of NO2 to NO per second. The total efficiency for the equipment described here is approximately 70 % as determined by measuring calibrated NO2 standards. The sample path always includes the photolytic converter, and typical experiments cycle the power for the LED lights to switch between measuring NO (lights off) and NOx (lights on). NO2 concentrations were then determined by subtracting the NO from the adjacent NOx measurements.

A customized proton transfer reaction quadrupole mass spectrometer (PTR-Q-MS) was used to measure volatile organic compounds including isoprene (C5H8), MACR (C4H6O), and MVK (C4H6O). The instrument was operated at 2.3 mbar drift pressure and 560 V drift voltage. Measurements reported here were obtained at a sampling rate of 10 Hz. In positive-mode operation, a given analyte [M] undergoes proton transfer reaction, producing an ion of the form [M+H]+; that is, isoprene is detected as ion C5H9+ (m/z 69) and MACR and MVK are both detected as ion C4H7O+ (m/z 71). The instrument background was collected by sampling the chamber air for at least 30 min prior to each experiment. Measured ion intensities for isoprene (C5H9+, m/z 69) and MACR and MVK (C4H7O+, m/z 71) were calculated as the signal of each ion (counts per seconds) normalized to the total ion signal of H3O+. The instrument sensitivities towards isoprene, MACR, and MVK were calibrated with a mixture of diluted gas standards. The instrument sensitivity towards MACR is identical to that of MVK and, as a result, the sum of MACR and MVK concentration in the sampling air can be calculated by applying one calibration factor to the measured C4H7O+ (m/z 71) signal intensity. Since artifacts in the measured C4H7O+ signal can be produced through thermal decomposition of isoprene oxidation products, such as the peroxides, nitrates, and epoxides, on contact with hot metal surface (Liu et al., 2013; Nguyen et al., 2014b; Rivera-Rios et al., 2014), a cold-trap system was used to avoid bias in the interpretation of the PTR-Q-MS data. Specifically, a 1 m section of Teflon tubing was submerged in a low-temperature ethanol bath (-40 ± 2 ∘C) that could trap oxidized products less volatile than the authentic MACR and MVK standards after steady state was established in the chamber. The quantification of the sum of MACR and MVK was then based on the PTR-Q-MS measured C4H7O+ (m/z 71) signal downstream of the cold trap.

Figure 1 shows the model-predicted steady-state mixing ratios of OH, HO2, NO3, NO, NO2, and O3 in the chamber after 20 h of photochemical reactions as a function of the H2O2 and NO concentrations in the continuous injection flow. Six blank chamber experiments were compared with simulations. In general, the model captures the evolution patterns of NOx and O3 well. The predicted mixing ratios of NO, NO2, and O3 agree within 69, 11, and 33 %, respectively, with the measurements (see Table S1 and Fig. S2 in the Supplement). The relatively large NO uncertainties originate from the measurements that were performed close to the instrument detection limit (0.5 ppb).

Simulated steady-state mixing ratios of OH radicals ([OH]ss) range from ∼5×105 to ∼4×106 moleccm-3, which over ∼4 h chamber residence time would be roughly equivalent to ∼1 to ∼8 h of atmospheric OH exposure (1×106 moleccm-3). As expected, [OH]ss increases with increasing NO influxes due to the enhanced NOx / O3 cycling but decreases with increasing H2O2 influxes due to the overwhelming reaction OH+H2O2→H2O+HO2. As a consequence, the steady-state mixing ratios of HO2 radicals ([HO2]ss) reach up to ∼7×109 moleccm-3 when 16.2 ppm H2O2 is continuously injected into the chamber. If 110 ppb H2O2 is used instead, the resulting [HO2]ss levels fall close to the ambient range (∼108 moleccm-3).

