Oxidation pathways and emission sources of atmospheric particulate nitrate in Seoul: based on δ¹⁵N and Δ¹⁷O measurements

PM_2.5 haze pollution driven by secondary inorganic ${\mathrm{NO}}_{\mathrm{3}}^{-}$ has been a great concern in East Asia. It is, therefore, imperative to identify its sources and oxidation processes, for which nitrogen and oxygen stable isotopes are powerful tracers. Here, we determined the δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) of PM_2.5 in Seoul during the summer of 2018 and the winter of 2018–2019 and estimated quantitatively the relative contribution of oxidation pathways for particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and investigated major NO_x emission sources. In the range of PM_2.5 mass concentration from 7.5 µg m⁻³ (summer) to 139.0 µg m⁻³ (winter), the mean δ¹⁵N was −0.7 ‰ ± 3.3 ‰ and 3.8 ‰ ± 3.7 ‰, and the mean Δ¹⁷O was 23.2 ‰ ± 2.2 ‰ and 27.7 ‰ ± 2.2 ‰ in the summer and winter, respectively. While OH oxidation was the dominant pathway for ${\mathrm{NO}}_{\mathrm{3}}^{-}$ during the summer (87 %), nighttime formation via N₂O₅ and NO₃ was relatively more important (38 %) during the winter, when aerosol liquid water content (ALWC) and nitrogen oxidation ratio (NOR) were higher. Interestingly, the highest Δ¹⁷O was coupled with the lowest δ¹⁵N and highest NOR during the record-breaking winter PM_2.5 episodes, revealing the critical role of photochemical oxidation process in severe winter haze development. For NO_x sources, atmospheric δ¹⁵N (NO_x) estimated from measured δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) considering isotope fractionation effects indicates vehicle emissions as the most important emission source of NO_x in Seoul. The contribution from biogenic soil and coal combustion was slightly increased in summer and winter, respectively. Our results built on a multiple-isotope approach provide the first explicit evidence for ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation processes and major NO_x emission sources in the Seoul megacity and suggest an effective mitigation measure to improve PM_2.5 pollution.

1 Introduction

In northeast Asia, air pollution characterized by high PM_2.5 (particulate matter with aerodynamic diameter smaller than 2.5 µm) and ozone concentrations has received significant attention due to its serious effects on human health (Lelieveld et al., 2015; Xie et al., 2019). As a result of extensive efforts by East Asian countries to improve the ambient air quality, anthropogenic emissions of SO_x, NO_x, and CO have been significantly reduced, particularly in China (Cheng et al., 2019). Nonetheless, the number of severe haze events and the duration have been increased, which is not understood clearly.

It is also noteworthy that there are common chemical and meteorological characteristics in the occurrence of the PM_2.5 haze pollution in northeast Asia, such as inorganic-species-dominated chemical composition (Liu et al., 2018; Shao et al., 2018; Wang et al., 2019a) and transboundary transport of haze aerosol depending on the synoptic atmospheric circulation (Quan et al., 2020; Shi et al., 2020; Zheng et al., 2019). Overall, approximately 40 % of PM_2.5 consists of ${\mathrm{NO}}_{\mathrm{3}}^{-}$, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$, and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (secondary inorganic aerosol, SIA) in both urban and background sites in China (Liu et al., 2018) for the period of 2012–2013. The mass contribution of SIA was further augmented up to 69 % at urban Beijing and island sites (Changdao) in the North China Plain (NCP) during the 2016 spring (Lim et al., 2020). A large increase in SIA mass is particularly associated with efficient formation of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ in cold months (most frequently in December to March), leading to a rapid increase in PM_2.5 concentration and developing a persistent haze pollution on a regional scale (Li et al., 2018; Xu et al., 2019). Recent studies suggest that the regional occurrence of the PM_2.5 haze events derived by ${\mathrm{NO}}_{\mathrm{3}}^{-}$ across the Asian continent has been associated with long-range transport of air pollutants promoted by cold fronts and their intrusion in downward regions by the development of the atmospheric boundary layer (Kang et al., 2019; Lee et al., 2019). However, scientific understanding is still limited because the rapid increase of fine-aerosol ${\mathrm{NO}}_{\mathrm{3}}^{-}$ is the result of the complex interplay of oxidation and transformation mechanisms producing ${\mathrm{NO}}_{\mathrm{3}}^{-}$ from various fossil and non-fossil sources of NO_x with micro-to-synoptic meteorology creating the conditions for ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation. In addition, considering that NO_x emissions from bottom-up emission inventory are smaller than top-down estimates by satellites (e.g., Goldberg et al., 2019), our understanding of NO_x emissions is poor.

In order to better understand processes of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation including NO_x emission sources in East Asia, a growing number of recent studies have suggested nitrogen and oxygen stable-isotope-based measurement as a robust and useful tool for providing important clues for ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation as well as NO_x emission sources (He et al., 2018; Lim et al., 2019, 2020; Song et al., 2019; Zong et al., 2017, 2020). The isotopic composition is expressed in parts per thousand (‰) and reported as δ (‰) $=(R_{\text{sample}}/R_{\text{standard}}-\mathrm{1})\times \mathrm{1000}$ with R being the ratio of the heavy isotope over the light isotope (e.g., ${}^{\mathrm{15}}\mathrm{N}{/}^{\mathrm{14}}\mathrm{N}$, ${}^{\mathrm{18}}\mathrm{O}{/}^{\mathrm{16}}\mathrm{O}$, ${}^{\mathrm{17}}\mathrm{O}{/}^{\mathrm{16}}\mathrm{O}$) in a sample (R_sample) and in the international standard (R_standard). The international standards are the Vienna Standard Mean Ocean Water (VSMOW) and atmospheric N₂ for oxygen and nitrogen ratios. In terms of NO_x emission sources, the major NO_x sources are distinguished in their nitrogen isotopic compositions: biogenic soil (driven by fertilizer use; −35.1 ‰ ± 10.2 ‰) (Felix and Elliott, 2014; Li and Wang, 2008; Yu and Elliott, 2017), biomass burning (1.8 ‰ ± 1.8 ‰) (Fibiger and Hastings, 2016), vehicle emissions (−2.5 ‰ ± 1.5 ‰) (Walters et al., 2015), and coal combustion (14.2 ‰ ± 4.5 ‰) (Felix et al., 2012; Heaton, 1990). Despite evidence that δ¹⁵N (NO_x) can serve as a useful tracer of NO_x emission sources, during NO_y photochemical cycling and its loss processes, the δ¹⁵N of initial NO_x can be significantly altered by both kinetic and equilibrium isotopic fractionation effects (ε_N), complicating the link between δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and δ¹⁵N (NO_x) (Freyer et al., 1993; Li et al., 2020).

The formation processes of HNO₃ in the atmosphere consist of NO−NO₂ photochemical cycle Reactions (R1)–(R3) and post NO₂ oxidation Reactions (R4)–(R8). During the NO_x cycling, nitrogen isotopic fractionation is affected by unidirectional reactions of the Leighton cycle and NO_x isotope exchange equilibrium. A recent laboratory experiment has shown that the Leighton cycle isotope effect (LCIE) associated solely with O₃ Reaction (R1) and equilibrium isotopic effect (EIE) were −10 ‰ and 28.9 ‰ at room temperature, respectively (Li et al., 2020). The relative importance of these two effects on δ¹⁵N (NO₂) also depends on NO_x levels, leading to an increase (decrease) in δ¹⁵N (NO₂) relative to δ¹⁵N (NO_x) values at high (low) NO_x conditions (Kamezaki et al., 2019; Li et al., 2020; Walters et al., 2018). Overall, the nitrogen isotope exchange equilibrium has been suggested to be the dominant fractionation process in NO−NO₂ cycling at urban atmosphere (Freyer et al., 1993) and oxidation reactions forming nitric acid (HNO₃) and particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}$ (Savarino et al., 2013).

