Photochemical reactions of methyl and ethyl nitrate: a dual role for alkyl nitrates in the nitrogen cycle

Additional keywords: nitric acid, OH-initiated reaction, peroxyacetyl nitrate, photolysis.

Oxides of nitrogen (NO_x) (i.e. NO₂ and NO) are usually considered as reactive nitrogen species and play an important role in atmospheric chemistry. The photolysis of NO₂ can initiate a series of reactions and is the sole known anthropogenic source of ozone in the troposphere.^[1] NO_x can transform into a variety of inorganic and organic nitrogen-containing substances in the troposphere, including peroxynitric acid (HOONO₂), dinitrogen pentoxide (N₂O₅), chlorine nitrate (ClONO₂), nitric acid (HNO₃), nitrous acid (HNO₂), aerosol nitrate (NO₃^–), peroxyacyl nitrates (RC(O)OONO₂, where R = H, CH₃, CH₃CH₂,…) and alkyl nitrates (RONO₂, ANs). These chemicals together with NO_x are collectively termed as ‘odd nitrogen’, namely, NO_y. However, the observed levels of NO_y were greater than the sum of the individually measured species (ΣNO_yi = NO, NO₂, HNO₃, NO₃^–_(aerosol) and peroxyacetyl nitrate (PAN, CH₃C(O)OONO₂)), implying missing NO_y components in the field NO_y measurement.^[2] Recently, ANs have come under increased scrutiny because the large abundances of ANs might represent the ‘missing NO_y’ in the atmosphere.^[2]

The mixing ratios of ANs lie in the range of tens to hundreds of parts per trillion by volume (pptv) in the air of polluted and remote regions. It should be noted that the ANs defined in the present study include monofunctional and multifunctional AN compounds. In polluted air masses, for example, in La Porte, Texas, the daytime mixing ratio of total straight chain alkyl nitrates (∑C₁–C₅) was found to range from 30 to 80 pptv^[3]; a long-term observation in coastal New England indicated that the median mixing ratio of ∑C₁–C₅ was 23–25 pptv in winter and 14–16 pptv in summer^[4]; the maximum mixing ratio of ∑C₁–C₅ was detected as 204 pptv in the Pearl River Delta, an industrial area of South China.^[5] In remote continental air masses, mixing ratios of the C₁–C₁₀ ANs were observed as 10 to 200 pptv.^[6–13] The measurements at a rural site in California over a full annual cycle showed that total ANs (∑ANs) were routinely 10–20 % of NO_y.^[2] Increasing evidence suggests that the biogenic volatile organic compounds, especially isoprene and terpenes, can significantly contribute to AN production, including simple and complex ANs with higher molecular weight and multifunctional structure.^[14] The measured concentration of isoprene nitrates reached ~90 pptv in a forest site at Pellston, Michigan,^[15] and as high as 150 pptv in rural Tennessee.^[16] ANs were also detected in the lower stratosphere by the NASA ER-2 aircraft, covering a latitude range of 60°N to 2°S, 115°W to 155°W and altitudes to 20.5 km.^[17] Methyl and ethyl nitrate account for ~20–40 % of the total light ANs.^[4,5,13] Methyl and ethyl nitrate have been detected in Antarctica and their sum mixing ratio accounted for more than 50 % of the 70 ppt of NO_y over Antarctica during the summer.^[6]

To date, the determinations of the concentrations of atmospheric ANs are subject to uncertainties because of limitations of handling and measurement techniques. Nonetheless, the collective pool of measurements to date point to the fact that ANs may be a significant component of NO_y in the atmosphere.

Tropospheric ANs appear to be mainly produced from the oxidation of parent hydrocarbons,^[18,19] as follows:

The branching ratio of the RONO₂ formation pathway (reaction 3b) can range from one percent to tens of percent, depending on the R group.^[20] This pathway is important because it is the major route for the formation of ANs and terminates the cycle between NO and NO₂. A possible alternative pathway for RONO₂ formation under polluted conditions has been suggested by^[21,22]:

where M represents the molecular N₂ or O₂ in air.

In a previous investigation, we established that the reaction of methyl hydroperoxide (CH₃OOH) and ethyl hydroperoxide (CH₃CH₂OOH) with NO₂ can respectively produce methyl nitrate and ethyl nitrate,^[23,24] although the rate coefficients are merely ~10^–20 cm³ molecule^–1 s^–1 at 293 K. The heterogeneous reaction of HCHO and HNO₃ on concentrated sulfuric acid aerosols has also been suggested to be a pathway for methyl nitrate formation.^[25] In addition to chemical formation, the oceanic source of ANs, particularly for methyl and ethyl nitrate, has been known for over a decade.^[7,26–28] In this case, ANs may come from the photosynthesis of sea plants, bacterial or algal processes, or from transformations involving dissolved nutrient NO₃^–.^[27] Biomass burning has been identified as a significant source of ANs.^[29] Moreover, several ANs have long been used as explosives and vasodilator drugs against angina pectoris. For example, the explosive nitroglycerin is a triester of glycerin and nitrous acid. Methyl nitrate itself is highly explosive^[30]; therefore, the production and use of ANs in this way are considered to be anthropogenic emissions.

The atmospheric fates of ANs determine their lifetimes and roles in atmospheric chemistry. Many of the simple ANs are thermally stable,^[31,32] and only slightly soluble in water, so their removal by thermal decomposition or wet deposition^[33] is expected to be minor, but their dry deposition accounts for a percentage for methyl nitrate.^[4] Photolysis and OH-initiated oxidations thus become the dominant sinks for ANs in the atmosphere.

Rebbert^[34] proposed three photolysis pathways for the photodissociation of C₂H₅ONO₂:

Several studies have suggested that the dominant pathway for photolysis of ANs is through formation of NO₂ and an alkoxy radical (reaction 6).^[33,35–39]

However, most previous photolysis investigations were carried out in a He or N₂ bath gas, neglecting the presence and importance of O₂ in the real atmosphere. The reaction of ANs with OH radicals initiates a series of successive reactions in which ANs release products into the atmosphere:

The rate coefficients of methyl and ethyl nitrate reactions with OH radicals have been determined in several studies.^[40–44] Two pathways for the reaction of ANs with OH have been proposed, namely, hydrogen-atom abstraction^[41,43,44] and OH addition.^[41] However, the specific products formed in the reaction of ANs with OH radicals are still unclear.