Simulated steady-state NO mixing ratios in the chamber range from ∼2 ppt to ∼0.9 ppb from combinations of 0.1–20 ppb NO and 0.11–16.22 ppm H2O2 in the injection flow. The ratio of inflow NO concentration to the steady-state NO concentration in the chamber ranges from 5 to 93, demonstrating the importance of chemical removal in controlling the overall steady-state NO levels. O3 accumulation is an inevitable consequence under photolytic conditions and, for example, the presence of 10 ppb O3 leads to the chemical removal term (kO3+NO⋅[O3]⋅τ) in Eq. (1) that reduces the steady-state NO concentration by a factor of 60. It is worth noting that under all simulated conditions in the continuous flow mode, O3 (∼1–126 ppb) coexists with NO (∼0.002–0.9 ppb). This particular chemical scenario, which is impossible to achieve in batch-mode reactors due to prompt conversion of NO to NO2, could then be used to mimic ambient ozonolysis chemistry, for example, in forest regions that frequently encounter polluted air masses from nearby urban areas. The steady-state mixing ratios of NO2 ([NO2]ss) exhibit a strong linear correlation with NO influxes. The use of less than 20 ppb NO in the injection flow results in a few to tens of ppb [NO2]ss, which is higher than the range typically found in the ambient. The potential “quenching” effect of NO2 on RO2 radicals through reversible termolecular reactions is discussed shortly.

In the so-called “high-NOx” chamber experiments, the NO3 radical is an unavoidable side-product when black lights are used as a representative of the solar radiation in mimicking the daytime photochemistry in the troposphere. The photolysis of NO3, although its primary sink in the atmosphere, proceeds rather slowly (JNO3∼1.8×10-3s-1) under the present chamber photolytic conditions, thereby leading to a significant accumulation of NO3 radicals (7.9×104–2.8×108 moleccm-3) at steady state. The simulated NO3 / OH ratio dictates the extent to which the NO3 (nighttime) chemistry competes with the OH-initiated (daytime) photochemistry. For compounds that are highly reactive towards NO3 such as isoprene, NO3-initiated oxidation accounts for up to ∼60 % of the overall isoprene degradation pathways at the highest NO3 / OH ratio (∼255) simulated. Low concentrations of NO (<20 ppb) and H2O2 (<2 ppm) in the injection flow are therefore necessary to limit the interferences of NO3-initiated chemistry. Again, taking isoprene as an example, the NO3 oxidation pathway contributes less than 0.1 % of the overall isoprene degradation kinetics at the lowest NO3 / OH ratio (∼0.13) simulated here.

Also given in Fig. 1 is the calculated lifetime (τRO2) of an RO2 radical with respect to reactions with NO and HO2 at 306 K. In most batch-mode chamber experiments, τRO2 of only several seconds or less can be achieved due to the presence of tens to hundreds of ppb levels of NO and HO2. Here τRO2 could extend to 60 s or even longer with the continuous input of low mixing ratios of H2O2 (≤110 ppb) and NO (≤0.2 ppb). Note that the presence of tens of ppb NO2 in the chamber might impose a “quenching” effect on the steady-state RO2 level through rapid reversible reactions (RO2+NO2+M↔RO2NO2+M). We evaluate this potential “quenching effect” using ethyl peroxy radical (C2H5O2) generated from OH oxidation of ethane as an example. Simulations shown in Fig. S3 in the Supplement reveal that incorporation of the C2H5O2+NO2+M↔C2H5O2NO2+M reaction into the mechanism in the presence of ∼1–80 ppb NO2 does not notably change the behavior of C2H5O2 radical. One exception is the peroxyacyl radical, which combines with NO2 yielding peroxyacyl nitrate. For example, under 0.1–16 ppb [NO2]ss as displayed in Fig. 1, we calculate that the time needed for peroxyacetyl radical (CH3C(O)O2) to reach equilibrium with peroxyacetyl nitrate (CH3C(O)O2+NO2+M↔CH3C(O)O2NO2+M) ranges from ∼ 1 to ∼ 10 s, suggesting that the lifetime of peroxyacyl radicals is ultimately controlled by NO2 instead of NO / HO2 in the reaction system and, consequently, peroxyacyl radicals are not expected to be long-lived in the current chamber configuration.

We further compare the photochemical oxidation environment created here with the “intermediate-NO” conditions achieved by other chambers that employed the experimental approaches introduced earlier. In terms of the oxidizing power, all approaches are capable of maintaining an atmospheric relevant OH level (∼106 moleccm-3) except the “slow chemistry” scenario that limits the photolysis rate of the OH precursor and results in an average OH mixing ratio of ∼105 moleccm-3 (Crounse et al., 2012; Teng et al., 2017). At comparable OH levels, the overall atmospheric OH exposure achieved in the flow tube reactor is rather limited due to the short residence time (e.g., ∼80 s in the PAM reactor). In terms of the NOx level, precisely controlled steady-state NO concentration can be achieved for an indefinite time period by operating chambers in the continuously mixed flow mode. However, NO2 accumulates during the continuous oxidation process and the resulting NO2 / NO ratio can be as much as an order of magnitude higher than that achieved in the static outdoor chambers.

