NO3 chemistry of wildfire emissions: a kinetic study of the gas-phase reactions of furans with the NO3 radical

Furans are emitted to the atmosphere during biomass burning from the pyrolysis of cellulose. They are one of the major contributing volatile organic compound (VOC) classes to OH and NO3 reactivity in biomass burning plumes. The major removal process of furans from the atmosphere at night is reaction with the nitrate radical, NO3. Here, we report a series of relative rate experiments in the 7300 L indoor simulation chamber at Institut de Combustion Aérothermique Réactivité et Environnement, Centre national de la recherche scientifique (ICARE-CNRS), Orléans, using a number of different reference compounds to determine NO3 reaction rate coefficients for four furans, two furanones, and pyrrole. In the case of the two furanones, this is the first time that NO3 rate coefficients have been reported. The recommended values (cm3 molec.−1 s−1) are as follows: furan, (1.49± 0.23)× 10−12; 2-methylfuran, (2.26± 0.52)× 10−11; 2,5-dimethylfuran, (1.02± 0.31)× 10−10; furfural (furan-2-aldehyde), (9.07± 2.3)× 10−14; α-angelicalactone (5-methyl-2(3H)-furanone), (3.01± 0.45)× 10−12; γ -crotonolactone (2(5H)-furanone),<1.4× 10−16; and pyrrole, (6.94± 1.9)× 10−11. The furfural+NO3 reaction rate coefficient is found to be an order of magnitude smaller than previously reported. These experiments show that for furan, alkyl-substituted furans, αangelicalactone, and pyrrole, reaction with NO3 will be the dominant removal process at night and may also contribute during the day. For γ -crotonolactone, reaction with NO3 is not an important atmospheric sink.


Introduction
Furans are five-membered aromatic cyclic ethers. Furans (and pyrroles -where N replaces O as the heteroatom) are generated during the pyrolysis of cellulose and are a major component of emissions from wildfire burning (Hatch et al., 2015(Hatch et al., , 2017Koss et al., 2018;Coggon et al., 2019;Andreae, 2019). Such emissions are likely to increase in the future, with the spatial extent, number, and severity of wildfires having increased markedly at the global scale in recent decades (Jolly et al., 2015;Harvey, 2016); this is predicted to continue as the climate warms (Krikken et al., 2021;Lohmander, fugitive emissions of these compounds during distribution as well as to emissions of unburned and partially oxidised products from vehicle exhaust. The oxidation of certain furan compounds has been shown to have large secondary organic aerosol yields (Hatch et al., 2017;Hartikainen et al., 2018;Joo et al., 2019;Ahern et al., 2019;Akherati et al., 2020), which could adversely impact air quality.
Oxidation of furans in the atmosphere has been shown to produce 2-furanones (monounsaturated five-membered cyclic esters) both via OH (notably hydroxy-furan-2-ones; Aschmann et al., 2014) and NO 3 (Berndt et al., 1997) reactions. Furan-2-ones are also produced from the OH oxidation of six-membered aromatic compounds (Smith et al., 1998(Smith et al., , 1999Hamilton et al., 2005;Bloss et al., 2005;Wyche et al., 2009;Huang et al., 2014). In both cases, the initial product is thought to be an unsaturated dicarbonyl, with production of the 2-furanone formed via photoisomerisation of the dicarbonyl to a ketene-enol (Newland et al., 2019), followed by ring closure of this molecule. In the case of aromatics, the ketene-enol can also be formed directly via decomposition of the bicyclic peroxy radical intermediate (Wang et al., 2020).
Furan-type compounds are removed from the atmosphere by reaction with the major oxidants OH, NO 3 , and O 3 . There have been a number of studies on the rates of reaction of furan-type compounds with the dominant daytime oxidant, OH (Lee and Tang, 1982;Atkinson et al., 1983;Wine and Thompson, 1984;Bierbach et al., 1992Bierbach et al., , 1994Bierbach et al., , 1995Aschmann et al., 2011;Ausmeel et al., 2017;Whelan et al., 2020). However, there have been fewer studies on the rates of reaction of furan-type compounds with the major nighttime oxidant, NO 3 (Atkinson et al., 1985;Kind et al., 1996;Cabañas et al., 2004;Colmenar et al., 2012).