ANs are usually considered to be the temporary reservoir of reactive nitrogen species and they can regenerate NO_x through photolysis and oxidation.^[5,45] Because compared to NO_x they have relatively long atmospheric lifetimes, ranging from several days to weeks,^[22,46,47] ANs may be transported large distances to remote regions of the atmosphere, potentially providing a source of NO_x to these areas and influencing O₃ chemistry. Therefore, ANs play a significant role in the NO_x–O₃–HO_x cycle in both polluted and remote atmospheres. However, to date, the current knowledge of the specific products and mechanisms for the transformation of ANs in the atmosphere has depended upon the speculations derived from the theory and limited experiments performed in the absence of O₂, which thus do not simulate real atmospheric conditions. The relationship between ANs and NO_x has been simplified in the current atmospheric models, possibly leading to large discrepancies in the simulation of the reactive nitrogen cycle. Detailed laboratory studies of the products and mechanisms in AN reactions under a simulated real atmosphere containing O₂ are urgently needed to better understand their roles in air quality and nitrogen exchange between the Earth’s atmosphere and surface.

In the present study, we have selected methyl and ethyl nitrate as models of atmospheric ANs and have investigated their photochemical reactions under simulated atmospheric conditions. Our study is focussed on the specific products and mechanisms in the photolytic and OH-initiated reactions of ANs. Our experimental results have revealed a dual role for ANs in the reactive nitrogen cycle, thus providing useful information for atmospheric chemical models.

Fuming nitric acid (95 %, Beijing Lisui Chemical Reagent Factory, Beijing, China), methanol (≥99.9 %, Fisher Chemical, Fair Lawn, NJ, USA), ethanol (≥99.9 %, Beijing Chemical Reagent Factory) N₂ (≥99.999 %, Beijing Pryx Applied Gas Company Limited, Beijing, China) and O₂ (≥99.999 %, Beijing Analytical Instrument Factory, Beijing, China) were used in this study. As methyl nitrate and ethyl nitrate are not commercially available, they were synthesised in our laboratory by the esterification of the corresponding alcohol and nitric acid with dehydration by sulfuric acid, and in the presence of a grain of urea, according to the procedures described by Desseigne et al.^[48] The purities of the methyl and ethyl nitrate obtained were confirmed to be ≥95 % by checking for possible impurities such as NO₂, CH₃ONO and C₂H₅ONO, etc. using Fourier-transform infrared (FTIR) spectroscopy.

Laboratory simulations were carried out in a 28.5-L cylindrical quartz reaction chamber (Q-6-54X2-BA-AU model, Infrared Analysis Inc., Anaheim, CA, USA). This chamber was equipped with a digital thermometer, a piezometer to measure the gas pressure, a vacuum system and a White-mirror system. The equipment has been described in detail in our previous work.^[49] A FTIR spectrometer (Nexus model, Thermo Nicolet, USA) was used to characterise the reactants and some of the products. There were five UV lamps (λ_max = 254 nm, 40 W, low-pressure mercury lamp, Beijing Haidian Konghou Complex Factory, Beijing, China) and a stainless steel enclosure around the reactor. The FTIR facility was mounted with a KBr beam splitter and a liquid nitrogen-cooled mercury–cadmium–telluride (MCT) detector. FTIR spectra were recorded by co-adding 64 scans at 1 cm^–1 spectroscopic resolution over the frequency region of 500–4000 cm^–1. The reaction temperature was controlled at 298 ± 2 K and the total pressure of the gas mixture was controlled at ~1.01 × 10⁵ ± 1333 Pa (760 ± 10 Torr).

For the offline analysis of products by high performance liquid chromatography (HPLC) and ion chromatography (IC), gas (1.2 L) was drawn from the reactor by a pump and was collected with a Horibe tube in a cold trap of ethanol–liquid nitrogen at ~–90 °C. After the collection was completed, the sample in the Horibe tube was quickly eluted by acetonitrile (≥99.9 %, Tedia, Fairfield, OH, USA), and the eluate was immediately mixed with a 2,4-dinitrophenyhydrazine (DNPH, 97 %, DryWT, Alfa Aesar, Ward Hill, MA, USA) solution in acetonitrile to derivatise the carbonyl compounds. The resulting sample solution was kept in the dark for ~12 h at room temperature, followed by analysis with HPLC. Similarly, a H₃PO₄ (1 mM) solution or pure water (MilliQ system, Millipore, Billerica, MA, USA) was substituted for the acetonitrile to trap the peroxides or inorganic and organic acids, and these two types of compounds were then immediately determined by HPLC or IC. Carbonyl compounds were analysed as their DNPH derivatives by HPLC (Agilent 1100, Waldbronn Vogel, Germany). For a 1.2-L sample of gas, the detection limit for formaldehyde and acetaldehyde using the DNPH method is ~0.05 ppmv (parts per million by volume). Peroxides were analysed using a post-column derivatisation method by using another HPLC instrument (Agilent 1100), in which the peroxides oxidised hydroxyphenylacetic acid to produce a fluorescent dimer in a reaction catalysed by hemin. Acids were analysed using an IC instrument with an ED₅₀ conductivity detector (ICS-2500 model, Dionex Corporation, Sunnyvale, CA, USA). The details of the sample processing and analysis are described in our previous work.^[50–52] Blanks were sampled and analysed using the same method before the reaction.

In order to reduce interference from the reactions occurring on the chamber wall, the quartz reaction chamber was deactivated with OH radicals (the generation method is described in the OH-initiated oxidation section below) for more than 12 h beforehand. To investigate the wall effect, attenuation experiments of methyl nitrate and ethyl nitrate in pure N₂ were performed in the chamber for a minimum of 12 h. The results of this experiment indicated that the wall loss of these two nitrates was negligible (i.e. <5 % loss over a 12-h period).

