Chemical loss processes of isocyanic acid, HNCO, in the 1 atmosphere 2

The impact of chemical loss processes of isocyanic acid was studied by a combined theoretical and modeling 13 study. The potential energy surfaces of the reactions of HNCO with OH and NO 3 radicals, Cl atoms, and ozone, 14 were studied using high-level CCSD(T)/CBS(DTQ)//M06-2X/aug-cc-pVTZ quantum chemical methodologies, 15 followed by TST theoretical kinetic predictions of the rate coefficients at temperatures of 200-3000K. It was 16 found that the reactions are all slow in atmospheric conditions, with k(300K) ≤ 7  10 -16 cm 3 molecule -1 s -1 ; the 17 predictions are in good agreement with earlier experimental work, where available. The reverse reactions of 18 NCO radicals, of importance mostly in combustion, were also examined briefly. The global model confirms that 19 gas phase chemical loss of HNCO is a negligible process, contributing less than 1%. Removal of HNCO by 20 clouds and precipitation is a larger sink, contributing for about 10% of the total loss, while globally dry 21 deposition is the main sink, accounting for ~90%. The global simulation also shows that due to its long chemical 22 lifetime in the free troposphere, HNCO can be efficiently transported into the UTLS by deep convection events. 23 Average daily concentrations of HNCO are found to rarely exceed levels considered potentially toxic, though 24 locally instantaneous toxic levels are expected. mol -1 below the reactants, 24 followed by H 2 O +  N=C=O, at 7.5 kcal mol -1 exoergicity. Despite the higher energy of the products, we predict 25 this latter reaction to have a lower barrier, 6 kcal mol -1 , compared to the addition process, 9 kcal mol -1 , in 26 to take into that this analysis limited computational output available in this study, which has only daily Therefore, it is expected that exceed daily averages of 1 ppbv are potentially areas in which peak HNCO can be observed above 10 ppbv throughout the day. between exceeding ppbv road This 25 evident road daily 26 of 1 ppbv Road traffic activities occur on a smaller spatial scale than biomass burning events. The EMAC model used is not capable to represent, for inner city road traffic activities, due to the spatial resolution of the model used (1.875 by 1.875 degrees in latitude and longitude). Therefore, we are not to draw any conclusion if 10 ppbv is exceeded regionally in densely populated areas, impacted by high 30 traffic emissions. phase lifetimes of decades. Yearly loss of HNCO towards these reactants 18 is ~5 Gg/y. Removal of HNCO by clouds and precipitation (“scavenging”), leading to hydrolysis to ammonia, is 19 also implemented in the global model, and was found to contribute significantly more, ~300 Gg/y, than the gas 20 phase loss processes. Still, these combined processes are overwhelmed by the loss of HNCO by dry deposition, 21 removing ~2700 Gg/y. These conclusions are robust against modifications of the emission scenarios, where two 22 distinct sets of emission factors were used, incorporating HNCO formation from biomass burning, as well as 23 anthropogenic sources such as formamide oxidation and road traffic. The inefficiency of gas-phase chemical loss processes confirms earlier assumptions; inclusion of the gas-phase chemical loss processes in kinetic 25 models appears superfluous except in specific experimental conditions with very high co-reactant 26 concentrations. The long gas-phase chemical lifetime (multiple decades) allows HNCO to be transported 27 efficiently into the UTLS demonstrating that surface emissions may impact the upper troposphere. Further research is necessary to identify the importance of strong biomass burning events coupled to strong vertical 29 transport processes (i.e. monsoon systems) on the chemical composition of the UTLS. exceedances are located in daily concentrations of order of 1 ppbv are encountered more frequently, with about 1/3th year exceeding limit. local concentrations might peak e.g. in urban 34 environments where road are or downwind plume of biomass events, regional air quality. regional effects studied in the current work, as the resolution of global model used is fine-grained.