The nitrate radical, NO 3 , is produced in the atmosphere, predominantly through the reaction of NO 2 with O 3 , and exists in equilibrium with N 2 O 5 . It has long been known to be an important night-time oxidant (Levy, 1972;Winer et al., 1984). While it is also produced during the daytime, it is rapidly converted back to NO 2 by reaction with NO and by photolysis. However, in environments with low NO, either due to low NO x emissions or suppression through high O 3 concentrations (e.g. , NO 3 oxidation has been observed to be significant during the day (Hamilton et al., 2021).

The CSA chamber
The ICARE-CNRS indoor chamber is a 7300 L indoor simulation chamber used for studying reaction kinetics and mechanisms under atmospheric boundary layer conditions. Further details of the chamber set-up and instrumentation are available elsewhere (Zhou et al., 2017). Experiments were performed in the dark at atmospheric pressure (ca. 1000 mbar), with the chamber operated at a slight overpressure to compensate for removal of air for sampling and to prevent ingress of outside air to the chamber. The chamber is in a climate-controlled room, and the temperature was maintained at 299 ± 2 K.

Experimental approach
Starting with the chamber filled with clean air, the volatile organic compounds (VOCs) of interest (ca. 3 ppmv) were added, followed by ∼ 1 Torr of the inert gas SF 6 to monitor the chamber dilution rate. A flow of 5 L min −1 of purified air was continuously added throughout the experiment, and air was then removed from the chamber to maintain a constant pressure (this is a slight overpressure to prevent possible ingress of air from outside the chamber). The chamber was left for at least 30 min prior to the start of the experiment to monitor the dilution rate and losses of the VOCs to the chamber walls. These losses, (1-8) × 10 −6 s −1 , were always smaller than dilution (∼ 1.2 × 10 −5 s −1 ). The reaction was then initiated by continuously introducing an N 2 O 5 sample, held in a trap at ∼ 235 K with a part of the purified air flow (2.5-5) L min −1 directed through it, for the duration of the experiment. The chamber was monitored until most of the VOC of interest was consumed, with experiments generally taking 0.5-2 h. The experiments were performed under dry conditions (RH ≤ 1.5 %).
VOC abundance was determined by in situ Fourier transform infrared (FTIR) spectroscopy using a Nicolet 5700 coupled to a White-type multipass cell with a pathlength of 143 m. Each scan was comprised of either 30 or 60 coadditions, taking a total of 2 or 4 min respectively, depending on the expected rate of loss of the VOCs, with a spectral resolution of 0.25 cm −1 .
N 2 O 5 was synthesised by reacting NO 2 with excess O 3 . First, NO and O 3 were mixed to generate NO 2 (Reaction R1). This NO 2 /O 3 mixture was then flushed into a bulb in which NO 3 and subsequently N 2 O 5 were generated through Reactions (R2)-(R3).

Analysis
VOC concentrations were monitored by FTIR. The furans generally have a number of major absorption bands in the infrared. The main bands used for analysis are shown in Table 1 (bold) along with other characteristic bands for each compound. Reference spectra of the major bands for each compound taken in the chamber at a resolution of 0.25 cm −1 are provided in the Supplement (Figs. S8-S14). The ANIR curve-fitting software (Ródenas, 2018), which implements a least squares fitting algorithm, was used to generate time profiles for each compound based on their reference spectra. Profiles were checked by doing a number of manual subtractions. Example time profiles from an experiment with α-angelicalactone and furan, with cyclohexene as the reference compound, are shown in Fig. 1. Further example plots are provided in the Supplement (Figs. S1-S7). All of the concentration-time profiles are provided in .txt format at https://doi.org/10.5281/zenodo.5721518, and all of the raw FTIR output is provided in .csv format at https://doi.org/10. 5281/zenodo.5721518. Relative rate plots for all of the experiments are shown in Fig. 2.