The reaction chamber was cleaned with N₂ and evacuated four to five times before each experiment. Methyl nitrate (or ethyl nitrate) was then introduced into the evacuated reactor. After it had been evaporated completely, high-purity nitrogen or synthetic air (O₂ : N₂ = 1 : 4) was added to reach a total pressure of ~1.01 × 10⁵ Pa, resulting in the initial gaseous concentrations of 60–210 ppmv for methyl nitrate and 50–160 ppmv for ethyl nitrate. These values, which were approximately six orders of magnitude higher than those in the atmosphere, were used to better characterise the possible products formed. After the gases were homogenously mixed, the UV lamps were turned on and in-situ FTIR spectra were recorded to investigate the concentration changes of the reactants and products. The FTIR spectrometer collected a spectrum every 5 min during the reaction, allowing us to record reactant and product concentrations. The gaseous samples were drawn out of the reaction chamber and collected by a Horibe tube during the reaction, followed by HPLC and IC analysis. Two samplings were usually done at the reaction times of 40 and 80 min during each experiment and a 1.2-L sample of gas was drawn out for each sample. After each sampling process, N₂ was quickly added into the reactor to keep the total pressure at ~1.01 × 10⁵ Pa, resulting in a slight dilution of the reactants and products. In order to obtain a uniform plot of the temporal evolution of reactants and products, the gas concentrations used in the plot, after each sampling process, were corrected by multiplying the corresponding concentrations measured in the reactor by the dilution ratio of 28.5/(28.5–1.2).

For this experiment, OH radicals were generated by UV photolysis of O₃ in the presence of water vapour:

The reactor was cleaned with N₂ and evacuated four to five times before the experiment. Methyl nitrate (or ethyl nitrate) and water were introduced into the evacuated reactor in turn. After both had been evaporated completely, ozone in O₂ gas, which was generated in a 5-L quartz cell beforehand, was added into the reactor, followed by addition of high-purity nitrogen to reach a total pressure of ~1.01 × 10⁵ Pa of synthetic air (O₂ : N₂ = 1 : 4). The initial concentrations of methyl nitrate and ethyl nitrate were the same as in the photolysis experiment described above. The initial concentration of O₃ in the reactor was ~45 ppmv. The initial relative humidity in the reaction chamber was ~30 %. After the gases were homogenously mixed, the UV lamps were turned on, immediately generating OH radicals. The methods for determining the reactants and products were the same as in the photolysis experiment described above.

To estimate the actual yield of HNO₃ in the OH-initiated reactions, we experimentally determined the loss rate of gaseous HNO₃ in the reactor using the same conditions, including the humidity levels and UV irradiation, as used for the OH-initiated oxidation of CH₃ONO₂ or C₂H₅ONO₂. In the HNO₃ loss experiment, liquid fuming HNO₃ was first introduced into the evacuated reactor and then liquid H₂O was introduced. After the liquid H₂O was evaporated completely, synthetic air was added to attain a total pressure of ~1.01 × 10⁵ Pa, resulting in a relative humidity of 30 % and the gaseous HNO₃ concentrations were within 40–120 ppmv in the reactor. The HNO₃ loss was monitored by FTIR spectroscopy at intervals of 5 min.

In the standard FTIR spectrum of CH₃ONO₂ (Fig. 1a), the absorbance peaks can be assigned as follows^[53]: 3014 cm^–1, stretching vibration of methyl (group); 2965 cm^–1, asymmetric stretching of methyl; 1469 cm^–1, the bending of methyl; 1178 and 1139 cm^–1, out of plane bend of methyl; 1675 cm^–1, asymmetric stretching of NO₂; 1291 cm^–1, symmetric stretching of NO₂ ; 1018 cm^–1, stretching of C–O; 855 cm^–1, stretching of N–O; 760 and 658 cm^–1, out of plane bend of NO₂.

Fig. 1.  Standard FTIR spectra of (a) CH₃ONO₂ and (b) C₂H₅ONO₂ in pure N₂ at ~1.01 × 10⁵ Pa.

C₂H₅ONO₂ has a similar structure, and thus, most of its characteristic IR absorption bands are similar to those of CH₃ONO₂ (Fig. 1b): 3004 cm^–1, stretching of methyl; 2996 cm^–1, asymmetric stretching of methyl; 2939 cm^–1, asymmetric stretching of methylene; 1652 cm^–1, nitro asymmetric stretching; 1287 cm^–1, nitro symmetric stretching; 1029 cm^–1, stretching of C–O; 852 cm^–1, stretching of N–O; 760 and 668 cm^–1, out of plane bend of NO₂. However, C₂H₅ONO₂ has one more absorption band at 912 cm^–1, which has not been assigned in this work.

Fig. 2 shows the temporal evolution of FTIR spectra during the photolysis of (a) CH₃ONO₂ and (b) C₂H₅ONO₂ in air (O₂ : N₂ = 1 : 4). Because the mixture of gases before photolysis was selected as the spectroscopic background, the negative absorbance represents the reactant consumption and the positive absorbance represents the product formation. As the aerobic reaction proceeded, the concentration of CH₃ONO₂ or C₂H₅ONO₂ decreased and the products, which are indicated by the new peaks in the spectra, formed. On the basis of the subtraction analysis of the spectra, HNO₃, HO₂NO₂, CO, HCHO and HC(O)OH were identified as the products in the photolysis of both CH₃ONO₂ and C₂H₅ONO₂, with major absorption bands appearing at: 1711, 1326 and 879 cm^–1 for HNO₃; 1735, 1397, 1304 and 803 cm^–1 for HO₂NO₂; 2229–2050 cm^–1 for CO; 2926–2685 cm^–1 for HCHO; and 1776 and 1105 cm^–1 for HC(O)OH. Moreover, PAN was identified as a photolytic product of C₂H₅ONO₂, with major bands at 1843, 1740, 1303 and 1164 cm^–1. It is worth noting that the expected nitrogen-containing substances, including NO₂, N₂O₅, NO, HONO, CH₃ONO and C₂H₅ONO, were not detected when each of their standard spectra was used to make comparisons. However, HO₂NO₂, PAN and a non-reactive nitrogen-containing substance, namely HNO₃, were found in these reactions conducted in air.