Introduction 26
The existence of isocyanic acid (HNCO) in the atmosphere has been established only recently (Roberts et al., 27 2011;Wentzell et al., 2013) despite being first recognized in the 19 th century (Liebig and Wöhler, 1830). HNCO 28 can form H-bonded clusters (Zabardasti et al., 2009(Zabardasti et al., , 2010Zabardasti and Solimannejad, 2007), and in 29 concentrated/pure form appreciably polymerizes to other species, but becomes fairly stable in the presence of 30 impurities (Belson and Strachan, 1982), such that the monomer is the near-exclusive representative especially in 31 the gaseous phase under ambient temperature conditions (Roberts et al., 2010). The background ambient mixing 32 ratios of HNCO vary in the range of a few pptv to tens of pptv (Young et al., 2012), while in urban regions, 33 HNCO mixing ratio increases from tens of pptv to hundreds of pptv (Roberts et al., 2014;Wentzell et al., 2013). Off-road fossil fuel activities (e.g., tar sands) also contribute to significant HNCO emissions on regional scales 27 (Liggio et al., 2017). Finally, secondary HNCO formation in the atmosphere is also known through the oxidation 28 of amines and amides [e.g., (Borduas et  The number of studies examining HNCO gas-phase chemistry is very limited, and the scarce data 30 suggests that HNCO destruction in the atmosphere by typical pathways such as reactions with oxidizing agents 31 or by photolysis is ineffective. We give a short overview here, to supplement a recent review (Leslie et al., 1992; Tsang, 1992;Tully et al., 1989;Wooldridge et al., 1996), where the extrapolated rate expressions lead to a 35 very low estimated rate coefficient of 5-12 × 10 -16 cm 3 molecule -1 s -1 at 298 K, i.e. a HNCO-lifetime towards OH 36 of over 20 years when assuming a typical OH concentration of 1 × 10 6 molecule cm -3 . Early theoretical work by 37 Sengupta and Nguyen (1997) at temperatures  500 K showed that the mechanism proceeds predominantly by 38 uncertain, and can only be determined by measuring HNCO absorption cross-sections directly for the relevant 11 tropospheric wavelengths as suggested by Young et al. (2012). HNCO has absorption bands in the infra-red 12 (Sharpe et al., 2004) but at these wavelengths generally the photon energy is too limited for photo-dissociation 13 (Hofzumahaus et al., 2002). The main atmospheric loss processes are considered to be the transition to the 14 liquid-phase via hydrolysis, and deposition. This process depends on the varying atmospheric liquid water 15 contents, relevant temperatures, and pH of cloud droplets. Therefore, the gas-to-liquid partitioning, in the 16 varying atmospheric properties, i.e. water content, temperature, and pH of cloud droplets become important Liu, 2019), the lifetime of HNCO due to heterogeneous processes is known to be of the order of a few hours (in-21 cloud reactions) to weeks (aerosol deposition). 22 The emissions and sources of HNCO have been focused on by many past studies, but there remain large 23 uncertainties in our understanding of HNCO removal process, especially in gas-phase chemistry. The limited 24 number of available studies suggests that the (photo)chemical HNCO loss processes in the atmosphere appear to 25 be slow, with liquid-phase processes acting as the dominant sink. To alleviate the dearth of direct data, we 26 provide in this work a theoretical analysis of the chemical reactions of HNCO with the dominant atmospheric 27 oxidants: OH and NO 3 radicals, Cl atoms, and O 3 molecules, predicting the rate coefficients of these reactions at 28 atmospheric conditions. The results are included in a global numerical chemistry and climate model to assess the 29 impact of chemical loss of HNCO in competition against hydrolysis within cloud droplets and against deposition 30 to the Earth's surface. The model is also used to provide an estimate of the relative importance of primary and 31 secondary HNCO sources. 32 2 Methodologies 33