$$\begin{array}{}\text{(R1)}& {\displaystyle }\mathrm{NO}+{\mathrm{O}}_{\mathrm{3}}\to {\mathrm{NO}}_{\mathrm{2}}+{\mathrm{O}}_{\mathrm{2}},\text{(R2)}& {\displaystyle }\mathrm{NO}+{\mathrm{RO}}_{\mathrm{2}}\left({\mathrm{HO}}_{\mathrm{2}}\right)\to {\mathrm{NO}}_{\mathrm{2}}+\mathrm{RO}\left(\mathrm{OH}\right),\text{(R3)}& {\displaystyle }{\mathrm{NO}}_{\mathrm{2}}+\mathrm{hv}\to \mathrm{NO}+\mathrm{O}{(}^{\mathrm{3}}\mathrm{P}),\text{(R4)}& {\displaystyle }{\mathrm{NO}}_{\mathrm{2}}+\mathrm{OH}\to {\mathrm{HNO}}_{\mathrm{3}},\text{(R5)}& {\displaystyle }{\mathrm{NO}}_{\mathrm{2}}+{\mathrm{O}}_{\mathrm{3}}\to {\mathrm{NO}}_{\mathrm{3}}+{\mathrm{O}}_{\mathrm{2}},\text{(R6)}& {\displaystyle }{\mathrm{NO}}_{\mathrm{2}}+{\mathrm{NO}}_{\mathrm{3}}+\stackrel{M}{⟷}{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}},\text{(R7)}& {\displaystyle }{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}+{\mathrm{H}}_{\mathrm{2}}\mathrm{O}\textcolor{red}{}\mathrm{2}{\mathrm{HNO}}_{\mathrm{3}}\left(\mathrm{aq}\right),\text{(R8)}& {\displaystyle }{\mathrm{NO}}_{\mathrm{3}}+\mathrm{RH}\to {\mathrm{HNO}}_{\mathrm{3}}+\mathrm{R},\end{array}$$

where M is an unreactive third body and R is any organic group.

The kinetic fractionation effect during daytime oxidation of NO₂ to HNO₃ Reaction (R4) is relatively minor, being estimated to be −3 ‰ (Freyer, 1991). During the nighttime when most NO is oxidized to NO₂ without NO_x photolysis, the isotopic equilibrium between NO₂, NO₃, and N₂O₅ should be achieved (Reaction R6). The nighttime thermal equilibrium likely favors the partitioning of ¹⁵N into N₂O₅ relative to NO₂ and consequently induces a large isotopic fractionation effect of 25.5 ‰ (Reactions R5–R6; Walters and Michalski, 2015). On the other hand, the nitrogen partitioning between NO₂ and NO₃ Reaction (R5) may induce a fractionation effect of about −18 ‰ (Walters and Michalski, 2015). These N isotope fractionation effects should be first evaluated to explore NO_x source contributions, which should be based on combining with a robust tracer for the contributions of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ oxidation pathways.

Lately, Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) has been used for tracing ${\mathrm{NO}}_{\mathrm{3}}^{-}$ oxidation pathways (Alexander et al., 2009, 2020; Morin et al., 2009; Savarino et al., 2007, 2013). Earlier researches observed that atmospheric ${\mathrm{NO}}_{\mathrm{3}}^{-}$ is anomalously enriched in ¹⁷O (Michalski et al., 2003), which stems from O₃ formation reactions, where a rare isotope effect leads to excess ¹⁷O enrichment relative to what is expected based on the ¹⁸O enrichments (Thiemens, 1999, 2006). This enrichment is quantified by Δ¹⁷O notation (¹⁷O excess, defined as δ¹⁷O$-\mathrm{0.52}\times {\mathit{\delta }}^{\mathrm{18}}$O). Since non-zero Δ¹⁷O strictly reflects a photochemical effect, ${\mathrm{NO}}_{\mathrm{3}}^{-}$ produced by denitrification in soils should have Δ¹⁷O of zero. The mass-independent Δ¹⁷O signature of O₃ is transferred to NO_x, in which the number of oxygen atoms from O₃ is involved in NO_x. In this way, the Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) is served as a conservative marker to track the chemical formation of atmospheric ${\mathrm{NO}}_{\mathrm{3}}^{-}$. Photochemical formation by peroxy radicals (HO₂ and RO₂) leads to a relatively low Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$), whereas nighttime formation through N₂O₅ and NO₃ results in high Δ¹⁷O (Michalski et al., 2003; Morin et al., 2009; Savarino et al., 2007). Consequently, the difference in Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) suggests approaching a proportional contribution of daytime and nighttime oxidation of ${\mathrm{NO}}_{\mathrm{3}}^{-}$.

To date, few field studies have coupled Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) to investigate NO_x-to-${\mathrm{NO}}_{\mathrm{3}}^{-}$ oxidation processes and emission sources of NO_x, e.g., field researches in Japan (Nelson et al., 2018); West Virginia, USA (Rose et al., 2019); Shanghai, China (He et al., 2020); and Beijing, China (He et al., 2018; Song et al., 2020).

In this study, we present the measurement results of δ¹⁵N and Δ¹⁷O of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ in Seoul during the summer of 2018 and the winter of 2018–2019, when we encountered the record-breaking PM_2.5 concentrations. Then, the δ¹⁵N and Δ¹⁷O measurements are used to evaluate seasonally distinct atmospheric oxidation pathways of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and to explore major NO_x source contributions in the study region.

2 Measurements and methods

2.1 Sampling

We collected PM_2.5 filter samples on the rooftop of the six-story Asan science building at the Korea University campus located in northeast Seoul, the capital of South Korea (37.59^∘ N, 127.02^∘ E; Fig. 1) during the summer of 2018 (26 May to 22 August; n=13) and the winter of 2018–2019 (27 December to 8 March; n=18). Seoul is a metropolitan area with a population of 9.77 million and known to be influenced by heavy road traffic all around. The PM_2.5 particulates were collected on quartz filters (20 cm× 25 cm; Pallflex Products, Putnam, USA) at a nominal flow rate of 96 m³ hr⁻¹ for 1 to 3 d using a high-volume air sampler (3000 series, Ecotech, Australia). Filters were analyzed for water-soluble ions, carbonaceous compounds, total nitrogen (TN) and carbon (TC), and stable nitrogen and oxygen isotopic ratios. Reactive gases including O₃, NO, NO₂, SO₂, and CO and a meteorological suite including air temperature, relative humidity, and wind speed and direction were measured hourly at the nearby monitoring sites run by the National Institute of Environmental Research (NIER) and the Korea Meteorological Administration (KAM), respectively. These data were averaged daily for comparison with filter-based chemical composition data, if necessary.

Figure 1 Location of the study region. (a) Seoul metropolitan area in South Korea; the map was generated using the open-source software R (http://www.R-project.org/, last access: 2 February 2021). (b) Sampling site of the Korea University campus in the northeast Seoul; the map was taken from © Kakao maps (https://map.kakao.com/, last access: 29 December 2021).

2.2 Chemical analyses

Filters were stored in a freezer pending chemical analysis. Chemical composition of PM_2.5 was determined for eight water-soluble ions (Cl⁻, ${\mathrm{NO}}_{\mathrm{3}}^{-}$, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$, Na⁺, ${\mathrm{NH}}_{\mathrm{4}}^{+}$, K⁺, Ca²⁺, and Mg²⁺) by ion chromatography (IC; Eco IC, Metrohm, Switzerland), organic carbon (OC) and elemental carbon (EC) by an OC-EC analyzer (Sunset Laboratory Inc., US) using the thermo-optical transmittance method (NIOSH870), and TC and total nitrogen (TN) by an elemental analyzer (EA, Fisons NA-1500NC, Thermo, Waltham, MA, USA). Mass concentrations of these constituents were corrected for laboratory and field blanks. The detection limit, determined as 3 standard deviations (SD) above blank concentrations, was <0.1 ppm for ionic species, 0.5 µg cm⁻³ for TC (the sum of OC and EC), and 0.8 g µgN and 0.5 µgC per punched filter area for TN and TC, respectively. Details of the analytical methods can be found elsewhere (Lim et al., 2020).

