New measurements reveal a large contribution of nitrogenous molecules to ambient organic aerosol

The study addresses the longstanding gap in quantifying the total mass and sources of organic nitrogen (ON) within ambient organic aerosol (OA). While OA can represent 20–90% of fine aerosol mass, prior research has been largely carbon-centric, with limited attention to nitrogen-containing organic aerosols (OAN). Ultrahigh-resolution mass spectrometry has revealed numerous N-containing molecules in OA, and some ON species are light-absorbing or active in new particle and secondary OA formation, implicating them in climate and air quality. However, the total aerosol ON (including both water-soluble and water-insoluble fractions) has remained poorly quantified due to analytical limitations. The purpose of this work is to quantitatively survey aerosol ON across diverse environments, estimate the contribution of OAN to total OA using joint ON-OC measurements and Monte Carlo analysis, and apportion the sources of ON using receptor modeling. The study’s importance lies in improving understanding of OA composition, refining models of aerosol impacts, and informing assessments of atmospheric nitrogen deposition to ecosystems.

Previous quantitative knowledge of aerosol ON focused on limited compound classes (e.g., urea, small amines, amino acids, nitroaromatics, nitro-PAHs) or subgroups like organic nitrates measured by AMS or TD-LIF. Bulk ON measurements have often targeted only the water-soluble fraction (WSON), calculated as WSTN minus inorganic nitrogen (IN), leaving water-insoluble ON (WION) largely unconstrained. Some attempts used elemental analyzers to measure total N and infer ON by difference, but detection limits and error propagation compromised accuracy, especially when ON is a minor fraction. Recent advances include an aerosol IN&ON analyzer employing programmed thermal evolution with chemiluminescent detection and multivariate curve resolution, enabling sensitive, simultaneous quantification of ON and IN without pretreatment. Literature also documents that ON can significantly contribute to light-absorbing "brown carbon," new particle formation, and secondary organic aerosol formation, and that anthropogenic nitrogen deposition affects the global nitrogen cycle. Despite ultrahigh-resolution mass spectrometry revealing numerous N-containing molecules in OA, comprehensive bulk ON abundance, its partitioning (WSON vs WION), and source contributions have remained insufficiently quantified.

- Sampling: Over 600 PM2.5 filter samples were collected across 12 sites in China (North China Plain and Pearl River Delta), spanning urban, suburban, and rural environments. Sites included Beijing (BJ), Qingdao (QD), Guangzhou (GZ), Nanhai (NH), Dongguan (DG), Guangzhou Institute of Geochemistry (GIG), Tsuen Wan (TW), Yuen Long (YL), Mong Kok (MK), Nansha (NS), Dinghushan (DHS), and HKUST (UST). Year-long datasets and comprehensive speciation were available at TW (urban), NS (suburban/rural), and BJ (suburban). Quartz fiber and Teflon filters were collected using calibrated high- or mid-volume samplers; field blanks were taken; samples were stored at −20 °C and transported cold. IN stability over storage was confirmed via strong agreement (R2=0.96) between ion chromatography and later IN&ON analyzer measurements. - Chemical analyses: The aerosol IN&ON analyzer quantified inorganic and organic nitrogen simultaneously by programmed thermal evolution with chemiluminescent detection and multivariate curve resolution. ON, IN, OC, EC, and major ions were measured for all samples; additional source markers (non-polar organics, sugars, fatty acids, sterols, SOA tracers) were measured for BJ, TW, NS, YL, and MK. - Estimation of OAN contribution to OA: Using co-located OC and ON, OAN/OA was modeled as (OAN/ON) × (ON/OC) × (OC/OA). Distributions for each factor were constrained: ON/OC (B) from 609 samples (lognormal; mean 0.11, SD 0.06, median 0.10); OA/OC (reciprocal of C) constrained to 1.3–2.5 (assumed lognormal); OAN/ON (A) built bottom-up from literature-based OAN component groups and their molecular weight per N (MW per N atom) and abundance ranges, simulated via Monte Carlo. A had mean 7.16 with 95% CI [6.45, 7.78]. Monte Carlo simulations assumed lognormal distributions and independence among variables; sensitivity analyses explored alternative distributions (normal, triangular), different confidence bounds (90–99%), perturbations in MW per N for components (±10–30%, including synchronous changes for WION components), and expanded component contribution ranges. - Alternative dependence-aware algorithm: An alternative formulation removing potential dependence between A and C was also simulated (Supplementary Note 3) for comparison. - Source apportionment: Positive matrix factorization (EPA PMF v5.0) was applied to year-long datasets at TW, NS, and BJ with comprehensive tracers to resolve primary and secondary sources of ON, and to derive ON/OC ratios for source profiles. Additional analyses using HK urban sites (YL, MK) with cooking tracers supported identification of cooking contributions where applicable.