Theoretical methodologies 34
The potential energy surfaces of the initiation reactions of all four reaction systems were characterized at the 35 M06-2X/aug-cc-pVTZ level of theory (Dunning, 1989;Zhao and Truhlar, 2008), optimizing the geometries and 36 rovibrational characteristics of all minima and transition states. The relative energy of the critical points was 37 https://doi.org/10.5194/acp-2019-1138 Preprint. Discussion started: 3 February 2020 c Author(s) 2020. CC BY 4.0 License. further refined at the CCSD(T) level of theory in a set of single point energy calculations using a systematic 1 series of basis sets, aug-cc-pVxZ (x = D, T, Q) (Dunning, 1989;Purvis and Bartlett, 1982). These energies were 2 extrapolated to the complete basis set limit (CBS) using the aug-Schwartz6(DTQ) scheme as proposed by 3 Martin (1996). The rate coefficients were then obtained by transition state theory (Truhlar et al., 1996) in a rigid 4 rotor, harmonic oscillator approximation, applying a scaling factor of 0.971 to the vibrational wavenumbers 5

HNCO + OH 19
The reaction of isocyanic acid with OH can proceed by 4 distinct pathways: H-abstraction, or OH addition on 20 the carbon, nitrogen, or oxygen atom of HNCO; a potential energy surface is shown in Figure 1. Formation of 21 the HN=C  OOH and HN(OH)C  =O adducts through OH-addition on the oxygen or nitrogen atom is highly 22 endothermic by 20 kcal mol -1 or more, and is not competitive at any temperature. The two remaining pathways 23 are exothermic, with HN  C(=O)OH being the most stable nascent product, 20 kcal mol -1 below the reactants, 24 followed by H 2 O +  N=C=O, at 7.5 kcal mol -1 exoergicity. Despite the higher energy of the products, we predict 25 this latter reaction to have a lower barrier, 6 kcal mol -1 , compared to the addition process, 9 kcal mol -1 , in 26 agreement with the theoretical predictions of Sengupta and Nguyen (1997). Furthermore, the H-abstraction 1 process allows for faster tunneling, making this process the fastest reaction channel, while addition contributes 2 less than 0.5% of product formation at temperatures below 400K. From this data, we derive the following rate 3 coefficient expressions (see also Figure 2

11
Our predictions are in very good agreement between 624-875K, when compared with experimental data from 12 Tully et al. (1989), which served as the basis for the recommendation of Tsang (1992); our predictions 13 reproduce the rate coefficients within a factor 1.7, comparable to the experimental uncertainty of a factor 1.5 14 (see Figure 2). Likewise, our predictions agree within a factor 1.7 with the experimental determination of 15 Wooldridge et al. (1996), over the entire 620-1860 K temperature range. Our predictions overshoot the upper 16 limit estimated by Mertens et al. (1992) by a factor of up to 4 at the upper end of the temperature range (2120 to 17 2500 K). At these elevated temperatures, it is expected that our kinetic model is less accurate since 18 anharmonicity, internal rotation, and possibly pressure effects are not fully accounted for. At this time, we 19 choose not to invest the computational cost to improve the model at these temperatures. The predicted rate at 20 room temperature is within a factor of 2 of the extrapolation of the recommended expression derived by Tsang 21 (1992), k(298 K)  1.2410 -15 cm 3 molecule -1 s -1 , and very close to the extrapolation of the expression by 22 Wooldridge et al. (1996), 7.210 -16 cm 3 molecule -1 s -1 . The good agreement of our rate coefficient with the 23 experimental data extrapolated to room temperature is mainly due to the curvature predicted in the temperature-24 dependence (see Figure 2), as our calculations have a slightly steeper temperature dependence than the 25 experiments in the high-temperature range. Though negligible at low temperature, , we find that OH addition on 26 the C-atom of HNCO accounts for 7 to 8 % of the reaction rate between 2000 and 3000 K, with other non-H-27 abstraction channels remaining negligible (<0.1%). 28 Typical concentrations of the OH radical during daytime are measured at ~10 6 molecule cm -3 (Stone et al., 29 2012), leading to an pseudo-first order rate coefficient for HNCO loss by OH radicals of k(298K) = 10 -10 s -1 , i.e. 30 a chemical lifetime of several decades, negligible compared to other loss processes like scavenging. Even in 1 extremely dry conditions, where aqueous uptake is slow, heterogeneous loss processes will dominate, or 2 alternatively atmospheric mixing processes will transport HNCO to more humid environments where it will 3 hydrolyze. 4 5 6

