PM_2.5 sample collection was carried out at Dongying Atmospheric Super Station (118.59° E, 37.45° N) from 27 October to 6 December 2021. The surrounding area is primarily residential and commercial, with well-developed transportation infrastructure and no significant sources of industrial pollution, making it a representative urban monitoring site (Fig. 1). A high-flow particulate sampler (TH-1000H, Wuhan Tianhong Instrument Co., LTD) was employed to obtain the samples. This instrument had a sampling flow rate equal to 1.05 m³ min⁻¹ and a flow accuracy of ±2.5 %. It comprised a quartz microfiber filter (Quartz, 203 mm × 254 mm, Whatman, UK) with an effective rectangular dimension of 180 mm × 230 mm. Samples were collected twice daily: once during the day from 08:00 am to 07:30 pm local time (LT) and once from 08:00 pm to 07:30 am LT. Meteorological statistics, including temperature, weather conditions, humidity, wind direction and speed, were measured concurrently with each sampling.

Organic species in PM_2.5 were detected following a pre-treatment process that included ultrasonic extraction along with derivatization. Initially, the sampled filter was divided into four equal surface area segments, and one quarter was utilized for experimental analysis. The quarter of the sampled filter was placed into a sample bottle and then immersed in a mixture of dichloromethane and methanol solution (2:1, v/v) to completely submerge the filters. After that, ultrasonic extraction was performed three times, with each session lasting 15 minutes. Following the process of extraction, filtration was carried out through glass wool using a Babbitt dropper into pear-shaped flasks. The filtrates were then concentrated to a small amount through a rotary evaporator inside a vacuum. Subsequently, they were transferred to GC bottles, which were dried with a nitrogen purifier, followed by the addition of 60 µL of N, O-bis-(trimethylsilyl) trifluoroacetamide (BSTFA+TMCS, 99:1, v/v) and pyridine (99.5 %) (5:1, v/v). The bottles were then placed in an oven at 70 °C for a total time duration of 3 h to complete the derivatization process. After the reaction, the solution was cooled, and an internal standard comprised of 40 µL tridecane was added. The solution was mixed completely and then placed in the refrigerator for examination. The derivatized samples were examined using gas chromatography coupled with a mass spectroscopy detector (GC/MS: HP 7890A, HP 5975C, Agilent Co., USA). The extraction and derivatization methods described above allowed for the simultaneous measurement of the samples' polar and non-polar constituents. This study analyzed 13 different PAHs, 8 different OPAHs, and 9 different NPs, as well as some organic tracers, e.g. levoglucosan. Specific species of PAHs, OPAHs, and NPs were listed in Table S1 in the Supplement.

Inorganic ions soluble in water were quantified through an ion chromatograph (Dionex-1100). The analysis focused on eight water-soluble inorganic ions: NO3-, Cl⁻, NH4+, SO4-2, Na⁺, Mg²⁺, K⁺, and Ca²⁺. The measurement of organic carbon (OC) and elemental carbon (EC) was performed through a thermal/optical carbon analyzer (DRI2015). Detailed tests for analyzing EC, OC, and inorganic ions are provided in a previous report (Ren et al., 2021).

Quantum chemical calculations both in the gas phase and liquid phase were performed using the Gaussian 09 software. The geometries and frequencies for the reactant monomers, reactant complexes (RC), transition states (TS), intermediate (IM), and product complexes (PC) were performed via density functional theory (DFT) calculations at the M06-2X/6-311++G(2df,2p) level of theory, which is well-established for studies of organic systems (Li and Wang, 2014; Zhao and Truhlar, 2007; Wang et al., 2025). The conductor-like polarizable continuum model (CPCM) was used for the optimization calculations to include the solvent effect in the liquid (Takano and Houk, 2005). Radicals were treated using unrestricted Hartree–Fock calculations and TS was optimized to ensure the presence of a single imaginary frequency. Intrinsic reaction coordinate (IRC) calculations were carried out to further guarantee that the TS connects the right pre- and post-complexes along the reaction coordinate (Gonzalez and Schlegel, 1989).