- Abundance of ON: Annual average PM2.5-bound ON ranged 0.4–1.4 µg N m−3 across 12 sites; ON comprised 17–31% of total aerosol nitrogen (TN). IN and ON were higher in northern than southern China (p < 0.0001), while ON/TN ratios were comparable (p = 0.08). In PRD, ON was slightly higher in urban vs suburban/rural sites (p < 0.001), IN comparable (p = 0.62), yielding higher median ON/TN in urban (22%) vs suburban/rural (17%) (p < 0.001). - OAN share of OA: Monte Carlo simulations yielded mean OAN/OA = 42.2%, median 38.6%, with 95% CI [14.1%, 87.3%]. Extensive sensitivity tests (distribution types, confidence ranges, MW per N perturbations, component contributions, WION uncertainties) resulted in means roughly 36.5–45.9% and similar 95% CIs, indicating robustness. An alternative dependence-aware algorithm gave mean 39.8%, median 37.2%, 95% CI [13.8%, 78.9%]. Authors conclude OAN typically contributes 37–50% of OA, with overall 95% CI approximately [12%, 94%]. - Molecular implications: Mean MW per N atom in OAN is ~100 Da (A ≈ 7.16), unlikely to exceed ~109 Da per N (upper 95% bound). Assigning unrealistic high MW per N (e.g., 200 Da) leads to implausible OAN/OA > 100% in simulations, implying many OAN molecules are low molecular weight or contain multiple N atoms. - Source apportionment of ON: Approximately 70% of ON arose from primary emissions overall. Biomass burning contributed 21–24% (0.14–0.29 µg N m−3); primary biological aerosol particles (PBAP) contributed 7.4% at TW (urban), 16.3% at NS (rural), and 17.6% at BJ (suburban). Vehicle emissions contributed 13.6% at TW but were minor at NS (3.4%) and BJ (4.6%). Cooking was significant at BJ (21.5%); at HK urban sites with tracers, cooking contributed 3.5% (YL) and 9.0% (MK). Other primary sources (industrial, ship, coal combustion, soil dust, sea salt) typically contributed <10% each annually. - Secondary formation: Secondary processes accounted for ~30% of ON at each of TW, NS, and BJ. Nitrate-rich factors contributed more at BJ (0.103 µg N m−3; 14.3%) than TW (5.8%) and NS (4.3%), consistent with higher NOx. Sulfate-rich factors and two SOA factors (at BJ) also contributed; mechanisms include organic nitrate formation (RO2 + NO; NO3 + alkenes), aqueous-phase reactions (glyoxal + nitric acid; carbonyls with ammonium/amines/amino acids), and formation of N-containing organic salts via NH3–organic acid interactions. - Seasonality: Biomass burning contributions peaked in winter (27–38%), were second highest in fall (18–23%), and lowest in summer (4–7%). PBAP contributions peaked in summer (14–32%). Sulfate-rich secondary ON was important year-round at TW and NS and rose in spring/summer at BJ (20.8% in summer vs 6.5% annual). - Source-specific ON/OC ratios (atomic): Biomass burning 0.077–0.083; PBAP 0.094–0.141; industrial/coal 0.046–0.067; vehicle 0.048–0.076; ship 0.036–0.052; soil dust 0.129–0.162; cooking varied (0.291 at BJ vs 0.052 in HK); secondary nitrate-linked 0.035–0.241; secondary sulfate-linked 0.097–0.136; BJ SOA factors 0.103 and 0.082. These values refine assumptions used in prior modeling studies.

The findings quantitatively establish that nitrogenous organic molecules constitute a large fraction of ambient OA, addressing a critical knowledge gap left by carbon-centric analyses. The robust estimate that OAN typically contributes 37–50% of OA implies that nitrogen is widely incorporated into particle-phase organic compounds. Source apportionment clarifies that both primary (notably biomass burning and PBAP) and secondary processes (nitrate- and sulfate-linked pathways and SOA formation) are major ON sources, with distinct seasonal patterns and urban–rural contrasts. These results have significant implications for atmospheric chemistry and climate modeling: they call for explicit representation of nitrogen-bearing organics in OA frameworks, improved parameterizations of ON/OC ratios by source, and refined assessments of reactive nitrogen deposition via aerosols. The molecular-weight constraints suggest analytical strategies focusing on lower-mass or multi-nitrogen compounds. Overall, the joint ON–OC measurements and modeling provide a quantitative basis to reassess OA composition and its environmental impacts.

This study provides the first extensive, quantitative survey of bulk organic nitrogen in PM2.5 across diverse Chinese environments using a sensitive, simultaneous IN–ON analyzer. It demonstrates that nitrogenous organic aerosols comprise a substantial portion of OA (mean ~42%, typically 37–50%), with ON representing 17–31% of aerosol total nitrogen. Source apportionment identifies biomass burning, secondary formation, and primary biological emissions as dominant ON sources, with clear seasonal and spatial patterns. The work refines source-specific ON/OC ratios for use in models and constrains the molecular characteristics of OAN. Future research should integrate bulk ON measurements with detailed molecular speciation, develop online ON measurement techniques, expand tracer coverage (e.g., cooking, SOA) at more sites, and better characterize secondary ON formation pathways and ON/OC variability across environments and seasons.

- Sampling and measurement: Potential loss of semi-volatile organic compounds during filter sampling may bias OC and ON and thus OAN/OA estimates. The analysis focuses on PM2.5; coarse-mode contributions (notably PBAP) are underrepresented. - Source apportionment: Lack of cooking and SOA tracer data at some sites (TW, NS) limited resolution of those sources in PMF, introducing uncertainty. PMF-derived source profiles and ON/OC ratios may carry biases inherent to receptor modeling and tracer selection. - Modeling assumptions: The Monte Carlo framework assumes distributions and partial independence among variables; although sensitivity tests and an alternative dependence-aware scheme were performed, uncertainties remain, particularly in the molecular-weight-per-N estimates for WION components and component contribution ranges. - Generalizability: Results are from Chinese urban to rural environments and may not directly extrapolate to other regions without further measurements.