Following the bacterial denitrifier method (Casciotti et al., 2002; McIlvin and Casciotti, 2011), the Δ¹⁷O of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ was measured simultaneously with δ¹⁸O and δ¹⁵N coupled with an isotope ratio mass spectrometry (IRMS) measurement using an in-house peripheral system at the Université Grenoble Alpes (Morin et al., 2009). In brief, ${\mathrm{NO}}_{\mathrm{3}}^{-}$ of samples was converted to N₂O via bacterial denitrification, and the N₂O was further converted into O₂ and N₂, which were separated via a gas chromatography column before being introduced to the IRMS system (Thermo Finnigan MAT 253 isotope ratio mass spectrometer). Samples were measured in batch with reference materials following strictly the identical treatment principles, including the same water matrix for standards and samples. Together with samples, a subset of international nitrate reference materials (US Geological Survey 32, 34, and 35, as well as their mixtures) was measured for correction and calibration of Δ¹⁷O and δ¹⁸O values relative to VSMOW and δ¹⁵N values relative to air N₂. The overall accuracy of the method is estimated as the reduced standard deviation of the residuals from the linear regression between the measured reference materials and their expected values (Morin et al., 2009). For these sets of analyses, the obtained uncertainties (1σ) were 0.4 ‰ and 0.3 ‰ for Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$), respectively. The analytical procedure used in this study strictly followed the method described in Morin et al. (2009), which adheres to the new standard bacterial method coupled with the gold catalyst for the thermal decomposition of N₂O into N₂ and O₂ (Kaiser et al., 2007).

2.3 Quantifying isotope fractionation effects

2.3.1 Isotope fractionation effects of NO₂ oxidation to atmospheric particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}$: δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$)

The HNO₃ forms through three major pathways including (i) OH pathway, (ii) O₃ pathway associated with N₂O₅, and (iii) O₃ pathway associated with NO₃.

OH pathway

When NO and NO₂ coexist in similar quantities due to the Leighton cycle, ¹⁵N is preferentially partitioned into NO₂ via the equilibrium isotope effect, leading to higher δ¹⁵N in NO₂ relative to NO and NO_x (Freyer et al., 1993; Walters et al., 2016). Considering the comparable concentrations of NO and NO₂ over a year in Seoul (Fig. S1 in the Supplement), the N isotope effects on δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) should be significantly affected by NO_x photochemical interactions. The relative importance of EIE and LCIE to δ¹⁵N(NO₂) can be assessed by comparing lifetimes of NO₂ with respect to isotope exchange with NO (τ_exchange) and photolysis (${\mathit{\tau }}_{{\mathrm{NO}}_{\mathrm{2}}+\mathrm{hv}}$). In this regard, the “A” factor was defined as ${\mathit{\tau }}_{\text{exchange}}/{\mathit{\tau }}_{{\mathrm{NO}}_{\mathrm{2}}+\mathrm{hv}}=j_{{\mathrm{NO}}_{\mathrm{2}}}/k_{\mathrm{1}}\times \left[\mathrm{NO}\right](k_{\mathrm{1}}=\mathrm{8.14}\times {\mathrm{10}}^{-\mathrm{14}}$ cm³ s⁻¹, Sharma et al., 1970) by Li et al. (2020, 2021). These studies demonstrated that A was as small as 0.01–0.5 in the EIE-dominated regime with NO_x >20 ppbv and the ratio of ${\mathrm{NO}}_{\mathrm{2}}/{\mathrm{NO}}_x$ <0.6. In Seoul, to simply evaluate the relative importance of EIE and LCIE only (not for ¹⁵N correction), the A factor was estimated to be 0.08±0.20 (median ± SD) and 0.60±0.51 during the winter and summer months, respectively, indicating the significant influence of the equilibrium isotope effect on NO−NO₂ isotopic fractionation. The time series $j_{{\mathrm{NO}}_{\mathrm{2}}}$ was calculated using the Master Chemical Mechanism (MCM) model (Saunders et al., 2003).

After NO_x photochemical cycling, NO₂ is oxidized by the reaction with OH radical to form atmospheric HNO₃ (HNO₃ Reaction R4 pathway). The N fractionation effect of particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}$ produced via the HNO₃ Reaction (R4) pathway (ε₁, unit in ‰) can be expressed as the following, neglecting kinetic isotope effects associated with the HNO₃ Reaction (R4) pathway (Walters and Michalski, 2016):

$$\begin{array}{}\text{(1a)}& {\displaystyle }\mathit{\delta }{}^{\mathrm{15}}\mathrm{N}\left({\mathrm{HNO}}_{\mathrm{3}}\right)\left(\mathrm{R}\mathrm{4}\right)={\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{NO}}_{\mathrm{2}}\right)={\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{NO}}_x\right)+{\mathit{\epsilon }}_{\mathrm{1}},\text{(1b)}& {\displaystyle }\begin{array}{rl}{\mathit{\epsilon }}_{\mathrm{1}}& =\left(\left({}^{\mathrm{15}}{\mathit{\alpha }}_{{\mathrm{NO}}_{\mathrm{2}}/\mathrm{NO}-\mathrm{1}}\right)\left(\mathrm{1}-f_{{\mathrm{NO}}_{\mathrm{2}}}\right)\right)/\left(\left(\mathrm{1}-f_{{\mathrm{NO}}_{\mathrm{2}}}\right)\right.\\ & +\left.\left({}^{\mathrm{15}}{\mathit{\alpha }}_{{\mathrm{NO}}_{\mathrm{2}}/\mathrm{NO}}\times f_{{\mathrm{NO}}_{\mathrm{2}}}\right)\right),\end{array}\end{array}$$

where $f_{{\mathrm{NO}}_{\mathrm{2}}}$ is the fraction of NO₂ relative to the total NO_x, and ${}^{\mathrm{15}}{\mathit{\alpha }}_{{\mathrm{NO}}_{\mathrm{2}}/\mathrm{NO}}$ is the temperature-dependent isotope equilibrium exchange fractionation factor for ${\mathrm{NO}}_{\mathrm{2}}/\mathrm{NO}$ (Walters et al., 2016). In the present study, the measured $f_{{\mathrm{NO}}_{\mathrm{2}}}$ was used individually for summer samples, and a seasonal mean $f_{{\mathrm{NO}}_{\mathrm{2}}}$ (0.69) was applied to winter samples due to the lack of availability of continuous NO−NO₂ measurement data.

Oxidation of NO_x to HNO₃ is regarded as the formation pathway of particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}$ via the HNO₃ Reaction (R4) pathway due to the unconstrained isotope fractionation effect between HNO₃ and ${\mathrm{NO}}_{\mathrm{3}}^{-}$, resulting in the following equation:

$$\begin{array}{}\text{(1c)}& {\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{HNO}}_{\mathrm{3}}\right)\left(\mathrm{R}\mathrm{4}\right)={\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{NO}}_{\mathrm{3}}^{-}\right)\left(\mathrm{R}\mathrm{4}\right).\end{array}$$