Quality control and QA procedures for sampling and laboratory analysis strictly follow the environmental monitoring technical specifications and analytical standards established by the Ministry of Environmental Protection. Periodic inspections of all analytical and testing instruments were carried out in accordance with the relevant metrological verification protocols. The quartz microfiber filter was burned inside a Muffle furnace at a temperature of 450 °C for 6 h before sampling to remove any organic contaminants. After natural cooling, it was removed and transferred to a controlled temperature and humidity box (25±1 °C, 50±5 % RH). The sampled filter was packed in a ziplock bag and stored in the refrigerator at a temperature below 0 °C until analysis. To minimize systematic errors, corresponding blank filter samples were simultaneously prepared and stored followed by analysis under the same conditions as the collected samples. Limits of detection (LOD) of the target compounds were calculated using signal-to-noise ratios of 3:1 following previous studies (Bandowe et al., 2014b; Li et al., 2016). In this work LOD are in the ranges of 0.003–0.008 ng µL⁻¹ for PAHs, 0.007–0.37 ng µL⁻¹ for OPAHs, and 0.0001–0.0014 ng µL⁻¹ for NPs (Table S1). The blank sample analysis showed no serious contamination (less than 5 % of real samples) and the data reported here were all corrected according to the blanks.

In northern cities of China, centralized heating generally starts on 15 November each year and continues until 15 March of the following year. Therefore, the entire period in this study was divided into two stages: before heating (27 October to 14 November 2021) and during heating (15 November to 6 December 2021). Figure 2 and Table 1 illustrate the temporal variations within the meteorological factors, concentrations of gaseous pollutants, as well as the primary components of PM_2.5. Before heating, the relative humidity (RH) and temperature (T) were higher, averaging 64 % and 12 °C, respectively. During the heating period, the average values were found to be lower at 60 % and 7.9 °C. Overall, the average values for the entire study period were determined to be 62 % and 9.6 °C. The average concentrations of SO₂ and NO during heating (11 and 17 ppb, respectively) were nearly twice that of the value observed before heating, with a slight increase in the level of NO₂. These increases are primarily associated with enhanced combustion activities during the heating season, which supported by the correlation analysis (Sect. 3.3) though synergistic effects (e.g., meteorological effect) require further quantification in future studies. In contrast, the levels of O₃ and O_X (the sum of NO₂ and O3) were significantly higher before heating, with an average value of 49 and 80 ppb, respectively, in comparison to that observed during the heating period. This variation was associated with the higher temperatures and light conditions before heating, which are conductive to photochemical reactions.

Figures 2 and 3 showed the time variations and absolute proportion of chemical components in PM_2.5 before and during heating, respectively, throughout the entire sampling period. Throughout the observation period, secondary inorganic aerosols (SIA) such as SO42-, NO3-, and NH4+ were determined as the main chemical components of PM_2.5, followed by OM (1.6 times the OC) (Turpin and Lim, 2001). Before heating, the concentrations of SO42-, NO3-, and NH4+ were found to be 5.5, 17, and 7.5 µg m⁻³, respectively, while during heating, they demonstrated the average values of 5.3, 17, and 7.7 µg m⁻³ (Table 1). The total concentrations and proportions of SIA remained relatively stable, with an average of 29.9 µg m⁻³ and 35 % during the whole process, respectively. Among these, NO3- was determined as the predominant species, constituting approximately 19 % of the total PM_2.5 during the sampling period. OM comprised 12.7 % of PM_2.5 before the heating period and increased significantly to 17.5 % during heating (Fig. 3b, c). Correspondingly, the average OC concentration increased from 5.4 to 7.7 µg m⁻³. However, the average ratio of OC/EC remained relatively unchanged, with average values of 2.5 before heating and 2.4 during heating. Generally, OC is generated from both primary emissions and secondary formation, while EC is predominantly derived from primary sources (Zhang et al., 2015). The significant increase in the concentration of OC alongside a stable OC/EC ratio suggests that the sources of carbonaceous aerosols in this study include both primary emissions and secondary formation processes (Zhang et al., 2015). A strong positive correlation between levoglucosan and Cl⁻ (r=0.58–0.69) and a weaker correlation with K⁺ (r=0.3–0.34) were observed (Fig. 4), with both ions showing elevated concentrations during the heating period, suggests additional combustion activities, e.g. biomass burning. It is worth noting that Cl⁻ is more robustly associated with levoglucosan, a well-established tracer for biomass burning. Thus Cl⁻ was prioritized as the inorganic tracer for biomass burning in subsequent analyses (Sect. 3.3).