HNCO + Cl 7
From the potential energy surface (PES) shown in Figure 1, we see that the reaction between HNCO and Cl 8 atom can occur by abstraction of the H atom from HNCO, or by addition of the Cl atom on the C-, N-or O-9 atoms. Contrary to the OH-reaction, all entrance reactions are endothermic, with formation of the HN  C(Cl)=O 10 alkoxy radical nearly energy-neutral (see Figure 1). Formation of this latter product, proceeding by the addition 11 of a Cl atom to the carbon atom of HNCO, also has the lowest energy barrier, 7 kcal mol -1 above the reactants.

28
We find that the overall rate coefficient of the HNCO + Cl reaction is almost one order of magnitude below that 29 for the OH radical. The HN  C(Cl)=O radical formed, however, has a weak C-Cl bond requiring only 5.4 kcal 30 mol -1 to redissociate. The rate coefficient of 810 8 s -1 for dissociation at room temperature (k(T) = 8.310 12 31 exp(-2760/T) s -1 ), makes redissociation to the reactants the most likely fate of the HN  C(Cl)=O adduct. Addition 1 is thus an ineffective channel for HNCO removal, and the effective reaction with Cl atoms is dominated by the 2 H-abstraction reaction, forming HCl +  NCO, with the following rate coefficient (see also Figure 3 years, which is much longer than toward the OH radial. Therefore, HNCO loss by Cl radicals is negligible. 8 The supporting information provides information on the extended potential energy surface of the HNCO + Cl 9 reaction, with information on 9 intermediates, 19 transition states, and 16 products. 10 11

HNCO + NO 3 12
The reaction of NO 3 with HNCO shows the same four radical mechanisms found for OH and Cl, i.e. H-13 abstraction and addition on the 3 heavy atoms. As for Cl-atoms, none of the reactions are exothermic, and the 14 energy difference between the two most stable products, is reduced to 3 kcal mol -1 , indicating that NO 3 addition 15 is even less favorable than Cl addition. Formation of HNO 3 +  NCO is more favorable than HCl + NCO 16 formation, by about 2 kcal mol -1 , owing to the greater stability of nitric acid. The barrier for H-abstraction, 17 however, is larger compared to abstraction by both OH and Cl, and exceeds 12 kcal mol -1 . The most favorable 18 addition process, forming HN  C(=O)NO 3 has a barrier of 15 kcal mol -1 , but contributes less than 0.01% to the 19 reaction rate at room temperature. The overall reaction thus proceeds near-exclusively by H-abstraction forming 20 HNO 3 +  NCO, for which we derived the following rate coefficients (see also Figure 3 While this rate coefficient is almost 5 orders of magnitude below that of the OH radical, the nitrate radical is 24 known to be present in higher concentrations during night time, reaching concentrations as high as 10 9 molecule 25 cm -3 (Finlayson-Pitts and Pitts, 1999). The effective rate of the NO 3 reaction at night time is similar to the 26 reaction with OH at day time. The NO 3 radical is thus still considered to be ineffective for atmospheric removal 27 of HNCO. 28