O₃ pathways associated with N₂O₅ and NO₃

During the nighttime when NO is oxidized into NO₂ without photolyzing back to NO and fresh emissions of NO are negligible, NO_x exists almost as NO₂, and thus δ¹⁵N (NO₂) should be reflective of the δ¹⁵N of NO_x sources. If NO₂ is oxidized to N₂O₅, the isotopic equilibrium is likely to be achieved between NO₂, NO₃, and N₂O₅ by chemical equilibrium Reaction (R6), and the δ¹⁵N values of N₂O₅ and NO₃ will reflect the isotope equilibrium fractionation factors relative to NO₂ (i.e., ${}^{\mathrm{15}}{\mathit{\alpha }}_{{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}/{\mathrm{NO}}_{\mathrm{2}}}$ and ${}^{\mathrm{15}}{\mathit{\alpha }}_{{\mathrm{NO}}_{\mathrm{3}}/{\mathrm{NO}}_{\mathrm{2}}}$; values were adopted from the Supplement Table S5 in Walters and Michalski, 2016). Finally, δ¹⁵N of particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}$ produced from dark pathways can be expressed as the following, neglecting currently unconstrained kinetic isotopic fractionation associated with Reactions (R7) and (R8) (Walters and Michalski, 2016):

$$\begin{array}{}\text{(2a)}& {\displaystyle }{\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{HNO}}_{\mathrm{3}}\right)\left(\mathrm{R}\mathrm{7}\right)={\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\right)={\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{NO}}_{\mathrm{2}}\right)+{\mathit{\epsilon }}_{\mathrm{2}},\text{(2b)}& {\displaystyle }{\mathit{\epsilon }}_{\mathrm{2}}=\left({}^{\mathrm{15}}{\mathit{\alpha }}_{{\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}/{\mathrm{NO}}_{\mathrm{2}}-\mathrm{1}}\right),\text{(2c)}& {\displaystyle }{\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{HNO}}_{\mathrm{3}}\right)\left(\mathrm{R}\mathrm{7}\right)={\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{NO}}_{\mathrm{3}}^{-}\right)\left(\mathrm{R}\mathrm{7}\right),\end{array}$$

$$\begin{array}{}\text{(3a)}& {\displaystyle }{\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{HNO}}_{\mathrm{3}}\right)\left(\mathrm{R}\mathrm{8}\right)={\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{NO}}_{\mathrm{3}}\right)={\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{NO}}_{\mathrm{2}}\right)+{\mathit{\epsilon }}_{\mathrm{3}},\text{(3b)}& {\displaystyle }{\mathit{\epsilon }}_{\mathrm{3}}=\left({}^{\mathrm{15}}{\mathit{\alpha }}_{{\mathrm{NO}}_{\mathrm{3}}/{\mathrm{NO}}_{\mathrm{2}}-\mathrm{1}}\right),\text{(3c)}& {\displaystyle }{\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{HNO}}_{\mathrm{3}}\right)\left(\mathrm{R}\mathrm{8}\right)={\mathit{\delta }}^{\mathrm{15}}\mathrm{N}\left({\mathrm{NO}}_{\mathrm{3}}^{-}\right)\left(\mathrm{R}\mathrm{8}\right).\end{array}$$

Therefore, the δ¹⁵N (NO_x) in the atmosphere can be expressed using measured δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and the net N isotope fractionation effect, ε_N.

$$\begin{array}{}\text{(4a)}& {\displaystyle }{\mathit{\delta }}^{\mathrm{15}}\mathrm{N}{\left({\mathrm{NO}}_x\right)}_{\text{atmosphere}}={\mathit{\delta }}^{\mathrm{15}}\mathrm{N}{\left({\mathrm{NO}}_{\mathrm{3}}^{-}\right)}_{\text{measured}}-{\mathit{\epsilon }}_{\mathrm{N}},\text{(4b)}& {\displaystyle }{\mathit{\epsilon }}_{\mathrm{N}}={\mathit{\epsilon }}_{\mathrm{1}}\times f_{\mathrm{1}}+{\mathit{\epsilon }}_{\mathrm{2}}\times f_{\mathrm{2}}+{\mathit{\epsilon }}_{\mathrm{3}}\times f_{\mathrm{3}},\end{array}$$

where ε₁, ε₂, and ε₃ are the abovementioned N isotope fractionation effect of pathways (i), (ii), and (iii), respectively, and the proportional contributions (f₁, f₂, and f₃) of the three ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation pathways were estimated from Δ¹⁷O measurements (Sect. 2.3.2).

2.3.2 Proportional contributions of three formation pathways to atmospheric particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}:$ Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$)

Due to its mass-independent nature, Δ¹⁷O of particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}$ is a conservative tracer of photochemical ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation (Michalski et al., 2003, 2004). At photochemical steady state, the Δ¹⁷O of NO₂ is determined by the relative production rate of NO₂ via O₃ oxidation Reaction (R1) in NO₂ production Reactions (R1) and (R2) ($f_{{\mathrm{O}}_{\mathrm{3}}}$) and the mass-independent Δ¹⁷O anomaly transferred from O₃ during Reaction (R1) (Δ¹⁷$\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast }$):

$$\begin{array}{}\text{(5)}& {\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{NO}}_{\mathrm{2}}\right)\left(\mathrm{‰}\right)=f_{{\mathrm{O}}_{\mathrm{3}}}\times {\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast }.\end{array}$$

$f_{{\mathrm{O}}_{\mathrm{3}}}$ was adopted from a previous study conducted in Beijing (Wang et al., 2019b), where the seasonal mean was 0.858 and 0.918 for summer and winter, respectively. Δ¹⁷$\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast }$ can be approximated as 1.5×Δ¹⁷O−O₃ because of the isotopic asymmetry of O₃ (Michalski and Bhattacharya, 2009). In this study, the Δ¹⁷$\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast }$ is 37.5 ‰ ± 2.2 ‰ (mean ± SD) averaged from literature sources as the isotopic composition of ozone shows a remarkable stability in the lower troposphere (Ishino et al., 2017; Vicars et al., 2012; Vicars and Savarino, 2014)

The Δ¹⁷O of particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}$ produced via the three formation pathways can be predicted by distinct Δ¹⁷O transfer functions as the following (Morin et al., 2011):

$$\begin{array}{}\text{(6a)}& \begin{array}{rl}{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{NO}}_{\mathrm{3}}^{-}\right)\left(\mathrm{R}\mathrm{4}\right)\left(\mathrm{‰}\right)& =\mathrm{2}/\mathrm{3}{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{NO}}_{\mathrm{2}}\right)\\ & =\mathrm{2}/\mathrm{3}{f}_{{\mathrm{O}}_{\mathrm{3}}}\times {\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast },\end{array}\end{array}$$

$$\begin{array}{}\text{(6b)}& \begin{array}{rl}{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{NO}}_{\mathrm{3}}^{-}\right)\left(\mathrm{R}\mathrm{7}\right)\left(\mathrm{‰}\right)& =\mathrm{5}/\mathrm{6}{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{N}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{5}}\right)\\ & =\mathrm{1}/\mathrm{3}{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{NO}}_{\mathrm{2}}\right)+\mathrm{1}/\mathrm{2}{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{NO}}_{\mathrm{3}}\right)\\ & =\mathrm{1}/\mathrm{6}{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast }\left(\mathrm{4}{f}_{{\mathrm{O}}_{\mathrm{3}}}+\mathrm{1}\right),\end{array}\end{array}$$

$$\begin{array}{}\text{(6c)}& \begin{array}{rl}{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{NO}}_{\mathrm{3}}^{-}\right)\left(\mathrm{R}\mathrm{8}\right)\left(\mathrm{‰}\right)& ={\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{NO}}_{\mathrm{3}}\right)\\ & =\mathrm{2}/\mathrm{3}{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{NO}}_{\mathrm{2}}\right)+\mathrm{1}/\mathrm{3}{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast }\\ & =\mathrm{1}/\mathrm{3}{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast }\left(\mathrm{2}{f}_{{\mathrm{O}}_{\mathrm{3}}}+\mathrm{1}\right).\end{array}\end{array}$$

Finally, the Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) can be expressed as the following:

$$\begin{array}{}\text{(7)}& \begin{array}{rl}{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}{\left({\mathrm{NO}}_{\mathrm{3}}^{-}\right)}_{\text{measured}}& ={\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{NO}}_{\mathrm{3}}^{-}\right)\left(\mathrm{R}\mathrm{4}\right)\times f_{\mathrm{1}}\\ & +{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{NO}}_{\mathrm{3}}^{-}\right)\left(\mathrm{R}\mathrm{7}\right)\times f_{\mathrm{2}}\\ & +{\mathrm{\Delta }}^{\mathrm{17}}\mathrm{O}\left({\mathrm{NO}}_{\mathrm{3}}^{-}\right)\left(\mathrm{R}\mathrm{8}\right)\times f_{\mathrm{3}},\end{array}\end{array}$$

where Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$)_measured is the measured value in this study, and the three endmember values of Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) Reaction (R4), Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) Reaction (R7), and Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) Reaction (R8) are calculated using Eqs. (6a)–(6c). The proportional contributions of the three ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation pathways ($f_{\mathrm{1}}+f_{\mathrm{2}}+f_{\mathrm{3}}=\mathrm{1}$) were estimated by the Stable Isotope Analysis in R (SIAR) model (Sect. 2.5) for the winter and the summer months.