Based on the data presented in Table 2, the total concentrations of NP compounds (∑9NPs) demonstrated a mean value equal to 72 ng m⁻³ across the sampling period. During the heating phase, the average concentration of ∑9NPs was 76 ng m⁻³ (range: 23–175 ng m⁻³), representing a 1.2-fold increase compared with before heating (average: 65 ng m⁻³, range: 4.2–149 ng m⁻³), although this difference was not statistically significant (p>0.05). These concentrations were approximately twice those observed during autumn and winter (18 October 2017–17 January 2018) (38 ng m⁻³) in earlier work conducted in Beijing (Ren et al., 2024). They were also considerably higher in comparison to the values recorded in the summer (1–30 July 2017) (8.5 ng m⁻³) and spring (10–21 April 2017) (8.6 ng m⁻³) (Ren et al., 2022). Different from OPAHs and PAHs, the total concentrations of ∑9NPs did not demonstrate significant diurnal variation before heating, but a significant nighttime increase was observed during the heating period, with approximately 1.4 times higher concentrations at night compared to the levels of daytime (Figs. 5d and 6c). However, the rank order of relative abundance for the nine NP species in PM_2.5 remained consistent throughout the observation period. Among all the species, 4M5NC demonstrated the highest concentration, contributing to 73 % of the total NPs before heating and 53 % during heating, followed by 4NP, comprising 20 % before heating and increased to 34 % during heating (the pie charts in Fig. 5d). The total concentration of ∑9NPs observed in the current research was generally comparable to previous measurements carried out in Beijing during winter (average: 74 ng m⁻³) (Li et al., 2020). Conversely, the levels were substantially higher compared to those recorded in Jinan (average: 48 ng m⁻³) (Wang et al., 2018), Hong Kong SAR (average: 12 ng m⁻³), and Xi'an (average: 17 ng m⁻³) (Wu et al., 2020; Chow et al., 2015). Compared to the previous measurements in other regions, the observed ∑9NPs concentrations were generally higher, such as those reported in Germany (16 ng m⁻³), the UK (19 ng m⁻³), and Belgium (32 ng m⁻³ in winter, 13 ng m⁻³ in autumn) (Teich et al., 2017; Mohr et al., 2013; Kahnt et al., 2013). The spatial differences in the above concentration levels were influenced by a combination of various factors. Besides meteorological conditions, the influence of different emission sources was very important.

Aromatic hydrocarbons are a major class of VOCs that play a vital role in secondary organic aerosols (SOA) formation, particularly in urban areas (Song et al., 2021). The present study reported a positive correlation of these aromatic compounds with the key precursors (i.e. benzene, toluene, xylene) and a negative correlation with NO₂, NO3- and O₃, which are key parameters during the reaction process (Fig. 7a–c). Therefore, these substances were primarily derived from the same source as the precursors, rather than their main products. Furthermore, considering the substantial effect of biomass combustion, these compounds are probably the secondary products formed through the oxidation process of gaseous pollutants released during the biomass burning process. Nitroaromatic hydrocarbons are critical species influenced by biomass combustion emissions where phenol, nitrobenzene, and P-Cresol are their significant precursors (Wang et al., 2020). Therefore, to better comprehend the formation mechanism of the dominant NPs species in this observation, DFT calculations at the M06-2X/6-311++G(2df,2p) level of theory was carried out to explore the formation mechanism of 4NP and 4M5NC through oxidation processes of the major precursors by OH, NO₂, and NO₃ (Huang et al., 2010; Roman et al., 2022; Shenghur et al., 2014; Wang et al., 2019b). Electronic energies (EDFT), Gibbs free energies (GDFT) of the reactant monomers, reactant complexes (RC), transition states (TS), intermediates (IM), and product complexes (PC), with the imaginary frequency of TS at 298.15 K and 1 atm, are collected in Tables S4–S7 of the Supplement. The overall formation mechanisms, detailed in Fig. 8a, include both H-abstraction and OH/NO₂ addition steps where Mechanisms 1 and 2 show the two-step formation mechanism for 4NP, while Mechanism 3 describes the four-step pathway for 4M5NC.