Fig. 2.  FTIR spectra for the photolysis of (a) CH₃ONO₂ and (b) C₂H₅ONO₂ in dry air (O₂ : N₂ = 1 : 4). Mixture of gases before photolysis was selected as the spectral background, resulting in the negative absorbance for the reactant consumption and the positive absorbance for the product formation. FTIR spectra were collected at 5-min intervals, and this figure only shows the selected four spectra. PAN, peroxyacetyl nitrate.

In the samples photolysed in air, we also identified peroxides, carbonyl compounds and acid ions using offline HPLC and IC as described in the experimental methods section. In addition to HCHO, HC(O)OH and HNO₃ that were observed by in-situ FTIR spectroscopy as described above, hydrogen peroxide (H₂O₂) was detected as a product for the photolysis of both CH₃ONO₂ and C₂H₅ONO₂. In C₂H₅ONO₂ aerobic photolysis, four more products, CH₃CHO, CH₃C(O)OH, methyl hydroperoxide (MHP, CH₃OOH) and peroxyacetic acid (PAA, CH₃C(O)OOH), were detected.

We also investigated the UV photolysis of CH₃ONO₂ or C₂H₅ONO₂ in pure N₂ and found that the formed products included HCHO/CH₃CHO, CO, HONO and NO, but excluded PAN, HNO₃ and HO₂NO₂, products that were detected when the photolysis experiments were conducted in air. This result indicated that the presence of O₂ will lead to a series of different nitrogen-containing products. Because the photolysis of CH₃ONO₂ and C₂H₅ONO₂ in the pure N₂ does not reflect the real atmosphere, we will not include any further discussion of this experiment.

Fig. 3 shows the time series spectra of the OH-initiated oxidation of (a) CH₃ONO₂ and (b) C₂H₅ONO₂ in air. The spectra of the mixture of gases obtained before the reaction was used as the spectroscopic background, resulting in negative absorbance for the reactant consumed and positive absorbance for the product formed. Based on the analysis of spectroscopic subtractions, HNO₃, HO₂NO₂, CO, HCHO and HC(O)OH were detected as the main products for both CH₃ONO₂ and C₂H₅ONO₂ reactions, and one more product, PAN, was also present for the C₂H₅ONO₂ reaction. As in the photolysis system in the presence of O₂, NO₂, N₂O₅, NO, HONO, CH₃ONO and C₂H₅ONO were also not detected in the OH-initiated oxidation of both methyl and ethyl nitrate.

Fig. 3.  FTIR spectra of OH-initiated reactions of (a) CH₃ONO₂ and (b) C₂H₅ONO₂ in humid air (O₂ : N₂ = 1 : 4, relative humidity 30 %). PAN, peroxyacetyl nitrate.

Offline HPLC and IC were employed to further indentify the products formed. In addition to verification of the formation of HC(O)OH, HCHO and HNO₃, which were detected by in-situ FTIR spectroscopy, this offline analysis revealed that H₂O₂ was produced in the CH₃ONO₂ reaction, and H₂O₂, CH₃CHO, CH₃OOH, CH₃C(O)OOH and CH₃C(O)OH were produced in the C₂H₅ONO₂ reaction.

We examined the temporal evolution of reactants and products in the UV photolysis of (a) ~200-ppmv CH₃ONO₂ and (b) ~150-ppmv C₂H₅ONO₂ conducted in dry air (Fig. 4). The alkyl nitrate concentration decreased by 42 % at 40 min and 56 % at 80 min for CH₃ONO₂ and by 53 % at 40 min and 68 % at 80 min for C₂H₅ONO₂. For both CH₃ONO₂ and C₂H₅ONO₂ photolysis, the kinetic characterisations for several carbon-containing products were as follows: (i) the concentration of HCHO increased in the early part of the reaction and then decreased after ~40 min; (ii) the concentration of HC(O)OH increased gradually and then levelled to a constant value; and (iii) the concentration of CO increased steadily throughout the course of the study. For the nitrogen-containing products, the reaction intermediates were observed as follows: (i) the concentration of HO₂NO₂ remained at a low level, ~1–3 ppmv for CH₃ONO₂ photolysis and even lower for C₂H₅ONO₂ photolysis; (ii) HNO₃ increased gradually, approaching 82 and 62 ppmv for CH₃ONO₂ and C₂H₅ONO₂ at 90 min; and (iii) the concentration of PAN in C₂H₅ONO₂ photolysis reached a maximum at ~50 min, and levelled off at a constant value 38 ppmv.

Fig. 4.  Plots of the temporal evolution of reactants and products in the photolysis of (a) CH₃ONO₂ and (b) C₂H₅ONO₂ in dry air (O₂ : N₂ = 1 : 4). PAN, peroxyacetyl nitrate.

The combined data obtained by FTIR, IC and HPLC analyses were used to estimate the yields of the products. In this study, a yield is defined as the ratio of the molar amount of a product formed versus the molar amount of the reactant consumed. The time series for observed yields of HNO₃, HO₂NO₂ and PAN are displayed in Fig. 5. The yield of HNO₃ was nearly constant throughout the reaction process, with ~76 % for CH₃ONO₂ photolysis and ~57 % for C₂H₅ONO₂ photolysis. In order to estimate the wall effect of HNO₃, we investigated the loss rate of gaseous HNO₃ with ~100 ppmv in the reactor, under the same reaction conditions as the photolysis reaction of CH₃ONO₂ or C₂H₅ONO₂. The result showed a loss of <5 % of HNO₃ over 90 min, indicating that HNO₃ adsorption onto the wall was minor under dry air conditions Thus, the yield of HNO₃ observed by FTIR spectroscopy in the photolysis of CH₃ONO₂ or C₂H₅ONO₂ was an accurate reflection of its actual yield. For CH₃ONO₂ photolysis, the yield of HO₂NO₂ reached 12 % at 5 min and decreased to 2 % at 90 min. For C₂H₅ONO₂ photolysis, the yield of PAN remained at ~40 % throughout the reaction process. Peroxides were observed with low yields for both CH₃ONO₂ and C₂H₅ONO₂ photolysis, probably because they are unstable and easily photodissociate by UV irradiation.