2.4 Estimation of PM_2.5 aerosol liquid water content (ALWC) and aerosol pH

ISORROPIA II is a thermodynamic equilibrium model for the Cl⁻, ${\mathrm{NO}}_{\mathrm{3}}^{-}$, ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$, Na⁺, ${\mathrm{NH}}_{\mathrm{4}}^{+}$, K⁺, Ca²⁺, Mg²⁺, and H₂O aerosol system (Fountoukis and Nenes, 2007). In the present study, the model was run as a forward and metastable mode to calculate aerosol liquid water content (ALWC) and pH. The detailed information of the model is found in Fountoukis and Nenes (2007). As input parameters, the concentrations of water-soluble ions that were measured by NIER and ambient RH and temperature were used for the model.

2.5 Bayesian stable isotope mixing model (stable isotope analysis in R, SIAR)

For quantifying proportional contribution of three ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation pathways (f₁, f₂, and f₃ in Eq. 4b), we used the Bayesian stable isotope mixing model (Parnell et al., 2013) implemented in the simmr package in R software, which is available at https://cran.r-project.org/web/packages/simmr/index.html (last access: 10 March 2021). The Bayesian Markov chain Monte Carlo approach is adequate to provide the relative contribution of the endmembers. Detailed information of the SIAR can be found in Parnell et al. (2010). As input data, measured Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and estimated Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) endmember values of each pathway were treated in the SIAR model. Similar use of the SIAR model can be found elsewhere (Song et al., 2020; Wang et al., 2019b).

2.6 Backward air mass trajectory

Two-day air mass backward trajectories were analyzed using the HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) model with meteorological input from the global data assimilation system (a regular 1^∘ × 1 ^∘ longitude–latitude grid) (Stein et al., 2015; https://ready.arl.noaa.gov/HYSPLIT.php, last access: 27 May 2020). The trajectories were calculated every 6 h at an elevation of 500 m a.s.l. The potential source contribution function (PSCF) was then applied to calculate the probable source location (latitude i and longitude j), which is determined here as the ratio of the number of trajectory end points associated with measured δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values higher than a threshold value (set to the 95th percentile of δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) (m_ij) to the total number of points (n_ij) in the ijth grid cell). The PSCF calculation and the determination of the gridded trajectory frequencies were made in the openair R package (https://cran.r-project.org/web/packages/openair/index.html, last access: 4 January 2021) (Carslaw and Ropkins, 2012).

3 Results and discussion

3.1 Seasonal characteristics of PM_2.5 concentration, δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$), and Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$)

PM_2.5 mass concentration varied from 7.5 to 139.0 µg m⁻³ for the whole sampling period. According to the typical synoptic weather patterns of East Asia (Kim et al., 2007), the measurements are divided into summer (May–August) and winter (October–March) groups. A clear seasonal difference in PM_2.5 concentration and its composition were observed with significantly higher concentrations of mass and inorganic constituent in the winter than in the summer (Table 1 and Fig. S1).

Table 1 Measurement summary of PM_2.5 chemical constituents and isotopic compositions in Seoul during the sampling period of May 2018–March 2019. Arithmetic mean ± 1 SD (mass fraction, %) for mass concentrations and concentration-weighted mean ± 1 SD for isotope ratios.

During the summer, PM_2.5 concentration ranged from 7.5 to 34.5 µg m⁻³ with a mean of 22.7±6.9 µg m⁻³. The mean concentration of TC and TN was 4.6±1.4 µgC m⁻³ and 1.9±1.0 µgN m⁻³, respectively, resulting in the mean TC $/$ TN ratio of 2.7±2.0. The mass concentrations of all measured species were much higher in the winter, during which PM_2.5 concentration was raised up to 139.0 from 10.6 µg m⁻³ with a mean of 61.7±39.2 µg m⁻³. Accordingly, the mean of TC and TN was 15.2±4.5 µgC m⁻³ and 11.8±7.7 µgN m⁻³, respectively, and the mean TC $/$ TN ratio of 0.9±0.7 was noticeably lower than that of the summer.

For the entire experiment, the mean mass fraction against PM_2.5 was the highest for ${\mathrm{NO}}_{\mathrm{3}}^{-}$ (26 % ± 23 %), followed by ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ (20 % ± 1 %) and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ (14 % ± 1 %), highlighting the contribution of SIA to PM_2.5 mass. The ${\mathrm{NO}}_{\mathrm{3}}^{-}$ mass concentration varied from 0.2 to 69.3 µg m⁻³ with a mean of 17.6±22.1 µg m⁻³. Seasonally, the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ concentration was significantly higher in the winter (29.7±22.1 µg m⁻³) than summer (0.8±0.9 µg m⁻³). TN existed completely as inorganic N from both ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ during the winter (regression slope of 1.0). During the summer, ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and ${\mathrm{NH}}_{\mathrm{4}}^{+}$ comprised 67 % of TN, and the rest (33 %) was assumed to be organic nitrogen (ON) components. In contrast, the mass fraction of ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}$ against PM_2.5 was higher in the summer (23 %) than winter (19 %). The seasonal characteristics of chemical composition imply the significant role of inorganic nitrogen species in PM_2.5 mass increase in winter and ON and sulfate in summer.

Both δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) exhibited an inverse correlation with ambient temperature ($r=-\mathrm{0.87}$ and $r=-\mathrm{0.55}$, respectively). The δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) exhibited higher values in winter with a weight mean of −0.7 ‰ ± 3.3 ‰ and 3.8 ‰ ± 3.7 ‰ in the summer and the winter, respectively. This seasonal pattern of δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) has been typically observed in East Asia regions (Li et al., 2019; Song et al., 2019; Zong et al., 2020). Analysis of backward air mass trajectory indicates that the lower and the upper bound of δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) were associated with air masses from the ocean by southerly and easterly winds in the summer and from Siberia by northerly winds in winter, respectively (Fig. S3 in the Supplement). In comparison with urban China (Fig. 2a), averaged summer δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values were comparable between Seoul and all urban Chinese sites reported here, whereas in winter, δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) of Seoul was similar to that of Shanghai and Guangzhou rather than δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) of Beijing, which was higher than 10 ‰. Another observation at a mountain station in Taiwan shows that the highest δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) was found in spring when the level of anthropogenic constituents was elevated (Guha et al., 2017). These observations over East Asia may suggest to some extent that the seasonal pattern of δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) is basically associated with a synoptic meteorological condition that controls the type and strength of emission sources, where low and high δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values indicate biogenic soil emissions and fossil combustion, respectively (Elliott et al., 2019).

Figure 2 (a) δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and (b) Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) PM_2.5 observed in northeast Asia are compared: Seoul in this study (red); Beijing (brown), Shanghai (orange), and Guangzhou (light) in China; and Rishiri (green) and Sapporo (blue) in Japan during summer (circle) and winter (square). Annual ranges were presented for Rishiri and Sapporo. Marker indicates mean value (concentration-weighted average for Seoul samples of the present study), and lower and upper whiskers denote minimum and maximum values (this study; He et al., 2018; Lim et al., 2020; Nelson et al., 2018), mean ± SD (Song et al., 2019; Wang et al., 2019b), or 25th and 75th percentiles (Zong et al., 2020). Different marker shapes indicate different seasons.