HNCO + O 3 29
The chemistry of ozone with organic compounds is drastically different from radicals, where O 3 typically reacts 30 by cycloaddition on double bonds in unsaturated compounds. For isocyanic acid, cycloaddition pathways have 31 been characterized for both double bonds (HN=C=O). Only cycloaddition on the N=C bond leads to an 32 exothermic reaction, with the oxo-ozonide product being 12 kcal mol -1 more stable than the reactants (see Figure  33 1). In addition to the traditional cycloaddition channels, three further channels were found, corresponding to H- The cyclo-addition channels on the hetero-double bonds have high energy barriers, exceeding 30 kcal mol -1 , 1 significantly larger than typical barriers for C=C bonds with aliphatic substitutions. Surprisingly, this allows H-2 abstraction to become competitive to cycloaddition, with a comparable barrier of 32 kcal mol -1 . For the overall 3 reaction, we obtain the following rate coefficients (see also Figure 3 At room temperature, H-abstraction contributes 80% to the total reaction, and cycloaddition on the N=C bond 7 the remaining 20%. All other channels are negligible. The rate coefficient is exceedingly low, ~10 -37 cm 3 8 molecule -1 s -1 , such that even in areas with very high ozone concentrations of 100 ppbv the loss by ozonolysis is 9 expected to be negligible. 10 The supporting information provides information on the extended potential energy surface of the HNCO + O 3 11 reaction, with information on 10 intermediates, 30 transition states, and 15 products. The lowest-energy 12 unimolecular product channel leads to formation of CO 2 + HNOO by breaking of the cyclic primary ozonide 13 (see Figure 1) following the traditional Criegee mechanism (Criegee, 1975). 14

Global impact 15
From our global simulations, we gain many insights on the impact of the described mechanism (Table 1  The atmospheric lifetime of HNCO is dominated by its heterogeneous loss processes, leading to an atmospheric 1 lifetime of multiple weeks, whereasthe gas-phase lifetime in the free troposphere is about 50 years. This long 2 gas-phase lifetime and the fact that mainly surface sources are relevant indicate that atmospheric HNCO is 3 highly impacted by transport processes. Our simulations show that HNCO is transported from the surface into 4 the UTLS (upper troposphere/lower stratosphere) and that about 10% of the total atmospheric HNCO mass is 5 located in the stratosphere, with modelled concentrations of HNCO in the lower stratosphere of typically tens of 6 pptv but reaching up to hundred pptv in tropical regions. Since OH is the only significant stratospheric sink, the 7 stratospheric lifetime increases to more than 330 years. During the monsoon period, the total stratospheric only a few days can be observed in which this limit is exceeded. The maximum number of days exceeding 10 20 ppbv is 10 days over Africa, compared to 120 days above 1 ppbv. It is important to take into account that this 21 analysis is limited by the computational output available in this study, which has only daily averages. Therefore, 22 it is expected that areas which frequently exceed daily averages of 1 ppbv are potentially areas in which peak 23 HNCO can be observed above 10 ppbv throughout the day. 24 No correlation exists between the number of days exceeding 1 or 10 ppbv and road traffic emissions. This 25 becomes evident since typical areas of high road traffic activities (i.e. USA and Europe) do not exceed daily 26 averages of 1 ppbv (see Figure 5). Road traffic activities occur on a smaller spatial scale than biomass burning 27 events. The EMAC model used is not capable to represent, for example, inner city road traffic activities, due to 28 the spatial resolution of the model used

H-abstraction reactions by NCO radicals 5
The radical reactions characterized above proceed by H-abstraction, forming the NCO radical with an H 2 O, 6 HNO 3 , or HCl co-product. Likewise, the ozonolysis reaction proceeds for a large part by H-abstraction, forming 7 NCO with a HO 3 coproduct that readily dissociates to OH + O 2 . Though NCO radical formation through these 8 reactions is found to be negligibly slow in atmospheric conditions, this radical remains of interest in other 9 environments. Examples include combustion chemistry, where it can be formed directly from nitrogen-10 containing fuels, and where it is a critical radical intermediate in e.g. the RAPRENOx nitrogen-oxide mitigation 11 strategy (Fenimore, 1971;Gardiner, 2000). The NCO radical has also been observed in space (Marcelino et al., 12 2018). There is extensive experimental and theoretical information of the reactions of NCO radicals, tabulated 13 e.g. in Tsang (1992) can readily abstract a hydrogen atom from most hydrogen-bearing species to produce HNCO, and that H-19 abstraction is the main reaction channel. Hence, despite that our potential energy surfaces do not include an 20 exhaustive search of the NCO radical chemistry, we expect that predictions of the H-abstraction rate for NCO 21 from H 2 O, HNO 3 and HCl are fair estimates of the total rate coefficients of these reactions. 22 The energy barriers for the NCO radical reactions with H 2 O, HNO 3 and HCl, being 14, 7, and 4 kcal mol -1 23 respectively (see Figure 1), follow the bond strength trend in these reactants, with D 0 (H-OH) = 118 kcal mol -1 , 24 D 0 (H-NO 3 ) = 104 kcal mol -1 , and D 0 (H-Cl) = 103 kcal mol -1 (Luo, 2007;Ruscic et al., 2002). of the NCO reactions in combustion or non-terrestrial environments is well outside the scope of this paper, and 7 reactions with other co-reactants not discussed in this paper are likely to be of higher importance, e.g. H-8 abstraction from organic compounds, or recombination with other radicals. In atmospheric conditions, the fate 9 of the NCO radical is likely recombination with an O 2 molecule, leaving H 2 O, HNO 3 , and HCl as negligible co-10 reactants. Hence, the NCO radical will not affect the atmospheric fate of any of these compounds to any extent. 11 Subsequent chemistry of the  OONCO radical is assumed to be conversion to an  ONCO alkoxy radical through 12 reactions with NO, HO 2 or RO 2 , followed by dissociation to NO + CO. 13