Fig. 5.  Temporal profiles for the HNO₃ yield (Y_HNO3-MN) and HO₂NO₂ yield (Y_HO2NO2-MN) in the photolysis of CH₃ONO₂ and the HNO₃ yield (Y_HNO3-EN) and peroxyacetyl nitrate (PAN) yield (Y_PAN-EN) in the photolysis of C₂H₅ONO₂ in dry air (O₂ : N₂ = 1 : 4).

It is interesting that CH₃C(O)OH was not detected by FTIR spectroscopy in either the photolysis or the OH-initiated oxidation of C₂H₅ONO₂, but rather it was detected by IC at a level of ~40 ppmv, a much higher value than its detection limit by FTIR analysis. Moreover, the yield of CH₃C(O)OH was found to be close to that of PAN which was detected by FTIR spectroscopy. We therefore propose that the CH₃C(O)OH was a technical artefact produced by the hydrolysis of PAN collected in water for IC analysis, rather than by the reaction that occurred in the reactor.

The observed product yields and the carbon and nitrogen mass balances at 40 min are listed in Table 1. The experimental error represents 1σ (each experiment was repeated three to five times). It can be seen that the carbon balance was ~100 % within experimental error for the photolysis of both CH₃ONO₂ (113 ± 13 %) and C₂H₅ONO₂ (110 ± 15 %), with major carbon products as HCHO, HC(O)OH and CO, whereas the C₂H₅ONO₂ photolysis generated two more products, namely CH₃CHO and PAN, both with high yields.

Table 1.  Yields (%) of products and the carbon and nitrogen mass balance in the photolysis in dry air (O₂ :_{N2 = 1 : 4) and OH-initiated reactions of CH3ONO2 and C2H5ONO2 in humid air (O2 : N2 = 1 : 4, relative humidity 30 %)}
The experimental error represents 1σ

For C₂H₅ONO₂ photolysis, the nitrogen balance observed by FTIR spectroscopy was ~100 % within experimental error (100 ± 12 %), with HNO₃ and PAN as the main nitrogen-containing products; for CH₃ONO₂ photolysis, however, the nitrogen balance was only 79 ± 16 %, with HNO₃ as the major product. To find the ‘missing nitrogen’ in this reaction, we employed IC to determine the HNO₃ and the result showed that this HNO₃ was higher than that detected by FTIR spectroscopy, affording a quantitative yield of HNO₃, within experimental error. This result suggests that the unknown nitrogen-containing compound, not distinguished in the mixture by FTIR analysis, was hydrolysed into HNO₃ during the sample preparation. By analogy with PAN formed in C₂H₅ONO₂ photolysis, we tentatively consider this unknown nitrogen-containing compound as the homologue of PAN, namely, peroxyformyl nitrate (PFN, HC(O)OONO₂). The formation of PFN in the irradiation of a HCHO + NO₂ + Cl system in air was assumed by Gay et al. based on IR spectra,^[54] but this was later corrected to be peroxy nitric acid by Hanst and Gay.^[55] We could not identify PFN by FTIR spectroscopy because of the lack of its standard spectrum. We suggest that the difference between the yields of HNO₃ detected by IC and FTIR analysis accounts for the yield of PFN. Noticeably, the carbon mass balance was ~100 % when only HC(O)OH was considered and PFN was not, as shown in Table 1. If PFN is added into this balance, the carbon mass balance will exceed 100 %, a condition that could not occur in the reaction. Unfortunately, it is difficult for us to quantify PFN and HC(O)OH simultaneously. For FTIR analysis, HC(O)OH might be overestimated by the absorption band overlap between the PFN and HC(O)OH, as both of them contain the HC(O) group. Contrastingly, for IC analysis, HC(O)OH might be produced by the hydrolysis of PFN in water. Obviously, the assumption of the formation of PFN is not convincing enough simply by the analogy with PAN, and further study is needed to identify this unknown nitrogen-containing compound.

Fig. 6 shows the plot of the temporal evolution of reactants and products in the OH-initiated oxidation of (a) ~200-ppmv CH₃ONO₂ and (b) ~150-ppmv C₂H₅ONO₂ in air with 30 % relative humidity. The alkyl nitrate concentration decreased by 53 % at 40 min and 74 % at 80 min for CH₃ONO₂, and 60 % at 40 min and 80 % at 80 min for C₂H₅ONO₂. For both CH₃ONO₂ and C₂H₅ONO₂ reactions, CO increased gradually with time; and HC(O)OH increased gradually and levelled off. For CH₃ONO₂, HCHO reached a maximum at 40 min and then decreased gradually, whereas for C₂H₅ONO₂, HCHO was not evident in the FTIR spectra.

Fig. 6.  Plots of the temporal evolution of reactants and products in OH-initiated reactions of (a) CH₃ONO₂ and (b) C₂H₅ONO₂ in humid air (O₂ : N₂ = 1 : 4, relative humidity 30 %). PAN, peroxyacetyl nitrate.

In the CH₃ONO₂ reaction, HO₂NO₂ was detected at low levels (~2 ppmv), and at even lower levels for the C₂H₅ONO₂ reaction. PAN formed in the C₂H₅ONO₂ reaction reached 59 ppmv at 90 min. HNO₃ formed in both CH₃ONO₂ and C₂H₅ONO₂ reactions was lower than that in their corresponding photolytic reactions, and obviously decreased with the reaction time. This result may imply that the rapid deposition of HNO₃ or the formation of a complex of HNO₃ with H₂O would occur during the reaction because water vapour with 30 % relative humidity was present in the reactor; water was added to the chamber in order to generate OH. HNO₃ is a ‘sticky’ molecule and is readily adsorbed onto surfaces, particularly if water is present.^[56] In humid air, HNO₃ can combine with H₂O molecules to form (HNO₃)_x(H₂O)_y.^[57] The loss rate of gaseous HNO₃ is therefore much higher in humid air than in dry air.

