Introduction

Particulate matter pollution is a worldwide environmental problem, especially in winter in the North China Plain1,2. CHON are critical organic components with substantial impacts on particulate matter pollution, atmospheric ozone cycling3, Earth’s radiative balance4,5, and human health6.

Field measurements indicate that the proportion of CHON molecules can exceed 60% within organic matrices7, drawing increasing attention. The proportion of CHON species can vary with seasonal changes, climatic conditions, and different pollution sources. In winter, due to the increase in coal burning and heating activities, the proportion and concentration of CHON species in northern China are often higher than those in southern regions8. CHON compounds are mainly formed through secondary processes, with key precursors including isoprene and terpenes from natural sources, as well as anthropogenic VOCs such as benzene derivatives (benzene, toluene, phenol), aliphatic alkanes, and lignin pyrolysis products (phenol, m-cresol, and guaiacol)9,10,11,12,13. These VOCs can undergo various reactions to form CHON, including gas-phase, aqueous, and heterogeneous reactions, influenced by agents such as OH radicals (·OH), NO3 radicals (·NO3), nitrite/nitrous acid (NO2−/HNO2), and nitronium ions (NO2+)14,15,16,17,18,19,20. However, our understanding of the molecular composition and formation mechanism of CHON in the real atmosphere remains limited.

In recent years, the proportion of CHON in organic aerosol has significantly increased in eastern China and the North China Plain, highlighting their growing importance in regional particulate pollution7,8,21,22,23,24. High-resolution mass spectrometry has been widely applied for the qualitative characterization of atmospheric organic aerosols due to its high mass resolution and high mass accuracy25,26. In this study, ultra-high performance liquid chromatography coupled with quadrupole Orbitrap mass spectrometry (UHPLC-Orbitrap MS) has been employed to analyze the molecular composition of organic species, particularly CHON compounds, in winter PM2.5 samples collected in Beijing. Furthermore, correlation analyses between the formula number and peak abundance of CHON molecules and various environmental parameters were performed to explore their potential formation pathways.

Results and discussion

Overview of winter haze and organic aerosols

Figure S1 illustrates the temporal variation of PM2.5, gaseous pollutants, aerosol water content (AWC), and meteorological conditions during clean and haze periods in Beijing winter of 2016. During the haze period, the average PM2.5 concentration reached 187.4 μg m−3, about 6 times higher than that during the clean period. The haze episode was accompanied by high humidity, low O3 concentration, and low wind speed, consistent with the previous report27. Additionally, the elevated NO2 levels during the haze could provide abundant precursors and oxidants for the formation of secondary compounds, further exacerbating the pollution28,29. PM2.5 concentration exhibited a diurnal pattern, with lower concentrations during the daytime and higher concentrations at nighttime, in line with the trend of relative humidity (RH) and AWC.

Table S1 summarizes the number of molecular formulas of six classes of organic species identified in the ESI- and ESI+ modes using Orbitrap, without distinguishing isomers. In total, over 85% of molecules belong to oxygen-containing categories (CHO, CHON, CHOS, and CHONS), of which CHON was dominant, accounting for over 40%. Due to the preference for the detection type, more molecules were detected in the ESI+ mode, as well as a higher proportion of CHON. Previous studies also demonstrated that CHON compounds substantially contribute to particulate matter in Chinese cities, emphasizing their critical role in regional air pollution7,8,21,22,23,24. Figure 1 shows the comparison of molecular composition between clean and haze, as well as daytime and nighttime. The formula number of total CHON molecules increased dramatically from clean to haze periods, particularly in the m/z range of 250–450 (Figure S2). During the haze period, the formula number of unique molecules was approximately 6.3 times higher than in the clean period. CHON was the main driver of this increase, accounting for 42% of haze-unique molecules. In addition, more molecules were detected at nighttime than during daytime in both clean and haze periods, suggesting more complex chemical processes at nighttime.