Conclusions 14
The isocyanic acid molecule, HNCO, is found to be chemically fairly unreactive towards the dominant 15 atmospheric gas phase oxidants, i.e. OH and NO 3 radicals, Cl atoms, and O 3 molecules. The reactions all occur 16 predominantly by H-abstraction, and have comparatively low rates of reactions with k(298) ≤ 710 -16 cm 3 17 molecule -1 s -1 , leading to chemical gas phase lifetimes of decades. Yearly loss of HNCO towards these reactants 18 is ~5 Gg/y. Removal of HNCO by clouds and precipitation ("scavenging"), leading to hydrolysis to ammonia, is 19 also implemented in the global model, and was found to contribute significantly more, ~300 Gg/y, than the gas 20 phase loss processes. Still, these combined processes are overwhelmed by the loss of HNCO by dry deposition, 21 removing ~2700 Gg/y. These conclusions are robust against modifications of the emission scenarios, where two 22 distinct sets of emission factors were used, incorporating HNCO formation from biomass burning, as well as 23 anthropogenic sources such as formamide oxidation and road traffic. The inefficiency of gas-phase chemical 24 loss processes confirms earlier assumptions; inclusion of the gas-phase chemical loss processes in kinetic 25 models appears superfluous except in specific experimental conditions with very high co-reactant 26 concentrations. The long gas-phase chemical lifetime (multiple decades) allows HNCO to be transported 27 efficiently into the UTLS demonstrating that surface emissions may impact the upper troposphere. Further 28 research is necessary to identify the importance of strong biomass burning events coupled to strong vertical 29 transport processes (i.e. monsoon systems) on the chemical composition of the UTLS. 30 On a global scale, the daily average concentrations of HNCO rarely exceed 10 ppbv, the threshold assumed here 31 for toxicity; the exceedances are mainly located in regions with strong biomass burning emissions. Average 32 daily concentrations of the order of 1 ppbv are encountered more frequently, with about 1/3th of the year 33 exceeding this limit. This suggests that local concentrations might peak to much higher values, e.g. in urban 34 environments where road traffic emissions are highest, or in the downwind plume of biomass burning events, 35 and could impact regional air quality. Such regional effects were not studied in the current work, as the 36 resolution of the global model used here is not sufficiently fine-grained. 37 https://doi.org/10.5194/acp-2019-1138 Preprint. Discussion started: 3 February 2020 c Author(s) 2020. CC BY 4.0 License.
Though not important for the atmosphere, we briefly examined the reactions of the NCO radical formed in the 1 chemical reactions studied. The rate coefficients of the H-abstraction reactions with H 2 O, HNO 3 and HCl 2 suggest that these reactions might contribute in high-temperature environments, such as combustion processes. 3

Supplement 4
The supplement related to this article is available online, and contains extended information on the chemical 5 model, and the quantum chemical characterizations (geometric, energetic and entropic data)