The chemical fate of HNO₃ was examined under the same conditions that were applied to those for the OH radical studies with the two nitrates, as fully described in the Experimental methods section. The result showed that the HNO₃ could undergo a significant loss in the reactor under humid conditions. Combining the observed HNO₃ yield and its rate of loss under the same conditions, we estimated the actual yield of HNO₃ formed in the OH-initiated oxidation of CH₃ONO₂ and C₂H₅ONO₂ (Fig. 7). The estimated actual yields remained almost constant, with ~50 % for the OH⁺ CH₃ONO₂ reaction and ~40 % for the OH⁺ C₂H₅ONO₂ reaction, although the observed yields decreased gradually with reaction time because of the deposition of gaseous HNO₃.

Fig. 7.  Temporal profiles for the observed and estimated yields of HNO₃ (Y_HNO3-obs and Y_HNO3-est) and the yields of HO₂NO₂ (Y_HO2NO2) and peroxyacetyl nitrate (PAN) (Y_PAN) in OH-initiated reactions of (a) CH₃ONO₂ and (b) C₂H₅ONO₂ in humid air (O₂ : N₂ = 1 : 4, relative humidity 30 %). Y_HNO3-est = Y_HNO3-obs + Y_HNO3-loss, where Y_HNO3-obs is the observed HNO₃ yield by FTIR, Y_HNO3-loss is the lost HNO₃ yield because of the loss of HNO₃ in the reactor and Y_HNO3-est represents the actual HNO₃ yield.

The yields of HO₂NO₂ and PAN from CH₃ONO₂ or C₂H₅ONO₂ are also shown in Fig. 7. For the CH₃ONO₂ reaction, the yield of HO₂NO₂ decreased from 16 % at 1 min to 1 % at 70 min because of its instability (Fig. 7a). For the C₂H₅ONO₂ reaction, the yield of PAN remained ~50 % throughout the reaction process. Peroxides were observed in the OH-initiated reactions but with a minor yield.

CO₂ is potentially a product and should be estimated. Unfortunately, we could not detect its concentration directly by FTIR spectroscopy, because a part of the FTIR spectrometer’s optical path was exposed to the ambient air, resulting in an interference by ambient CO₂. However, CO₂ can be estimated by the kinetics of CO in the reactor, because CO₂ is usually considered to be produced by the oxidation of CO. We investigated the respective oxidation rate of CO by O₃ and OH radicals, under the same conditions as for the photolysis and OH-initiated oxidation of CH₃ONO₂ or C₂H₅ONO₂. The result indicated that ~20 % of CO was oxidised by OH radicals in 60 min, whereas there was no obvious oxidation of CO in the CO + O₃ reaction in 90 min. For the OH-initiated oxidation of CH₃ONO₂ or C₂H₅ONO₂, the CO₂ yield was then estimated using a linear fit on the basis of the observed CO concentration and OH⁺CO reaction rate (Table 1).

The yields of products and the carbon and nitrogen mass balances in the OH-initiated reactions are shown in Table 1. The carbon balance for reactions with both ANs approached 100 %, with high yields of CO, HCHO and HC(O)OH for the CH₃ONO₂ reaction, and high yields of CO and PAN for the C₂H₅ONO₂ reaction. The results of nitrogen mass balance in the OH-initiated reactions were similar to the corresponding photolysis reactions, as follows: for the OH⁺C₂H₅ONO₂ reaction, the main nitrogen-containing products were HNO₃ and PAN; for the OH⁺CH₃ONO₂ reaction, HNO₃ was detected as one main nitrogen-containing product, and PFN was proposed as another main nitrogen-containing product.

For CH₃ONO₂ photolysis, two possible cleavage pathways exist that produce two different fragments, namely NO₂ and ONO₂, as follows:

If photolysis is dominated by the reaction 10a pathway and occurs in the absence of O₂, HCHO, CO and HONO could be expected to form by the following reactions:

However, we observed HCHO, CO and HONO formation during CH₃ONO₂ photolysis in pure N₂. The formation of HONO cannot be explained by the reaction 10b pathway. If the two pathways 10a and 10b coexist, HNO₃ would be produced by the combination of ONO₂ with CH₃O or HCHO in the absence of O₂:

However, this outcome is inconsistent with our experiment result which indicated no formation of HNO₃ in the absence of O₂. In summary, we suggest that the photolysis of CH₃ONO₂ follows the reaction 10a pathway exclusively. This is in agreement with the recent report for a NO₂ quantum yield of unity from photodissociation of methyl and isopropyl nitrate.^[58]

In the presence of O₂, the H and HCO radicals formed from reactions 11 and 13b, will rapidly react with O₂, resulting in the formation of HO₂, H₂O₂ and OH (reactions 16–18), and the oxidisation of NO₂ into HNO₃ (reaction 20) and HO₂NO₂ (reaction 21). HC(O)O₂ was reported to be the intermediate product in the HCO + O₂ reaction, and it can decompose rapidly to CO and HO₂ because of its instability (reactions 19).^[59,60] In the present reaction system, we suggest that part of the HC(O)O₂ reacts with NO₂ to give HC(O)O₂NO₂ (PFN) before its decomposition (reaction 22):

Like CH₃ONO₂, C₂H₅ONO₂ will undergo cleavage of the O–N bond, giving fragment pairs of CH₃CH₂O and NO₂ (reaction 23). The subsequent set of reactions leads to the formation of HCHO, CH₃CHO, CO, HNO₃, HO₂NO₂, H₂O₂, etc. In addition, PAN, PAA and MHP are formed (reactions 28, 29a and 31).

For the photolysis of both CH₃ONO₂ and C₂H₅ONO₂, formic acid was observed. It appears to be formed from HCHO and a HO₂ radical^[61]:

The above mechanisms agree well with the products observed in our experiments, which gave the observed carbon and nitrogen mass balances, although PFN was not directly identified.