Fig. 1: Differences in the molecular composition of the CHONS, CHOS, CHON, CHO, CHNS, and CHN from clean to haze period and day-night contrast in the combined ESI- and ESI+ modes.
figure 1

Overlapping molecules between samples (clean and haze; CD, HD, CN, and HN) were marked with shadow. Pie charts demonstrate the category proportion of unique molecules.

Full size image

Molecular characteristics of CHON species

Elemental distribution of CHON molecules varied between positive and negative ESI modes in both daily and half-day samples (Figs. S3-S4). In ESI- mode, CHON molecules were predominantly characterized by C7-C20 (more than 80%) and were largely N1 compounds, with a high degree of O/N ratio of 2.8, suggesting the dominance of -NO2 or -ONO2 functional groups30. In contrast, CHON species detected in ESI+ mode displayed higher carbon numbers (C8-C23) and a broader nitrogen distribution, encompassing both N1 and N2 compounds. The average O/N ratio was lower at 1.8, and 62.4% of CHON species exhibited an O/N ratio below 2. This indicated that the nitrogen-containing groups in ESI+ mode were predominantly imine (N = H) or amine (N-H) groups, rather than -NO2 or -ONO2 groups. Carbonyl-containing precursors may undergo aqueous reactions with ammonium or amines to form heterocyclic compounds with N = H or N-H functional groups, which can subsequently incorporate more nitrogen atoms through intermolecular acetal reactions31,32,33. The O/N ratio observed in this study is similar to values reported for urban China, yet remains lower than those found in European and North American studies34,35,36.

CHON molecules were predominantly anthropogenic aromatic compounds in all samples, highlighting the significant impact of human activities on organic compound generation in Beijing. In general, the formula number of aromatic molecules was 3.5 and 2.4 times that of aliphatic molecules in ESI- and ESI+ modes, respectively (Fig. S5). During the haze period, both aliphatic and aromatic molecules increased compared to the clean period, but the rise in aromatic compounds was more pronounced. Comparing the molecules that are commonly present in the clean and haze samples, we found that the haze-unique aromatic compounds were primarily monocyclic and bicyclic in ESI-, while in ESI + , they were predominantly monocyclic, bicyclic, and tricyclic (Fig. 2). Aromatic compounds typically originate from anthropogenic activities36. Aromatic CHON products can be formed from the reaction of anthropogenic VOCs in the presence of NOx, facilitated by ·OH gas-phase chemistry during the daytime, ·NO3 radical-mediated reactions at night, and electrophilic nitration under acidic aqueous-phase conditions17,20,29. CHON molecules detected in the ESI- mode in Beijing exhibited a higher H/C ratio (~1.0) and a lower O/C ratio (~0.4) compared to those reported in other Chinese cities such as Changchun, Shanghai, and Guangzhou8,21. This suggests a lower degree of unsaturation and aging in Beijing aerosols. Compared with the ESI- mode in this study, CHON molecules detected in the ESI+ mode showed an even higher H/C ratio (~1.4) and a lower O/C ratio (~0.2), which are comparable to values observed in other cities. We further investigated 14 representative CHON compounds in which the concentrations of anthropogenic compounds were nearly 2-4 orders of magnitude higher than those from natural sources. All anthropogenic CHON compounds were ubiquitous throughout the haze event, with their abundance proportion in the overall CHON abundance rising significantly from 34.6% during the clean period to 51.8% during the haze period. These results confirmed that anthropogenic aromatic molecules are the primary CHON components.

Fig. 2: Characteristics of CHON molecules during clean and haze periods.
figure 2

KMD diagrams and Van Krevelen (VK) diagrams of ubiquitous CHON molecules (a, b) in the clean period and (c, d) in the haze period in ESI- mode. KMD diagrams and VK diagrams of ubiquitous CHON molecules (e, f) in the clean period and (g, h) in the haze period in ESI+ mode. The green dotted circles indicate CHON molecules that newly appeared during the haze period in ESI- mode and ESI+ mode, respectively. The color bar denotes the aromaticity equivalent.