MACR and MVK are major first-generation products from the OH-initiated oxidation of isoprene in the presence of NO (Wennberg et al., 2018). They are produced from the decomposition of β-ISOPO alkoxy radicals that are primarily formed from the reaction of β-ISOPOO peroxy radicals (β-1-OH-2-OO and β-4-OH-3-OO) with NO; see mechanisms displayed in Fig. 2a. Reactions of β-ISOPOO peroxy radicals with HO2 and RO2 also partially yield β-ISOPO alkoxy radicals that ultimately lead to MACR and MVK, although these pathways are considered to be minor in the presence of hundreds to thousands of ppt NO in the atmosphere. The molar yields determined from previous studies range from 30 to 35 % for MVK and from 20 to 25 % for MACR under high-NO conditions (NO >60 ppb) (Tuazon and Atkinson, 1990; Paulson and Seinfeld, 1992; Miyoshi et al., 1994; Ruppert and Becker, 2000; Sprengnether et al., 2002; Galloway et al., 2011; Liu et al., 2013). It has been recently shown that the six hydroxyl peroxy radicals (ISOPOO) produced from the initial OH addition to the double bonds of isoprene undergo rapid interconversion by removal or addition of O2 that competes with bimolecular reactions under atmospherically relevant NO levels (Peeters et al., 2014; Teng et al., 2017). As a result, the distribution of ISOPOO radical isomers and their subsequent reaction products varies with their lifetimes with respect to bimolecular reactions. In the presence of hundreds of ppb NO as done by most previous experimental studies, the reaction of ISOPOO radicals with NO dominates over their interconversion, and the production of β-ISOPOO peroxy radical is less favored, leading the experiments to underestimate the MACR and MVK yields typically obtained in ambient conditions. Measurements by Karl et al. (2006) and Liu et al. (2013) conducted at NO concentrations comparable to the moderately polluted urban environment (∼0.2 ppb in Karl et al. and ∼1 ppb in Liu et al.) found higher MACR (∼27 % in Karl et al. and ∼31.8 % in Liu et al.) and MVK (∼41 % in Karl et al. and ∼44.5 % in Liu et al.) yields than other studies.

Here we perform a steady-state continuous-mode experiment to measure the production of MACR and MVK from the OH-initiated oxidation of isoprene in the presence of ∼0.45 ppb NO. Figure 3 shows the observed and simulated temporal profiles of NOx, O3, C5H8, and C4H6O over 24 h photooxidation of isoprene. In this experiment, C5H8, H2O2, and NO were continuously fed into the chamber, with constant inflow concentrations of 19.9, 600, and 19 ppb, respectively. An outgoing flow at 40 Lmin-1 continuously withdrew air from the chamber to balance the pressure. After approaching steady state, the sampling tube was submerged into a low-temperature ethanol bath (-40 ± 2 ∘C) to trap oxidized products that would otherwise undergo thermal decomposition introducing interferences in the C4H7O+ (m/z 71) signal. The measured concentrations of C5H8 and C4H6O upon cold trapping agree within 2.8 and 4.6 % uncertainties with the model simulations (see Fig. 3c and d).

To calculate the total molar yield (YC4H6O) of MACR and MVK, two reactions are considered: C5H8+OH⟶k1YC4H6O⋅C4H6O+products,C4H6O+OH⟶k2products, where k1 is the rate constant for OH reaction with isoprene, and k2 is taken as the average of rate constants for OH reactions with MACR and MVK. Uncertainties associated with the simplification of k2 in calculating the MACR and MVK yields will be discussed shortly. Note that the ozonolysis and NO3-initiated oxidation in total account for less than 6 % of isoprene degradation pathway under current experimental conditions and are neglected in the calculation. The ozonolysis and photolysis in total account for ∼6 % of the C4H6O degradation pathway and are neglected here as well.