The low level of the observed HO₂NO₂ can be explained as follows. First, we estimated the reaction rates of HO₂ + NO₂ and HO₂ + HO₂. The rate constants are 1.4 × 10^–12 cm³ molecule^–1 s^–1 for k_HO2+NO2 and 2.9 × 10^–12 cm³ molecule^–1 s^–1 for k_HO2+HO2 at 1 atm and 298 K.^[62] Because of the lack of experimental values, the concentrations of NO₂ and HO₂ in the CH₃ONO₂ photolysis were estimated using the same box model method as that for C₂H₅ONO₂ photolysis described below. The model output gave the concentrations of ~3.7 × 10¹¹ molecule cm^–3 for HO₂ and ~2.2 × 10¹² molecule cm^–3 for NO₂. Combining the concentrations with rate constants, the reaction rates are estimated as 1.1 × 10¹² molecule cm^–3 s^–1 for the HO₂ + NO₂ reaction and 0.4 × 10¹² molecule cm^–3 s^–1 for the HO₂ + HO₂ reaction. This result indicates that the formation of HO₂NO₂ will compete with that of H₂O₂. However, HO₂NO₂ will be rapidly lost by photodissociation,^[63] thermal decomposition,^[64] and reaction with OH,^[65] resulting in its low photostationary state concentration. This is verified by the model, which gives its photostationary state concentration as ~3 ppmv.

As shown in Table 1, the observed yields of peroxides, including H₂O₂, MHP and PAA, are considerably low (~1 %) in the photolysis of C₂H₅ONO₂, owing to their instability. CH₃C(O)OH was not detected by FTIR spectroscopy, although it should be produced by the CH₃C(O)O₂ + HO₂ reaction (reaction 29b). This outcome implies that the branching ratio of reaction 29b is minor compared with that of reaction 29a and 29c.^[66,67]

Although our instrument did not detect the NO₂ that was expected to form according to the proposed mechanism, we have simulated its concentration in the photolysis of C₂H₅ONO₂ using a box model on the basis of the Carbon Bond Mechanism-Version IV (CBM-IV) developed by Gery et al.^[68] In the present study, we modified the CBM-IV mechanism by deleting some irrelevant reactions and adding the C₂H₅ONO₂-related reactions. We estimated the photon flux at 254 nm of the UV irradiation by the measured photolysis rate of C₂H₅ONO₂ (its cross-section is known from the literature) and then estimated the photolysis rates of the other species using this photon flux. The experimental initial reactant concentrations, namely, 155 ppmv for C₂H₅ONO₂ and 2 × 10⁵ ppmv for O₂, were input into the model. Fig. 8 shows the model results for the concentration variations of C₂H₅ONO₂ and the main nitrogen-containing products. The modelled concentrations of PAN (maximum ~80 ppmv) are higher than those measured in our experiments (maximum ~60 ppmv) and those of HNO₃ are lower (modelled ~35 ppmv, measured ~50 ppmv). This discrepancy may arise from the uncertainties in the kinetic parameters used in the model. Several of the rate constants used in the model are limited and uncertain, including those for the formation of PAN (k_PAN,f) and HNO₃ (k_HNO3,f). Obviously, decreasing k_PAN,f or increasing k_HNO3,f would improve the modelled results. It is evident that NO₂ was produced and then was rapidly transformed into PAN and HNO₃, resulting in a maximum level of ~2 ppmv for the photostationary state of NO₂, and a much lower NO level. The NO_x photostationary state was achieved at ~40 min after the photolysis reaction began. As a result, the FTIR analysis was not able to detect NO_x at such a low level. Moreover, it is suggested that the initial excitation in the photodissociation of CH₃ONO₂ is located on the NO₂ moiety, and most of the available energy is deposited in it.^[69,70] It is worth mentioning that we carried out an additional experiment in which ~40 ppmv of NO₂ was added to the CH₃ONO₂ photolysis system. The result showed that the production rate of HNO₃ in this NO₂-suplemented system had no obvious changes, compared with the control system containing no NO₂. This result implies that the added NO₂ reacts much more slowly than does the NO₂ in the excited-state derived from the photodissociation of CH₃ONO₂. This outcome may also indicate that the reaction of excited-state NO₂ (NO₂*) to form HNO₃ and peroxyacyl nitrates is competitive with its quenching. Therefore, we suggest that the NO₂* fragment from the CH₃ONO₂ or C₂H₅ONO₂ photolysis may have extra energy (in an electronically excited state), making it much more reactive than ground state NO₂ in the atmosphere, resulting in the formation of more HNO₃ and peroxyacyl nitrates (RC(O)OONO₂, PAN and the homologue of PAN). To our knowledge, we have not found specific reports on the amount of extra energy in the NO₂* formed in the photolysis of methyl and ethyl nitrates. Unfortunately, we currently cannot make an estimation on the extra energy in the NO₂* and the dominant modes.

Fig. 8.  Temporal profiles of C₂H₅ONO₂ and main nitrogen-containing products in the modelled photolysis reaction of C₂H₅ONO₂ in dry air (O₂ : N₂ = 1 : 4). PAN, peroxyacetyl nitrate.

In the OH-initiated oxidation of ANs, the OH radicals were produced by UV photolysis of O₃ in the presence of water vapour. Because of the use of UV irradiation, the photolytic reaction should occur simultaneously. However, we observed that the consumption rate of CH₃ONO₂ or C₂H₅ONO₂ in the OH-initiated oxidation was ~30 % higher than that in the corresponding photolysis, indicating that the OH-initiated oxidation was the dominant pathway in the reaction system.

For this study, the reactions of CH₃ONO₂ and C₂H₅ONO₂ with OH radicals were carried out in simulated atmospheric conditions, and HNO₃ was detected as a major product. The OH reaction has two different pathways, namely, addition and H-atom abstraction. As the H-atom abstraction could produce bifunctional organic nitrates such as carbonyl-nitrates,^[71] rather than HNO₃ or PAN, we think that this pathway should not predominate. Alternatively, the OH addition is possibly the dominant pathway in these reactions, resulting in the formation of an unstable adduct (reactions 37 and 38). This view is supported by Nielsen et al.^[41]

The adduct has two routes for its decomposition: (1) the direct formation of HNO₃, accompanied by the formation of the alkoxy radical (reactions 37a and 38a); and (2) formation of the excited NO₂* and alkyl peroxides (reactions 37b and 38b). CH₃OOH and C₂H₅OOH, which are formed in reactions 37b and 38b, will be rapidly transformed into other products by their photodissociation and reactions with OH, leading to their low photostationary state concentrations. HNO₃, HO₂NO₂, PFN or PAN are produced in the subsequent reactions of NO₂*. This mechanism is similar to the photolysis system described previously.