Along with nitrogen isotope, heavier oxygen isotopes were also enriched in ${\mathrm{NO}}_{\mathrm{3}}^{-}$ during the winter compared to the summer, when the weight-mean values of δ¹⁷O, δ¹⁸O, and Δ¹⁷O in ${\mathrm{NO}}_{\mathrm{3}}^{-}$ were 70.4 ‰ ± 5.4 ‰, 82.0 ‰ ± 6.2 ‰, and 27.7 ‰ ± 2.2 ‰ for the winter and 57.3 ‰ ± 4.9 ‰, 65.7 ‰ ± 6.2 ‰, and 23.2 ‰ ± 2.2 ‰ for the summer. These results of high winter and low summer Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) were consistent with previous observations at urban Beijing (He et al., 2018; Wang et al., 2019b; Fig. 2b), indicative of relatively greater contribution of nighttime oxidation pathways in winter. It is also noteworthy that our summer and winter Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values were similar to annual Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values of an urban (Sapporo) and a rural (Rishiri) site in Japan, respectively (Nelson et al., 2018). It is likely suggestive that the winter Δ¹⁷O value in Seoul has undergone a considerable atmospheric processing on a regional scale.

Given that PM_2.5 concentrations reflect the seasonality, atmospheric chemical composition and meteorological properties were examined in relation to PM_2.5 concentrations (Fig. 3). Clearly, for the winter samples with PM_2.5 concentration greater than 40 µg m⁻³, meteorological conditions varied relative to low PM_2.5 samples; relatively high temperature and RH staying at 0 to 10^∘ and 45 % to 65 %, respectively, and a low wind speed of 1 to 2 m s⁻¹, representing the meteorological characteristics of winter PM_2.5 episodes. Particularly in winter, a strong linearity of PM_2.5 was found with most chemical constituents considered in this study, such as SIA species, TN, NOR, and SOR (sulfur oxidation ratio). However, although elevated in level, NO₂ and O₃ mixing ratios showed non-linearity with PM_2.5 concentrations. Isotope ratios were correlated either linearly or inversely with PM_2.5 concentrations. In winter, Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) showed the best correlation with PM_2.5 concentrations, but δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) was inversely related with PM_2.5 level.

Figure 3 Relationships of PM_2.5 mass concentration with meteorological parameters (a–c), concentrations of reactive gases (d–f, k) and aerosol chemical constituents (g–j, l–n), and N and O isotopic compositions (o–q) measured in Seoul during the summer (open circle) and the winter (blue ribbon).

3.2 Graphical representation of dual isotopes: Δ¹⁷O and δ¹⁵N of ${\mathrm{NO}}_{\mathrm{3}}^{-}$

The isotope ratios of source endmembers are scarce in the study region and in the aerosol measurements, and the isotope fractionation effect during gas-to-particle conversion is often estimated with thermodynamic constants and reasonable assumptions. In this regard, the graphical representation of dual isotopes enables the complex signatures of atmospheric samples to be distinguished and allows their oxidation processes to be constrained. Figure 4 displays measured δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) coordinates with related chemical parameters in colors.

Figure 4 Relationship of measured δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values with key parameters during the summer (open circle) and winter (closed ribbon). The predicted Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values are presented as dotted rectangles for the three major ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation pathways (Reactions R4, R7, and R8; see Sect. 2.3.2 for details) in (a) summer and (b) winter. Data are color coded by (a, b) nitrate concentrations, (c) NOR, (d) ALWC, (e) pH, (f) $\left[{\mathrm{NH}}_{\mathrm{4}}^{+}\right]/\left(\right[{\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}]+[{\mathrm{NO}}_{\mathrm{3}}^{-}\left]\right)$ equivalent ratio, (g) ${\mathrm{NO}}_{\mathrm{2}}/{\mathrm{O}}_{\mathrm{3}}$ ratio, and (h) $f_{{\mathrm{NO}}_{\mathrm{2}}}$. Marker size is proportional to PM_2.5 concentration ranging from 10 to 100 µg m⁻³.

In Fig. 4a and b, the range of Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) endmembers for the three ${\mathrm{NO}}_{\mathrm{3}}^{-}$ oxidation pathways calculated in Sect. 2.3.2 is presented by black dotted boxes, which are successfully distinguished. The Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values of all samples fell into the predicted ranges of the Δ¹⁷O endmembers but were separated into seasonal groups. Most of summer Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values indicate the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ Reaction (R4) pathway. It is in accordance with what has been reported in temperate areas during summer (Alexander et al., 2009; Michalski et al., 2003). In field measurements of δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and δ¹⁸O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) in Chinese cities, the proportional contribution of the OH-oxidation pathway was significantly correlated with latitudes (Zong et al., 2020), confirming the evidence of the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation through OH oxidation depending on UV radiation intensity.

In comparison, more than half of winter samples are located in the domains indicating ${\mathrm{NO}}_{\mathrm{3}}^{-}$ Reactions (R7) and (R8) pathways. In particular, for winter samples with ${\mathrm{NO}}_{\mathrm{3}}^{-}$ mass concentration higher than ∼30 µg m⁻³, Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values clearly represent the Reaction (R7) pathway. The winter Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) was proportional to the NOR (Fig. 4c), which is indicative of the efficient conversion of NO_x to ${\mathrm{NO}}_{\mathrm{3}}^{-}$ via O₃ oxidation pathways. In the summer periods when ${\mathrm{NO}}_{\mathrm{3}}^{-}$ concentrations were relatively high (above ∼3 µg m⁻³), the contributions of O₃ oxidation pathways became elevated, which is a similar pattern to what was observed in the winter. This result suggests that the O₃ oxidation pathways are likely to be responsible for the rapid conversion of particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}$ observed during winter and summer PM_2.5 episodes in the study region.

The formation of N₂O₅ Reaction (R6) is dependent on both NO₂ and O₃, which are reciprocally correlated in source regions. The equilibrium of Reaction (R6) shifts to N₂O₅ at low temperature, and its lifetime against photolysis is long under the low sunlight. For these reasons, the dark formation of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ Reaction (R7) would be favorable in urban outflows in winter. The hydrolysis of N₂O₅ on aerosol surface Reaction (R7) is highly dependent on RH, aerosol liquid water content (ALWC), and chemical composition (Hallquist et al., 2003; Wahner et al., 1998). To evaluate the Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) signature observed, we calculated ALWC and acidity of aerosol, pH, using the ISORROPIA II model (Sect. 2.4). Given the high concentration of nano-particles from various sources in the urban areas, it is assured that aerosol surface is enough for the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ Reaction (R7) pathway. The calculated ALWC was higher by 90 % (19.1±22.8 µg m⁻³) in the winter and lower by 21 % (7.8±5.2 µg m⁻³) in the summer relative to the annual mean of the 2 years. When ${\mathrm{NO}}_{\mathrm{3}}^{-}$ mass concentrations were greater than ∼3 µg m⁻³ in the summer and ∼30 µg m⁻³ in the winter, ALWC was 14.3 and 31.7 µg m⁻³, respectively, which were greater by 166 % and 183 % than the seasonal mean, respectively (Fig. 4d). Furthermore, the high Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and ALWC coincided with the lower bound (about 4) of winter pH varying from 4 to 6, whereas aerosol pH was 2–3 during the summer. The low aerosol pH concurrent with the high Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and ALWC is coherent with the aqueous-phase HNO₃ formation on liquid aerosol through Reaction (R7).

While Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) allows insights into the oxidation pathway of ${\mathrm{NO}}_{\mathrm{3}}^{-}$, δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values are affected by various factors, most of which are not well constrained due to the complicated chemistry of nitrogen oxides in the Earth's environment. In this study, δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) variability versus PM_2.5 concentration resembled the variability of NO₂ versus PM_2.5 concentration (Fig. 3f and o), implying a close link between NO₂ oxidation and δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$). Therefore, in addition to the role of Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) as a tracer of an ${\mathrm{NO}}_{\mathrm{3}}^{-}$ oxidation pathway, we evaluated whether a variation of δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) indicates NO_x oxidation efficiency between NO and NO₂ (Freyer et al., 1993; Nelson et al., 2018; Walters et al., 2016). A possible effect on δ¹⁵N during NO_x photochemical cycling cannot be predicted in the same way under varying conditions of NO, NO₂, and O₃ abundances. The shift in δ¹⁵N (NO₂) relative to δ¹⁵N (NO_x) in the atmosphere will change depending on $f_{{\mathrm{NO}}_{\mathrm{2}}}$ and temperature-dependent ${\mathit{\alpha }}_{{\mathrm{NO}}_{\mathrm{2}}/\mathrm{NO}}$ (Freyer et al., 1993; Li et al., 2020; Walters et al., 2016). When O₃ mixing ratio is high, NO is almost completely oxidized to NO₂, leading to an increasing $f_{{\mathrm{NO}}_{\mathrm{2}}}$ value, and the δ¹⁵N (NO₂) should correspond to the δ¹⁵N of NO_x sources (Freyer et al., 1993).

In the dual isotope coordinates of δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) (Fig. 4), it is evident that the samples taken during a record-breaking winter's PM_2.5 pollution events are associated with the lower bound values of δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) (−1 ‰ to 4 ‰) and the highest Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values (28 ‰ to 31 ‰). Simultaneous measurements of PM₁ chemical composition in Seoul and Beijing demonstrated that the regionally processed air masses were long-range transported to Seoul within approximately 2 d during these episodes (H. Kim et al., 2020). Their ${\mathrm{NO}}_{\mathrm{2}}/{\mathrm{O}}_{\mathrm{3}}$ ratio (1 to 2) was clearly lower with higher $f_{{\mathrm{NO}}_{\mathrm{2}}}$ (0.7 to 0.8) compared to the other winter samples (Fig. 4g and h). This result implies that O₃ level was high enough to efficiently oxidize NO to NO₂ during the severe PM_2.5 pollution events. In this condition, the shift in δ¹⁵N (NO₂) relative to δ¹⁵N (NO_x) is insignificant, and, consequently, δ¹⁵N (NO₂) would be lower than those of other winter samples unless both emission sources and ${\mathit{\alpha }}_{{\mathrm{NO}}_{\mathrm{2}}/\mathrm{NO}}$ changed significantly. Considering that ${\mathrm{NO}}_{\mathrm{3}}^{-}$ is the key driver of the high PM_2.5 in Seoul, the higher degree of NO_x oxidation efficiency is worth highlighting, in conjunction with a strong linear relationship between Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and NOR (Fig. 4c) revealing an efficient conversion of NO₂ to ${\mathrm{NO}}_{\mathrm{3}}^{-}$ through ${\mathrm{NO}}_{\mathrm{3}}^{-}$ pathway Reactions (R5)+(R6)+(R7)+(R8) via O₃.

3.3 Contributions of major HNO₃ oxidation pathways

Combining the Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) measurements and the calculation of Δ¹⁷O transferred from O₃ to HNO₃, the contributions of three major ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation pathways were quantitatively accounted for, despite the inherent uncertainties in the calculation (Sect. 2.5).

${\mathrm{NO}}_{\mathrm{3}}^{-}$ pathway Reaction (R4) dominated the total ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation (87 % ± 6 %) in summer. In contrast, the nighttime pathways through N₂O₅ and NO₃ (pathways Reactions R7 and R8) were responsible for 24 % and 14 % of the ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation in the winter, respectively. The contributions increased further to 40 % and 30 %, respectively, on haze days when PM_2.5 concentration exceeded 75 µg m⁻³, the “very bad” alert of the national air quality standard. The significant nighttime oxidation of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ has been observed broadly at urban sites in northeast Asia. The results of this study are consistent with those conducted in Beijing, showing low Δ¹⁷O values in summertime (about 17 ‰ to 25 ‰) and substantially high Δ¹⁷O values (about 25 ‰ to 34 ‰ ) in wintertime when ${\mathrm{NO}}_{\mathrm{3}}^{-}$-driven PM_2.5 haze pollution occurred (He et al., 2018; Song et al., 2020; Wang et al., 2019b). In general, nighttime pathways accounted for more than 60 % and up to 97 % in Beijing (He et al., 2018; Song et al., 2020; Wang et al., 2019b). From a global perspective, the chemical transport model demonstrated that N₂O₅ heterogeneous hydrolysis was of comparable importance to NO₂+OH (41 %) for ${\mathrm{NO}}_{\mathrm{3}}^{-}$ formation at below 1 km altitude (Alexander et al., 2020).

It is noteworthy that although the seasonal patterns are similar in northeast Asia, the average proportional contributions estimated from Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) are highly sensitive to input parameters (i.e, $f_{{\mathrm{O}}_{\mathrm{3}}}$ and Δ¹⁷$\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast }$). In this study, the seasonal $f_{{\mathrm{O}}_{\mathrm{3}}}$ was assumed to be equal to those estimated for Beijing and set to 0.858±0.05 and 0.918±0.05 for the warm and cold months, respectively (Wang et al., 2019b), which are comparable to other estimates for Beijing (0.86) and Shanghai (0.97) (He et al., 2018, 2020) and the annual mean of 0.85 in a global model (Alexander et al., 2020).

The uncertainty associated with Δ¹⁷$\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast }$ has been recognized as the largest source of uncertainty in estimating ${\mathrm{NO}}_{\mathrm{3}}^{-}$ production pathways from Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) (Alexander et al., 2009, 2020). In this study, the Δ¹⁷$\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast }$ of 37.5 ‰ ± 2.2 ‰ was averaged from previous observations, corresponding to Δ¹⁷O-bulk O₃ of 25 ‰ (Ishino et al., 2017; Vicars et al., 2012; Vicars and Savarino, 2014). Our mean Δ¹⁷$\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast }$ of 37.5 ‰ was 2.8 ‰ higher and 1.5 ‰ lower than what was used in the field studies (Song et al., 2020; Wang et al., 2019b) and other studies (He et al., 2018, 2020), respectively, for urban Beijing and Shanghai. When a sensitivity test was conducted for the proportional contribution of the three oxidation pathways, a 2.8 ‰ change in Δ¹⁷$\mathrm{O}-{\mathrm{O}}_{\mathrm{3}}^{\ast }$ value caused a 1.6 ‰, 2.1 ‰, and 2.5 ‰ change in the endmember for the pathways for Reactions (R4), (R7), and (R8), respectively. As a result, the average contribution of the nighttime pathways, including ${\mathrm{NO}}_{\mathrm{3}}^{-}$ Reaction (R7) and ${\mathrm{NO}}_{\mathrm{3}}^{-}$ Reaction (R8), increased to 23 % in summer and 65 % in winter. This suggests that a proper use of key parameters driving endmember values is a prerequisite for more realistic quantification of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ oxidation pathway contributions.

3.4 Major NO_x emission sources

To investigate major emission sources of atmospheric NO_x in Seoul, δ¹⁵N (NO_x) in the atmosphere (δ¹⁵N (NO_x)_atmosphere) was estimated from measured δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) considering nitrogen isotopic fractionation effects (ε_N) based on a dual isotopes approach (Δ¹⁷O and δ¹⁵N). Estimated ε_N values were 5.9 ‰ ± 1.5 ‰ and 12.2 ‰± 0.5 ‰, and accordingly the δ¹⁵N (NO_x)_atmosphere was −8.7 ‰± 3.3 ‰ and −5.8 ‰± 4.2 ‰ in summer and winter, respectively. The larger winter ε_N reveals the enhanced contribution of the nighttime oxidation pathway via N₂O₅. As a result, a seasonal difference in δ¹⁵N (NO_x)_atmosphere was as small as 2.9 ‰ on average, which is suggestive of little seasonal difference in major NO_x emission sources in Seoul.

