Observation of isoprene hydroxynitrates in the Southeastern United States and implications for the fate of NO x

δ-RO 2 senario. When k(+O 2 ) values for δ-RO 2 by 1.5 times and k(-O 2 ) values for β-RO 2 were increased by 1.5 times, the yield of 1,2-RO 2 and 4,3-RO 2 became 50% and 26%, and the combined MVK+MACR yield of 76% was still within the uncertainty reported by experimental studies.This set of RO 2 isomer distribution was treated as the high β-RO 2 scenario. The three set of isomer distributions, base scenario, high δ-RO 2 senario and high β-RO 2 scenario were used to estimate the uncertainty associated with the LIM1 mechanism. The total error in RO 2 distribution is 20%, calculated as the weighted relative error of each RO 2 isomer.


CIMS calibration results for 4,3-IN
Multiple CIMS calibrations for 4,3-IN have been conducted overtime, and the results were stable ( Figure S1). Figure S1. Calibration results for 4,3-IN. Each data point represents one calibration curve. The blue shade shows 1σ standard deviation of the measured sensitivity.

CIMS sensitivities for cis-and trans-1,4-IN
Three CIMS calibrations were conducted using standard solutions that contained a mixture of cisand trans-1,4-IN. The relative abundance of the cis and trans isomers was determined from NMR spectra. The overall sensitivity and relative isomer abundance are listed in Table S1. Table S1. Overall sensitivity and isomer composition for cis-and trans-1,4-IN calibration.

Calibration
Sensitivity (ppt -1 ) Isomer composition (trans relative to cis) If we assume x is the sensitivity for trans-1,4-IN and y is the sensitivity for cis-1,4-IN, the isomerweighted sensitivity measured by CIMS can be written as following.
Due to the uncertainties in the coefficients in equation (1) to (3), x and y in the above equation system cannot be solved. Therefore, we define function z through the following expression.  Figure S3 shows the changing IN sensitivity derived from the 0D model.

Isomer distribution for IN during SOAS
The 0D model was used to estimate the relative abundance of IN isomers during the SOAS study.
The diurnal average of isoprene, OH, NO, NO2, O3 and HO2 were calculated and the 0D model was used to simulate the relative concentrations of the IN isomers as they were produced from isoprene oxidation and lost to OH, O3 and deposition throughout the day. The reaction was initiated at 6:00 AM (reaction time equals 0) and the duration was 24 hours. The simulated IN isomer composition is shown in Figure S8. A diurnal isomer-weighted IN sensitivity was calculated based on the simulated IN isomer distribution ( Figure S9). The same diurnal calibration factors were applied to interpret IN raw data for each individual day.

Isoprene RO2 distribution and RO2 lifetime
During SOAS, the RO2 loss rates to NO and HO2 are slow, compared with 1,6-H shift rate constant for the cis-δ-RO2. As a result, the yield of total RO2 from OH addition to isoprene, defined as the amount of RO2 produced relative to the amount of isoprene consumed, can decrease with RO2 lifetime, as cis-δ-RO2 radicals isomerize into hydroperoxy aldehyde (HPALD) and other RO2 covert to cis-δ-RO2 through O2 loss and addition (Peeters et al., 2014). The yield of the products with respect to RO2 lifetime was calculated with the Kintecus software, and the result is shown in Figure S10. The cis-δ-RO2 radicals become less important with longer RO2 lifetime. Besides RO2 and HPALD, OH addition to isoprene also forms a stable carbonyl product, with a yield of 2% (Fan and Zhang, 2004;Peeters et al., 2014). The daytime total RO2 loss rate to NO and HO2 was on the order of 0.05 s -1 , so the RO2 yield at 20 s was chosen to calculate the IN production rate during SOAS. With an RO2 lifetime of 20 s, the isoprene oxidation products consist of 83% RO2, 15% HPALD and 2% carbonyl product. The 83% RO2 products include 1% cis-δ-RO2, 2% transδ-RO2 and 81% β-RO2. Figure S10. Product yield from OH addition to isoprene.

Sensitivity tests on LIM1 mechanism
The uncertainties in kinetics data for RO2 interconversion and 1,6-H shift can cause error in the isoprene RO2 and IN isomeric distribution. For the LIM1 mechains, the uncertainties for the equlibrium constants Keq=k(+O2)/k(-O2) are a factor of 1.5, and the uncertainties for the 1,6-H shift rate constants k1,6-H are a factor of 2.4 (Peeters et al., 2014). Sensitivity tests were performed by varying the rate constants k(+O2) (or k(-O2)) by 1.5 times and k1,6-H by 2.4 times, and calculating the relative abundance of the RO2 isomers. We found changing k1,6-H had no influence on the relative abundance of RO2 isomers, although it significantly influences the yield of HPALD.
Changing k(+O2) or k(-O2) only affected the production rate of total RO2, but had no ifluence on the relative abundance of RO2 isomers. The isomeric distribution was affected most when k(+O2) or k(-O2) were varied differently for β-RO2 and for δ-RO2. When k(+O2) values for δ-RO2 were increased by 1.5 times and k(-O2) values for β-RO2 were decreased by 1.5 times, the yield of 1,2-RO2 and 4,3-RO2 became 32% and 17% repectively, which will lead to a total MVK+MACR yield of 49%. This is significantly lower than the experimental MVK+MACR yield from isoprene high NO oxidation (Liu et al., 2013;Park et al., 2004). When k(+O2) values for δ-RO2 were increased by 1.5 times and k(-O2) values for β-RO2 were kept as orignial, the yield of 1,2-RO2 and 4,3-RO2 were 37% and 19%, more consistent with the experimental MVK+MACR yield. This set of RO2 isomer distribution was treated as the high δ-RO2 senario. When k(+O2) values for δ-RO2 were decreased by 1.5 times and k(-O2) values for β-RO2 were increased by 1.5 times, the yield of 1,2-RO2 and 4,3-RO2 became 50% and 26%, and the combined MVK+MACR yield of 76% was still within the uncertainty reported by experimental studies.This set of RO2 isomer distribution was treated as the high β-RO2 scenario. The three set of isomer distributions, base scenario, high δ-RO2 senario and high β-RO2 scenario were used to estimate the uncertainty associated with the LIM1 mechanism. The total error in RO2 distribution is 20%, calculated as the weighted relative error of each RO2 isomer.

4,3-IN sensitivity and sample humidity
The stability of the CIMS signal for 4,3-IN under different humidity was investigated with the setup in Figure S11a.
If we define rate as variable r, the concentration of IN as x and the concentration of MVK+MACR as y, we will have the following expression.
Since the concentration of IN (x) and the concentration of MVK+MACR (y) both changed with time. Equation (6) can be written as the following.
In Equation (7), dy/dt is the growth rate of MVK+MACR, and dx/dt is the growth rate of IN.
The 2-hour period from 7:00 AM to 9:00 AM for the 12-day average was chosen as the time window to estimate the contribution of downward mixing. To simplify the calculation, the average changing rate was used for IN and MVK+MACR, instead of the instantanous rate. That modifies Equation (7) to the form of Equation (8)  instead, which will make the result slightly biased high.
The calculated total [IN] growth rate with robs is 9.93×10 -3 ppt/s, which is consistent with the [IN] growth rate derived directly from IN measurement ( Figure S13d). When rmod was applied in Equation (9)