PAN was observed with a high yield in the OH-initiated oxidation of C₂H₅ONO₂, and this implies that the excited NO₂* was produced by the second decomposition pathway for the adduct (reactions 37b and 38b) because no evidence was reported for the formation of PAN by direct decomposition of the adduct. However, Nielsen et al.^[41] suggested a pathway leading to the direct decomposition of the adduct into HNO₃ (reactions 37a and 38a). These two routes for the adduct decomposition likely coexist and the branch ratio possibly approaches 1 : 1, considering the observed HNO₃/peroxyacyl nitrates ratio of ~1 : 1, as seen in Table 1. Unfortunately, we cannot determine the relative contributions of these two channels in the present experimental study.

In summary, for the OH-initiated oxidation of ANs, there are two pathways to form HNO₃, namely, direct OH-adduct decomposition and OH+NO₂* reaction, whereas for the photolysis of ANs, the OH+NO₂* reaction is the only pathway. We note that the yield of the OH+NO₂* pathway leading to HNO₃ could be different between the laboratory conditions presented in this paper and atmospheric conditions. In the atmosphere, whether NO₂* competes effectively for OH to form HNO₃ very much depends on the atmospheric composition and conditions. In the photolytic reaction, small concentrations of OH radicals were produced by a series of reactions after the cleavage of the alkyl nitrate O–N bond; in the OH-initiated reaction, the OH radicals were produced by the photolysis of O₃ in the presence of water vapour (reactions 8 and 9). It is expected that the OH radicals in the OH-reaction are much more abundant than those in the photolysis. This can also explain the discrepancy regarding the CH₃CHO and PAN concentrations observed between the photolysis and OH-initiated reaction of C₂H₅ONO₂ (Table 1). As mentioned above, acetaldehyde is transformed into PAN through the OH-initiated reactions (reactions 26–28). The higher OH level in the OH-initiated reactions of C₂H₅ONO₂ results in the lower CH₃CHO level as well as higher PAN level relative to those in the photolytic oxidation reaction of C₂H₅ONO₂.

Under simulated atmospheric conditions, we have for the first time carried out a laboratory study of the atmospheric photochemical reactions of CH₃ONO₂ and C₂H₅ONO₂, which served as models of ANs. A series of intermediate and end products, including carbon- and nitrogen-containing compounds, were well characterised, and the carbon and nitrogen balances based on these observed compounds were ~100 %, within experimental error (1σ) for the reactions. Notably, for both photolysis and OH-initiated reactions of CH₃ONO₂ and C₂H₅ONO₂, HNO₃ was found to be a major nitrogen-containing product with a yield of ~50 % or more, whereas PAN was observed in C₂H₅ONO₂ reactions and PFN was proposed as a product in CH₃ONO₂ reactions with significant yields. The detailed mechanisms of these reactions have been deduced based on the experimental results. Noticeably, we used a Hg lamp (λ_max = 254 nm) as the light source because it can provide enough UV intensity to encompass the AN photolysis lifetime in one hour at a controlled temperature. We propose that this UV wavelength, which is shorter than that in the troposphere, would not significantly change the reaction mechanism of ANs, which is intensively concerned in the present study, although the reaction rates and the product yields would be different from that in the troposphere.

The present study has revealed that the photochemical reactions of CH₃ONO₂ and C₂H₅ONO₂ in the atmosphere would ultimately produce PAN or PFN, a reactive nitrogen-containing substance, and HNO₃, a photochemically non-reactive nitrogen-containing substance. It is expected that the higher-weight ANs might undergo a similar atmospheric fate. This finding implies that ANs may serve as not only a reservoir but also as a sink for reactive nitrogen in the atmosphere. Such a dual role for ANs in the nitrogen cycle should be considered in the models in regard to air quality and nitrogen exchange between atmosphere and surface.

Herein, we have attempted to estimate the production of HNO₃ and peroxyacyl nitrates derived from NO_x by way of ANs as intermediates in the atmosphere on the basis of their yields obtained by our experiments. The average yields from CH₃ONO₂ and C₂H₅ONO₂ photolytic and OH-initiated decompositions (Table 1) are 57 % for HNO₃ and 36 % for peroxyacyl nitrates. We assume that these experimental yields approximate their yields in the real atmosphere. Recent measurements indicated that ∑ANs comprise a large fraction of NO_y^[2,3,14]; the result from the model analysis also suggested that ∑ANs represent a large fraction of NO_y, with 5–42 % for a rural scenario and 2–16 % for an urban scenario.^[72] Summarising the observed and model results, we suggest that ∑ANs comprise ~10 % of the NO_y in urban areas and ~20 % in rural areas. As mentioned earlier, ANs in the continental atmosphere are mainly produced from the reactions of NO_x. This means that 10 and 20 % of NO_x is converted into ANs in urban and rural atmospheres. Combining the percentages of ANs in NO_x transformation with HNO₃ and PAN yields from ANs, we estimate here that ANs account for the sink and reservoir fractions of NO_x by two scenarios: (i) in the urban atmosphere, HNO₃ makes up ~6 % of the nitrogen sink and ~4 % of the nitrogen reservoir is composed of peroxyacyl nitrates; and (ii) in the rural atmosphere, HNO₃ comprises ~11 % of the nitrogen sink and ~7 % of the nitrogen reservoir is peroxyacyl nitrates. This estimation is illustrated in Fig. 9. These sink and reservoir contributions are significant for the total NO_x cycle, especially in rural areas.

Fig. 9.  Roles for alkyl nitrates (ANs) in the cycle of atmospheric NO_x. (NO₂)* represents the excited-state of NO₂ (NO₂*); for the OH-initiated reactions, it also includes the OH-adduct containing NO₂* group (as seen in reactions 37a and 38a). PANs, peroxyacyl nitrates (RC(O)OONO₂, where R = H, CH₃, CH₃CH₂,…).

Recently, a modelling study on global atmospheric chemistry pointed out that PAN has increased globally by 9 % from megacities under year 2000 conditions.^[73] We hope our findings will be useful for a better understanding of this global PAN change.