Full size image

Nocturnal and aqueous formation of CHON species

Figure 3 shows the correlations between the number of CHON molecular formulas and various environmental parameters in both ESI modes for daily samples and half-day samples, respectively. The CHON species correlated well with PM2.5 concentrations. The good correlation between CHON and NO2 suggests that NO2, either directly or indirectly, promotes CHON formation. During the daytime, VOCs are initially oxidized by hydroxyl radicals (·OH), leading to the formation of peroxy radicals (RO2·) or phenoxy radicals. These intermediates subsequently react with NOx to yield CHON species. The oxidation occurs under sunlight, and the NOx absorbs photons with wavelengths shorter than 420 nm16. At night, O3 can oxidize NO2 to generate ·NO3 radicals, which in turn initiate oxidation reactions of VOCs. These ·NO3-driven pathways are analogous to the ·OH-initiated processes occurring during the daytime. For example, ·NO3 radicals are highly reactive with unsaturated hydrocarbons such as alkenes and biogenic terpenes, forming nitrooxyalkyl radicals, which can then yield nitrooxyalkyl peroxy radicals (R(NO2)O2·) upon reaction with oxygen37,38. Recent research also indicates a significant rise in ·NO3 radical concentrations in China39,40, promoting the nocturnal chemistry.

Fig. 3: Influencing factors of CHON formation.
figure 3

Correlation of the formula number of total, aromatic, and aliphatic CHON species with various parameters (PM2.5, NO2, RH, AWC, O3) in both modes for (a) daily samples and (b) half-day samples, respectively. The color bar shows the degree of correlation between various parameters. *p < 0.05, **p < 0.01, ***p < 0.001.

Full size image

Aqueous chemistry played a crucial role in the formation of CHON compounds. The increase in CHON molecules is strongly linked to RH and AWC, indicating the importance of aqueous chemistry. Some high-abundance molecules identified in this study are closely linked to aqueous-phase formation pathways. Aqueous reactions under acidic conditions are a significant pathway for the formation of the nitrobenzene series20,41. In the presence of NO2, Kyodai nitration drives the transformation of phenol (C6H6O) into nitrophenol (C6H5NO3)10. And mononitro compounds, such as mononitrophenol (C6H5NO3) and mononitrocatechol (C7H7NO4), can undergo further radical nitration reactions to produce dinitro products42,43. Additionally, methylcatechol (C7H8O2) can undergo the aqueous electrophilic substitution to form the methylnitrocatechol (C7H7NO4) within acid particles20,44. Besides the CHON compound in ESI- mode, the aqueous reactions between carbonyls and ammonium/amine are key routes for the formation of nitrogen-containing heterocycles detected in the ESI+ mode33. CHON species did not show an expected positive correlation with O3, which indicates that O3 is not a key factor influencing the formation of CHON.

More CHON molecules were produced at night. CHON molecules exhibited clear diurnal variation, with both formula number and abundance being higher at night than during the day (Fig. S6). This trend closely followed changes in RH and AWC (Fig. S1), and correlation analysis confirmed strong positive relationships between CHON abundance and these parameters (Fig. S7). Notably, such patterns were more pronounced in anthropogenic CHON compounds than in natural ones. In contrast, NO2 and O3 showed differences between clean and haze conditions, but did not display notable diurnal variation. These findings confirm that nocturnal and aqueous chemistry promoted the formation of CHON compounds. In the future, more field, laboratory, and modeling research is needed to elucidate the formation mechanism of CHON in ambient air.