Relative rate experiments were performed, whereby a compound (or two) with an unknown reaction rate coefficient (k VOC ) with NO 3 was added to the chamber with a reference compound with a known NO 3 reaction rate coefficient (k ref ). A plot of the relative loss of the compound against the reference compound following addition of NO 3 (via N 2 O 5 decomposition), accounting for both chamber dilution and wall losses (k d ), gives a gradient of k VOC /k ref (Eq. 1).
A number of reference compounds were used for each VOC; they were chosen so that the reference rate coefficient was roughly within a factor of 5 of the expected unknown rate coefficient, and an attempt was made to use different references that had both larger and smaller NO 3 reaction rate coefficients than the VOC. Rate coefficients of the reference compounds (Table 2) are taken from the Database for the Kinetics of the Gas-Phase Atmospheric Reactions of Organic Compounds v2.1.0 (McGillen et al., 2020), available at https://data.eurochamp.org/data-access/kin/#/home (last access: 27 January 2022). N 2 O 5 was not present at detectable levels (by FTIR) during most of the experiments. The only experiments in which N 2 O 5 concentrations built up in the chamber were those with the slowest reacting VOCs (i.e. furfural and γcrotonolactone). NO 2 concentrations increased throughout all experiments, typically up to 2-3 ppmv. NO 2 is initially produced from the decomposition of N 2 O 5 and, later, potentially by the loss of NO 2 from nitrated VOCs/nitrated radicals. HNO 3 concentrations increased throughout the experiments, typically up to 3-4 ppmv. This could be caused by either impurities in the N 2 O 5 sample or H-abstraction reactions of NO 3 . It is not thought that this level of HNO 3 will cause any interference in the rate coefficient determinations.
It is noted that no OH scavenger was used in these experiments (as is the case for most, if not all, previous NO 3 relative rate studies to the authors' knowledge). NO 3 reaction with alkenes tends to proceed by electrophilic addition to the double bond followed by addition of O 2 to the resulting radical, leading to a nitrooxy peroxy radical (β-ONO 2 -RO 2 ) (Barnes et al., 1989;Hjorth et al., 1990). It has recently been shown (Novelli et al., 2021) that there is the possibility of OH formation through the reactions of β-ONO 2 -RO 2 with HO 2 . HO 2 could be generated in these experiments from the abstraction of an H atom by O 2 from a β-ONO 2 -RO radical with available H atoms. The initial NO 3 reaction with furans is not thought to form β-ONO 2 -RO 2 radicals, with NO 3 addition to the C 2 carbon followed by O 2 addition to the C 5 carbon (Berndt et al., 1997), analogous to the OH-addition reaction (Bierbach et al., 1995;Mousavipour et al., 2009;Yuan et al., 2017;Whelan et al., 2020). However, some of the reference compounds used in the experiments will form such radicals. For example, the reaction of HO 2 with the β-ONO 2 -RO 2 radicals formed from α-pinene + NO 3 has been reported to have an OH yield of up to 70 % (Kurtén et al., 2017). An additional minor source of HO 2 during the experiments will be H-abstraction reactions by NO 3 . These will produce RO 2 that can react to form RO radicals which may yield HO 2 following abstraction of an H atom by O 2 . However, the rate coefficient of H abstraction by NO 3 is generally expected to be negligible relative to that of the NO 3 -addition pathway. A box model run was performed to test the impact of this chemistry in this study. The α-pinene scheme from the Master Chemical Mechanism version v3.3.1 (MCMv3.3.1; Jenkin et al., 1997; http://mcm.york.ac.uk, last access: 27 January 2022) was incorporated into the AtChem box model (Sommariva et al., 2020), and an OH yield of 0.5 was assigned to the reaction of HO 2 with the initial β-ONO 2 -RO 2 radicals formed from the α-pinene+NO 3 reaction. The model was initiated with   2,3-dimethyl-2-butene (5.70 ± 1.71) × 10 −11 2-carene (2.0 ± 0.3) × 10 −11 α-pinene (6.20 ± 1.55) × 10 −12 camphene (6.60 ± 1.65) × 10 −13 cyclohexene (5.60 ± 0.84) × 10 −13 3-methyl-3-buten-1-ol (2.60 ± 0.78) × 10 −13 cyclohexane (1.35 ± 0.20) × 10 −16 2-methylfuran and α-pinene concentrations of 3 ppmv, representative of the experiments performed here. NO 3 concentrations were constrained to give a lifetime of ∼ 1 h for the VOCs, typical of the experiments. OH reaction was found to account for less than 1 % of the removal of 2-methylfuran or α-pinene through the model run. Consequently, it can be assumed that OH chemistry is a negligible interference in these experiments.