Figure 5 Variations in measured δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) (open) and estimated δ¹⁵N (NO_x)_atmosphere (closed) in Seoul during the summer (circle) and winter (triangle). The δ¹⁵N (NO_x) range (mean ± 1 SD) of source endmembers is distinguished by dotted boxes in different colors: coal combustion (14.2 ‰±4.5 ‰; Felix et al., 2012; Heaton, 1990), biomass burning (1.8 ‰ ± 1.8 ‰; Fibiger and Hastings, 2016), vehicle emissions (−2.5 ‰ ± 1.5 ‰; Walters et al., 2015), and biogenic soil emissions (−35.1 ‰ ± 10.2 ‰; Felix and Elliott, 2014; Li and Wang, 2008; Yu and Elliott, 2017).

Figure 5 displays the measured δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and estimated δ¹⁵N (NO_x)_atmosphere of individual samples together with δ¹⁵N (NO_x) domains of emission source endmembers reported in literature: biogenic soil (driven by fertilizer use, −35.1 ‰± 10.2 ‰; Felix and Elliott, 2014; Li and Wang, 2008; Yu and Elliott, 2017), biomass burning (1.8 ‰ ± 1.8 ‰; Fibiger and Hastings, 2016), vehicle emissions (−2.5 ‰ ± 1.5 ‰; Walters et al., 2015), and coal combustion (14.2 ‰ ± 4.5 ‰; Felix et al., 2012; Heaton, 1990). Both in summer and winter, δ¹⁵N (NO_x)_atmosphere values are the closest to the domain of vehicle emissions δ¹⁵N (NO_x), highlighting the largest contribution of vehicle emissions to NO_x in Seoul (Fig. 5). However, the winter δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) without isotope fractionation effect apparently points to coal emissions, which could lead to misleading conclusions about major NO_x sources. This result is supported by a better correlation of NO₂ with CO than with SO₂ for both seasons. According to an (anthropogenic) bottom-up emissions inventory, the Clean Air Policy Support System (CAPSS), total mobile sources account for about two-thirds of NO_x emissions in Korea, followed by combustion sources such as energy and manufacturing industries (33 %). The highest NO₂ column densities are distinct in Seoul metropolitan areas, which is mostly due to the emissions from transportation (H. C. Kim et al., 2020). Our results of the isotope measurements are in fairly good agreement with the national emission inventories and satellite observations, highlighting the largest contribution of vehicle emissions to NO_x sources in Seoul.

Other than vehicle emissions, the lower bound summer δ¹⁵N (NO_x)_atmosphere and the upper bound winter δ¹⁵N (NO_x)_atmosphere were relatively closer to the δ¹⁵N (NO_x) of fertilized soil emissions and coal combustion, respectively. These seasonally distinct contributions from soil emissions and coal combustion were consistent with the results from previous studies conducted in China, revealing common seasonal emission characteristics on a regional scale in East Asia. It was evident that the upper bound winter δ¹⁵N (NO_x)_atmosphere coincided with an elevated ${\mathrm{SO}}_{\mathrm{4}}^{\mathrm{2}-}/{\mathrm{NO}}_{\mathrm{3}}^{-}$ mass ratio by 2–3 times. During these periods, trajectory analysis indicates that air masses originated from cold regions in northeastern China, such as Liaoning province, where heavy and coal-fired industries are located. In Beijing, located in northeastern China, coal combustion is an important fossil-fuel source as the highest contribution source to atmospheric NO_x emissions during winter, about 30 % to 40 %, regardless of approaches used for estimating ${\mathrm{NO}}_{\mathrm{3}}^{-}$ oxidation pathways, either Δ¹⁷O or δ¹⁸O (Li et al., 2019; Song et al., 2019, 2020; Zong et al., 2020). Thus, in this study, the upper bound winter δ¹⁵N (NO_x)_atmosphere suggests enhanced contributions of coal combustion to atmospheric NO_x emissions. It is particularly noteworthy that the lower winter δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and thus relatively lower contribution of coal combustion of the present study confirms the recent trend that emissions from coal combustion have been reduced in China (Cheng et al., 2019; Tong et al., 2018). Our study was conducted in later years (2018–2019) than their studies (2013–2017), in which δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values were rather comparable to the winter mean δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) observed in Seoul during January 2014 to February 2016 (11.9 ‰ ± 2.5 ‰; Park et al., 2018). The lower bound of winter δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) values associated with the highest ${\mathrm{NO}}_{\mathrm{3}}^{-}$ and PM_2.5 concentrations (Figs. 3 and 4) was considered a result of complex effects of isotopic fractionation, as discussed above.

This study region is under the influence of various biomass burning sources throughout a year, such as agricultural combustion in the vicinities of Seoul and over eastern China from spring to fall (Chen et al., 2017; Zhao et al., 2017), wild fires over Siberia and the Russian Far East in summer (van der Werf et al., 2010), and residential biomass combustion for heating over east Asia in winter. Some winter δ¹⁵N (NO_x)_atmosphere values fell in the δ¹⁵N (NO_x) range of biomass burning endmembers (Fibiger and Hastings, 2016; Fig. 5). When assessing contributions of biomass burning based on few endmember δ¹⁵N (NO_x), however, caution should be exerted considering the fact that δ¹⁵N (NO_x) varied among biomass types from −7 ‰ to 12 ‰ (Fibiger and Hastings, 2016), and there is currently a lack of understanding of biomass combustion that could potentially affect air quality in Seoul.

4 Conclusions and outlook

Our δ¹⁵N- and Δ¹⁷O-based study confirms that vehicle emissions are the main source of ${\mathrm{NO}}_{\mathrm{3}}^{-}$ in PM_2.5 in Seoul during the summer and the winter. In addition, the qualitative estimates of NO_x emission sources provide suggestive evidence for enhanced contributions from coal combustion and biogenic soil emissions in the winter and the summer, respectively. Moreover, severe winter haze events with daily PM_2.5 exceeding 100 µg m⁻³ were mainly driven by ${\mathrm{NO}}_{\mathrm{3}}^{-}$ (up to ∼60 % in PM_2.5). Reducing NO_x emissions from vehicles is, therefore, essential for an effective mitigation measure to improve fine aerosol pollution in the study region. In particular, the highest PM_2.5 was concurrent with the lowest δ¹⁵N (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and the highest Δ¹⁷O (${\mathrm{NO}}_{\mathrm{3}}^{-}$) and NOR, revealing the efficient NO-to-NO₂ conversion and heterogeneous conversion of HNO₃ to particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}$ through the O₃ oxidation pathway during the winter haze episodes. The multiple-isotope signatures of particulate ${\mathrm{NO}}_{\mathrm{3}}^{-}$, including ¹⁷O, ¹⁸O, and ¹⁵N, highlight the secondary nature of fine aerosol pollution intimately coupled with the photochemical oxidation process.

While our results demonstrate that δ¹⁵N and Δ¹⁷O are robust tracers for major NO_x sources, quantitative source apportionment using the isotope method requires further elaboration of isotope equilibrium and/or kinetic fractionation effects involved in photochemical cycling of nitrogen oxides and δ¹⁵N of NO_x source endmembers representing local or regional emissions in East Asia. In well-designed field studies, the δ¹⁵N and Δ¹⁷O measurements of multiphase and their vertical structures allow us to test the isotope fractionation effects suggested by laboratory experiments and theoretical calculations, as well as to characterize the atmospheric processing that influences them. In addition, there is an urgent need to document the δ¹⁵N (NO_x) values of emissions from vehicles with/without selective catalytic reduction (SCR) and from biomass combustion as a function of biomass type and combustion conditions. Consequently, a comprehensive and quantitative understanding of the oxidation pathways and emission sources of nitrogen oxides using δ¹⁵N and Δ¹⁷O measurements will be able to elucidate the detailed mechanisms driving severe haze development in megacities of northeast Asia, including Seoul.