In the continuous-mode operation, two mass conservation equations are satisfied at steady state: d[C5H8]ssdt=[C5H8]0/τ-[C5H8]ss/τ-k1⋅[OH]ss⋅[C5H8]ss=0d[C4H6O]ssdt=YC4H6O⋅k1⋅[OH]ss⋅[C5H8]ss-k2⋅[OH]ss⋅[C4H6O]ss-[C4H6O]ss/τ=0 where [C5H8]ss and [C4H6O]ss are the PTR-Q-MS measured steady-state concentrations of isoprene and the sum of MACR and MVK when using the cold trap, respectively, [C5H8]0 is the initial concentration of isoprene, and τ is the chamber mean residence time and can be calculated as the total chamber volume divided by the incoming and outgoing flow rate. The steady-state OH radical concentration ([OH]ss) can be derived by solving Eq. (2). The calculated [OH]ss (3.13×106 moleccm-3) is 12 % higher than the model prediction (2.74×106 moleccm-3). The molar yield of the sum of MACR and MVK from isoprene OH oxidation pathway in the presence of ∼0.45 ppb NO is thus given by Eq. (4) and calculated as 76.7 ± 5.8 %: YC4H6O=[C4H6O]ss+k2⋅[OH]ss⋅τ⋅[C4H6O]ssk1⋅[OH]ss⋅τ⋅[C5H8]ss×fβ-ISOPOO+NO. Here a 5.8 % uncertainty originates from the assumption that MACR + OH and MVK + OH proceed with equal reaction rate, although the rate constant for MVK + OH is ∼31 % lower than that of MACR + OH. Another potential uncertainty relates to the accuracy of the simulated steady-state HO2 and RO2 concentrations and the contribution of β-ISOPOO + HO2 and β-ISOPOO + RO2 reaction pathways to the overall β-ISOPOO fate. The fraction of β-ISOPOO radicals that reacts with NO (fβ-ISOPOO+NO) was predicted as 0.88, which was used here to scale the final MACR + MVK yield, see Eq. (4).

In summary, the measured yield of the sum of MACR and MVK in this study is close to that reported by Karl et al. (2006) and Liu et al. (2013), but ∼27–52 % higher than the majority of previous measurements performed under high-NO conditions (NO >60 ppb). This is consistent with the dynamic nature of the six ISOPOO radical isomers that undergo rapid interconversion by addition or removal of O2. In the presence of ∼0.45 ppb NO as performed in this study, lifetimes of the β-1-OH-2-OO peroxy radical with respect to reaction with NO and loss of O2 are estimated as 9.4 and 0.2 s, respectively, implying that the rapid interconversion between β-ISOPOO and δ-ISOPOO radicals essentially governs their distribution; under such conditions, the production of thermodynamically more stable β-ISOPOO isomers is favored, leading to higher yields of MACR and MVK. Here the reported MACR and MVK yield from isoprene OH oxidation in the presence of ∼0.45 ppb NO represents an illustration of chamber operation at steady-state continuous flow mode for the establishment of certain experimental conditions that are not easily accessible from traditional batch-mode chamber experiments. A complete measurement of first-generation oxidation products from isoprene OH reaction under a wide range of NO levels (ISOPOO bimolecular lifetimes) will be forthcoming in a future publication.

Figure 4 shows the model simulated steady-state mixing ratios of HO2, NO3, NO, NO2, and O3 after 16 h of dark reactions in the chamber as a function of the HCHO concentration and O3 / NO ratio in the continuous chamber inflow. Blank experiments were compared with simulations in five cases (see Table S1 in the Supplement). The model captures the evolution patterns of NOx and O3 well. The observed mixing ratios of NO2 and O3 agree with the simulations to within 11 and 6 %, respectively (Fig. S4 in the Supplement).

Compared with the photochemical reaction schemes discussed earlier, the nocturnal chemistry is rather straightforward; that is, the inflow O3 / NO ratio governs the steady-state concentrations of NO3, NOx, and O3, while the inflow HCHO concentration ultimately controls the steady-state HO2 level. Increasing the O3 / NO ratio from 1 to 9 in the continuous inflow leads to increased NO3 from 2.4×105 to 1.1×109 moleccm-3, but decreased NO from 1.8 ppb to 20 ppt and decreased NO2 from 18 to 7 ppb. At a fixed inflow O3 / NO ratio, doubling the NO and O3 concentrations leads to elevated NO3, NOx, and O3 by a factor of 2.0–3.2, 1.5–2.0, and 1.4–2.0, respectively. The use of HCHO as an effective dark HO2 source does not significantly impact the steady-state mixing ratios of NOx and O3, but slightly weakens the NO3 production.