A further potential interference with the current experimental set-up is the reaction of NO 2 with the compounds used. Rate coefficients have been measured for reaction of NO 2 with a number of unsaturated compounds (Atkinson et al., 1984b;Bernard et al., 2013). For conjugated dienes, these values can be large enough (∼ 10 −18 cm 3 molec. −1 s −1 ) to provide a significant loss under the experimental conditions employed here; NO 2 is formed during these experiments Figure 2. Relative rate plots for (a) furan relative to cyclohexene (red), camphene (blue), and α-pinene (pink); (b) 2-methylfuran relative to 2-carene (blue), 2,3-dimethyl-2-butene (red), and α-pinene (pink); (c) 2,5-dimethylfuran relative to 2-carene (blue), 2,3-dimethyl-2-butene (red), 2-methylfuran (black), and pyrrole (green); (d) furfural relative to camphene (blue), cyclohexene (red), furan (black), and 3-methyl-3buten-1-ol (pink); (e) pyrrole relative to 2-carene (blue), 2,3-dimethyl-2-butene (red), 2-methylfuran (black), and 2,5-dimethylfuran (green); and (f) α-angelicalactone relative to cyclohexene (red), furan (black), and α-pinene (pink). Different shapes are used for different experiments with the same reference compound. from the decomposition of N 2 O 5 , with the NO 2 mixing ratio typically increasing up to roughly 3 ppmv through the experiment. Separate experiments were performed to look at the potential reaction of NO 2 with furan, 2,5-dimethylfuran, and pyrrole. The experiments were performed with initial VOC mixing ratios of 3 ppmv and initial NO 2 mixing ratios of roughly 5 ppmv, similar to the maximum amount of NO 2 observed during the NO 3 experiments. For all three compounds, their loss in the presence of NO 2 (allowing for dilution) was indistinguishable from zero, allowing an upper limit of <2 × 10 −20 cm 3 molec. −1 s −1 to be placed on their k(NO 2 ) rate coefficients. Based on these experiments, it was assumed that the k(NO 2 ) rate coefficients for 2-methylfuran, furfural, and α-angelicalactone are likely to be of a similar magnitude and, hence, provide negligible interference under the experimental conditions employed.

Results and discussion
The k(NO 3 ) rate coefficients determined with each reference compound are given in Table 3 and Fig. 3. Overall recommended values for the rate coefficient for each compound are calculated by taking the mean (weighted by the reported uncertainty of the reference) of the rate coefficient derived from each experiment with each reference compound, including using the recommended values for the other furans presented in Table 3. Uncertainties for the relative rates in Table S1 are assumed to be <10 % and to be dominated by statistical errors in fitting to the absorption bands. Uncertainties for the rate coefficients reported in Table 3 2020) is 150 %; however, based on the fact that the rate coefficients derived using 2,3-dimethyl-2-butene for 2-methylfuran, 2,5-dimethylfuran, and pyrrole agree very well with those using other references with much smaller uncertainties, a conservative estimate of 30 % is used here. It is noted that the rate coefficients derived with different references agree very well, to within 10 %, for all compounds. The experimentally determined k(NO 3 ) rate coefficients of the furans relative to each other are in good agreement (to within 6 %) with those calculated using the weighted means shown in Table 3 (Table S2). This gives further confidence in the k(NO 3 ) values used for the reference compounds.