In the H-abstraction step, either an OH radical, NO₂ or NO₃ abstracts a hydrogen atom from the aromatic ring, forming an intermediate free radical (IM) and resulting in the formation of H₂O, HONO or HNO₃. In the OH/NO₂ addition step, the IMs react with OH or NO₂ without any reaction barriers. In Table S4, the calculated binding energy (E), Gibbs free energy change (ΔG) for the reactant monomers, Gibbs free energy (G), transition states (TS), reactant complexes (RC), intermediate (IM), and PCs are presented. While Fig. 8b exhibited the calculated dominant formation mechanism of 4NP and 4M5NC along with the reaction barrier in kcal mol⁻¹.

For the formation of 4NP, two precursors, phenol, and nitrobenzene, were selected which contain three distinct types of hydrogen atoms in their aromatic rings (ortho-, meta-, and para-H). In the case where the -OH and -NO₂ functional groups in 4NP are at the para-position, only the para-H in phenol and nitrobenzene was chosen for the formation of 4NP. During the daytime, when the gas-phase concentration of OH radicals is significant, phenol initially reacts with OH to yield a reactant complex RC_ph-OH with a ΔG of 3.16 kcal mol⁻¹. Then RC_ph must overcome a reaction barrier of 8.64 kcal mol⁻¹ to transition to the TS_ph-OH state, resulting in the generation of IM1, a free radical intermediate with a ΔG of -1.03 kcal mol⁻¹. The overall H-abstraction step was slightly exothermic, with a reaction barrier of 11.80 kcal mol⁻¹. The generated IM1 rapidly proceeds through an addition reaction with NO₂ without any energy barrier, forming 4NP with a ΔG of -64.50 kcal mol⁻¹.

During nighttime, a parallel H-abstraction step occurred where NO₂ or NO₃, instead of OH, abstracted a hydrogen atom from phenol. However, this step possess higher overall reaction barriers than that of OH, reaching 40.31 and 15.79 kcal mol⁻¹ in the gas phase, respectively. Moreover, the reactions were highly endothermic with the ΔG of 36.51 and 7.42 kcal mol⁻¹ in the gas phase, respectively, indicating that the H-abstraction step initiated by NO₂ or NO₃ were not feasible. Therefore, the calculations for Mechanisms 2 and 3 focused solely on the H-abstraction step initiated by OH.

For an alternative formation pathway of 4NP from nitrobenzene during the daytime, the overall H-abstraction step was slightly exothermic (-1.08 kcal mol⁻¹) with a reaction barrier of 12.66 kcal mol⁻¹. The generated IM2 then proceeded through a highly exothermic and barrier-free OH addition step to form 4NP. In a comparative analysis, it was possible to initiate 4NP formation with both phenol and nitrobenzene. For mechanism 1, phenol was preferred to undergo the H-abstraction reaction during the daytime and the NO₂ addition reaction at nighttime. For Mechanism 2, nitrobenzene preferred to undergo both reactions involving H-abstraction and OH addition during the daytime.

The formation of 4M5NC followed a four-step mechanism with P-Cresol as the precursor. During the daytime, P-Cresol initially reacted with OH to form free radical intermediate IM3, overcoming a reaction barrier of 7.19 kcal mol⁻¹. After that, IM3 underwent a highly exothermic OH addition step without any energy barrier to form IM4 which was then proceeded through a reaction barrier of 11.83 kcal mol⁻¹ to produce IM5. At nighttime, the IM5 formed during the daytime quickly reacted through a highly exothermic NO₂ addition step with no significant energy barrier, resulting in the formation of 4M5NC, with a ΔG of -61.14 kcal mol⁻¹ in the gas phase and -63.47 kcal mol⁻¹ in the liquid phase.

A comprehensive reaction mechanism (Fig. S2) of P-Cresol with OH, NO₂, NO₃ has been drawn to display the potentially different pathways. As can be seen, there are two parallel pathways leading to the formation of 4M5NC. One involves the initial addition of an OH radical on C2 of the benzene ring, followed by the addition of NO₂ on C5 mentioned in Mechanism 3. While the other one involves the initial addition of a NO₂ on C5 of the benzene ring, followed by the addition of OH on C2. Based on the calculated reaction barrier and the reactant monomer concentrations, the OH radical induced H-abstraction step possesses a much lower barrier, resulting in the following OH addition step more feasible. Hence, Mechanism 3 may be the dominant formation pathway of 4M5NC.