The calculated RO2 lifetime (τRO2) with respect to reactions with NO, NO3, and HO2 at 295 K ranges from 3 to 225 s. The highest τRO2 was achieved in the absence of any HCHO source and corresponds to a chemical regime that can be employed to study the intramolecular isomerization (autoxidation) pathway of RO2 radicals, if any. Adding a continuous flow of HCHO to the system leads to the production of 107–109 moleccm-3 HO2 radicals that then constitute a significant sink of RO2 radicals and represents prevailing forest environments during nighttime.

NO3-initiated oxidation of isoprene proceeds by the NO3 addition to the carbon double bonds followed by O2 addition, yielding six distinct nitrooxy peroxy radicals (INOO), including two isomers (β-INOO) with O2 added on the β-carbon to the nitrate group (see Fig. 2b for schematic illustration). The β-INOO radicals react further with NO3, HO2, NO, and RO2, producing nitrooxy alkoxy radicals (β-INO) with molar yields of 1.00, 0.53, 0.97, and 0.40, respectively (Wennberg et al., 2018). The further decomposition of β-INO radicals produces MACR and MVK, together with HCHO and NO2. Depending on the actual fate of β-INOO radicals, the yields of β-INO radicals can then vary from 0.4 to 1.0, resulting in a distinct distribution of final oxidation products. It is thus important to specify the ultimate fate of INOO radicals during quantification of oxidation products from isoprene reaction with NO3. As an illustration, we performed one continuous-mode experiment that targets on controlling the steady-state fate of INOO radicals to be their reaction with NO and HO2 (46 and 38 % INOO radicals are predicted to react with NO and HO2, respectively, as shown in Fig. 5). Note that by adjusting the concentrations and fractions of inflow reactants (O3, NO, HCHO, and C5H8), different chemical fates and lifetimes of INOO radicals can be achieved.

Figure 5 shows the observed and predicted temporal profiles of NOx, O3, C5H8, and C4H6O over 25 h of isoprene oxidation by NO3, with continuous input of 10.2 ppb C5H8, 205 ppb O3, and 59 ppb NO into the chamber and a balancing outgoing flow at 40 Lmin-1 carrying well-mixed reactants and products. It took >16 h to reach steady state for all the species displayed. In general, the model captures the trends of O3 and NO well, while underpredicting the steady-state NO2 by ∼26 %. After ∼18 h of dark reaction, the PTR-Q-MS sampling tubing was submerged into a cold ethanol bath (-40 ± 2 ∘C) to trap artifacts in the PTR-Q-MS measured C4H7O+ (m/z 71) signal. The simulated steady-state concentration of isoprene agrees within 9 % with the measurements. The derived concentration of the sum of MACR and MVK from the measured C4H7O+ ion intensity upon cold trapping is ∼1.1 ppb, which is ∼129 % higher than the model predictions (∼0.48 ppb). This disagreement can be attributed, to a large extent, to the oversimplified representation of the six different INOO radicals as one δ-INOO isomer in the MCMv3.3.1 mechanism. As a result, the production of β-INOO radical, the important precursor of MACR and MVK, from NO3-initiated oxidation of isoprene is suppressed in the simulations. The measured molar yield of the sum of MACR and MVK is 36.3 ± 12.1 %, with uncertainties arising from the fact that 10.5 % isoprene is predicted to react with OH as an additional source of MACR and MVK. Using this value, the fraction of β-INOO over the sum of nitrooxy peroxy radicals is estimated as 48.6 ± 16.2 %, which is close to that (∼46.3 %) reported by Schwantes et al. (2015), although the estimated bimolecular lifetime of INOO radical in that study (∼30 s) is lower than that predicted in the present work (∼50 s). As discussed above, the hydroxyl peroxy radicals produced from OH oxidation of isoprene could undergo rapid interconversion through addition or removal of O2 at atmospherically relevant lifetimes. This interconversion significantly impacts the subsequent chemistry of individual ISOPOO radical isomers in terms of reaction rates and product distributions. It is likely that the INOO radicals follow similar interconversion due to the small R-OO bond dissociation energy, although no experimental evidence exists. A full examination of the INOO chemistry, i.e., their kinetic and thermodynamic properties as well as their chemical fate at different lifetimes, will be the focus of future studies using this continuous flow chamber operation method.