The recommended rate coefficients from this work are compared to those previously reported in the literature in Table 4. The rate coefficient derived for furan agrees well with the value previously reported by Atkinson et al. (1985) from a chamber relative rate experiment. However, there is significant differences between the values reported here for furan, 2-methylfuran, and 2,5-dimethylfuran, and those reported by Kind et al. (1996) from relative rate experiments in a flow reactor. While the value reported for 2-methylfuran agrees within the uncertainties between the two studies, the values for furan and 2,5-dimethylfuran reported here are ∼ 50 % and 100 % greater respectively. It is unclear what is behind this observed disparity; the good agreement between the two studies for the 2-methylfuran rate coefficient suggests that there is not a systematic difference between the experimental set-ups. For pyrrole, the rate coefficient determined here is about 50 % faster than the value reported by Atkinson et al. (1985) from a chamber relative rate experiment using N 2 O 5 thermal decomposition. Cabañas et al. (2004) reported an upper limit of <1.8 × 10 −10 cm 3 molec. −1 s −1 (298 K) using an absolute technique of fast flow discharge.
For 2-furanaldehyde (furfural) + NO 3 , the rate coefficient recommended here is an order of magnitude slower than the only previously reported values (Colmenar et al., 2012), derived from small chamber relative rate experiments with 2methyl-2-butene and α-pinene as references. The rate coefficient from Colmenar et al. (2012) is very similar to the reported rate coefficient for furan + NO 3 . This is surprising, as the presence of a formyl group attached to a double bond is expected to be strongly deactivating with respect to addition to that bond, due to the electron withdrawing mesomeric effect of the −C(O)H group (Kerdouci et al., 2014). This has also been observed for other electrophilic addition reactions, such as those with OH and O 3 (Kwok and Atkinson, 1995;McGillen et al., 2011;Jenkin et al., 2020). Furthermore, while there is the possibility of H abstraction from the formyl group, which would increase the overall rate coefficient, such reactions are typically of the order of 10 −14 cm 3 s −1 (Kerdouci et al., 2014); hence, they would not be expected to compensate for the reduction in the contribution to the overall rate coefficient of the addition reaction.
For 5-methyl-(3H)-furan-2-one (αangelicalactone) + NO 3 , this is the first reported rate coefficient. For (5H)-furan-2-one (γ -crotonolactone), relative rate experiments with several reference compounds were attempted, with the slowest reacting of these being cyclohexane (k NO 3 = 1.4 × 10 −16 cm 3 molec. −1 s −1 ). In this experiment, roughly 10 % of the cyclohexane was removed by reaction with NO 3 (accounting for loss by dilution), whereas there was no appreciable chemical loss of γ -crotonolactone. Therefore, we can deduce that k(γcrotonolactone + NO 3 ) 1.4 × 10 −16 cm 3 molec. −1 s −1 . Again, this is the first time a NO 3 reaction rate coefficient has been measured for this compound. A comparison of the two furanones shows that 5-methyl-(3H)-furan-2one reacts more than 4 orders of magnitude faster than (5H)-furan-2-one. This can be explained in part by the presence of a methyl group, which is seen to increase the rate coefficient by roughly an order of magnitude from e.g. furan to 2-methylfuran to 2,5-dimethylfuran. Berndt et al. (1997) derived an NO 3 reaction rate coefficient of 1.76 × 10 −13 cm 3 molec. −1 s −1 for (3H)-furan-2-one. However, the majority of the difference must be explained by the structure of the two compounds, namely the conjugated nature of the C = C and C = O bonds in (5H)-furan-2-one. The carbonyl group removes electron density from the C = C bond, greatly reducing the rate coefficient. A similar relationship is seen for analogous acyclic compounds e.g. the NO 3 rate coefficient of the conjugated ester methyl acrylate is almost 2 orders of magnitude greater than that of the non-conjugated isomer vinyl acetate.