Methods

Field observation and UHPLC-Orbitrap MS analysis

We conducted online field measurements and offline PM2.5 sampling from 01–22 December 2016 in Beijing (Table S2). The online field observations are detailed in supporting information (SI, Text S1). PM2.5 samples were collected on 90 mm quartz filters at a flow rate of 100 L min−1 with a middle-volume air sampler (Laoying 2030, Qingdao, China). The filters were baked in a Muffle furnace at 550 °C for 6 h, then put in the cassettes and wrapped in aluminum foil before sampling, and all samples were stored at −20 °C before analysis. In total, 12 daily samples, 10 half-day samples, and 1 blank sample were acquired. The detailed information about sample pretreatment is described in SI (Text S2). The organic molecular composition was analyzed using an ultra-high performance liquid chromatograph (Dionex UltiMate 3000, USA) coupled to a Q-Exactive Orbitrap MS (Thermo Fisher Scientific, Germany)45. The electrospray ionization source was operated in both negative (ESI-) mode and positive (ESI + ) mode. Because of the diverse ionization mechanisms, the types of functional groups measured by the two modes are different. The ESI- mode provides better detection of acidic deprotonated functional groups (e.g., -NO2, -ONO2, and -COOH), while the ESI+ mode tends to measure basic protonated functional groups such as -NH246. More details about the measurement method were described in SI (Text S3).

Data processing

Raw chromatogram-mass spectrometry data were processed with MZmine software (http://mzmine.github.io) to obtain the m/z value (with a mass tolerance of ±2 ppm), formulas, retention times, and peak areas of detected organic compounds47. The detailed processing steps and settings have been reported in previous literature25,48. Formula assignment was subject to the following limitations: C ≤ 50, H ≤ 160, O ≤ 80, N ≤ 5, S ≤ 3. To remove chemically meaningless molecules, formulas were further constrained: H/C and O/C were in the range of 0.3–3 and 0–3, respectively; in ESI-, N/C and S/C were in the range of 0–0.5 and 0–0.2, respectively; in ESI + , these two ratios were in the range of 0–1.3 and 0–0.8, respectively21,49. The abundance of a compound in the following discussion refers to its chromatographic peak area and was corrected by subtracting blanks. In addition, the abundance of each sample was corrected by sampling volume for inter-sample comparison.

For the chemical formula CcHhOoNnSs, the double bond equivalent (DBE) is calculated as DBE = (2c + 2 + n - h)/2, and the aromaticity equivalent (Xc) is calculated as Xc = (3DBE - 1.5o - 2) / (DBE - 0.5o). Xc has been suggested to separate aliphatic compounds (Xc < 2.5) and aromatic compounds (Xc ≥ 2.5)25,50. Notably, certain aromatic molecules, such as C6H5NO3, C7H7NO3, and other nitrophenols, have Xc values below 2.5 when calculated using traditional formulas, leading to an overestimation of the aliphatic content30. The aromatic compounds were further divided into monocyclic compounds (benzene-core structure, 2.5 ≤ Xc < 2.714), bicyclic compounds (naphthalene-core structure, 2.714 ≤ Xc < 2.8), tricyclic compounds (anthracene-core structure, 2.8 ≤ Xc < 2.833), and tetracyclic compounds (pyrene-core structure, 2.833 ≤ Xc < 2.923). Kendrick mass defect (KMD) is effective in distinguishing similar compound groups from numerous molecular formulas. In this study, CH2 was chosen as a base unit, and KMD-CH2 was calculated as follows (Eqs. 1 and 2)51:

$$\mathrm{Kendrick\; mass}\,(\mathrm{KM})=\mathrm{observed\; mass}\times \frac{14}{14.01565}$$
(1)
$$\mathrm{KMD}-{\rm{C}}{{\rm{H}}}_{2}=\mathrm{normal\; mass}-\mathrm{KM}$$
(2)

where observed mass represents the precise molecular mass measured by the instrument, and nominal mass represents the molecular mass rounded to the nearest integer value.

We identified 14 specific compounds from both anthropogenic and natural sources based on their high concentrations in PM2.5 samples and significance in the atmosphere (Table S3)52,53,54,55, and analyzed their differences between clean and haze samples. These compounds are categorized into anthropogenic CHON compounds (including C6H5NO3, C6H4N2O5, C7H7NO3, C7H6N2O5, C7H5NO5, C6H5NO4, C6H4N2O7, C7H7NO4, C7H6N2O6, C9H11NO3, and C9H11NO4), and natural CHON compounds (including C5H9NO4, C5H7NO4, and C10H16N2O8)56,57.