the γ -crotonolactone + NO 3 reaction suggests that this will not be an important atmospheric sink. γ -crotonolactone has also been shown to have a very slow reaction with O 3 (lifetime >100 years; Ausmeel et al., 2017), whereas the lifetime is much shorter for reaction with OH, and this will be the predominant gas-phase sink for γ -crotonolactone . Such a slow NO 3 reaction might be expected to extend to all 2-furanones with a conjugated structure, e.g. hydroxyfuranones -major products of OH oxidation of methyl-substituted furans (Aschmann et al., 2014), such that the nitrate reaction may be unimportant in the atmosphere for these structures. However substitution at the double bond is likely to increase the rate coefficient somewhat, as observed for OH and O 3 reactions with the methyl-substituted form of γ -crotonolactone (Ausmeel et al., 2017). One of the major sources of furan-type compounds to the atmosphere is wildfires. Wildfire plumes can even be regions of high NO 3 during the day due to suppressed photolysis rates in optically thick plumes (Decker et al., 2021). NO 3 oxidation of furans may be even more important in such plumes than in the background atmosphere. Such plumes can extend over hundreds of kilometres and, hence, affect air quality on a local and regional scale (e.g. Andreae et al., 1988;Brocchi et al., 2018;Johnson et al., 2021). Domestic wood burning is an increasing trend in northern European cities (Chafe et al., 2015). Burning generally occurs in the winter, during which time (with short daylight hours and peak daytime OH often an order of magnitude lower than during the summer) the reaction with NO 3 is likely to be the dominant fate of furantype compounds in such emissions, contributing significantly to organic aerosol in urban areas (Kodros et al., 2020). Berndt et al. (1997) identified the major first-generation products of furan + NO 3 to be the unsaturated dicarbonyl, butenedial, and 2(3H)-furanone, with the NO 3 recycled back to NO 2 . However, Tapia et al. (2011) andJoo et al. (2019) found that the major products of the 3-methylfuran + NO 3 reaction predominantly retain the NO 3 functionality. In this case, furan + NO 3 oxidation chemistry may be a significant sink for NO x , sequestering it in nitrate species; these nitrate species may then release the NO x far from the source on further gas-phase oxidation or (due to their low volatility) be taken up into aerosol (Joo et al., 2019).

Conclusions
Rate coefficients are recommended for reaction of seven furan-type VOCs with NO 3 at 298 K and 760 Torr, based on a series of relative rate experiments. These new recommendations highlight the importance of NO 3 chemistry to the removal of furans and other similar VOCs under atmospheric conditions. The measured rate coefficients suggest that for the three furans reported here, as well as for pyrrole and α-angelicalactone, reaction with NO 3 is likely to be their dominant night-time sink. For the alkyl furans and pyrrole, reaction with NO 3 may also be a significant sink during the daytime. This work also extends the existing database of VOC + NO 3 reactions, providing valuable reference values for future work. Data availability. Further example plots and experiment information are provided in the Supplement. All of the response-time profiles from the FTIR are provided in .txt format at https://doi.org/10. 5281/zenodo.5724967 (Newland, 2021a), and all of the raw FTIR output is provided in .csv format at https://doi.org/10.5281/zenodo. 5721518 (Newland, 2021b).
Author contributions. MJN performed the experiments with technical support from YR and MRM; MJN was also responsible for performing the data treatment, data interpretation, and writing the paper. All co-authors revised the content of the original manuscript and approved the final version of the paper.
Competing interests. The contact author has declared that neither they nor their co-authors have any competing interests.

Disclaimer.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Special issue statement.
This article is part of the special issue "Simulation chambers as tools in atmospheric research (AMT/ACP/GMD inter-journal SI)". It is not associated with a conference.