Investigating Dual Character of Atmospheric Ammonia on Particulate NH₄NO₃: Reducing Evaporation Versus Promoting Formation

1. Introduction

Ammonium nitrate (NH₄NO₃), a significant water-soluble inorganic component in atmospheric particles, plays a critical role in influencing regional visibility, cloud condensation nuclei (CCN) activity, and radiative forcing through its spatial distribution and gas-particle partitioning behavior [1,2,3,4,5,6,7,8,9,10,11,12]. The thermodynamic properties of NH₄NO₃ make its concentration highly sensitive to changes in ambient temperature (T) and relative humidity (RH) [2,13,14,15,16,17,18,19,20]. For example, elevated RH enhances the partitioning of both HNO_3gas and NH_3gas into the submicron particle phase, as well as the transfer of HNO_3gas to supermicron particles through the neutralization of negative charges by non-volatile metal ions, thereby further regulating the formation and stability of NH₄NO₃ and nitrate metal salts [2,14,15,17,18,19]. Conversely, elevated temperatures and low humidity conditions tend to drive the dissociation of NH₄NO₃ back into the gas phase [17,18,19]. These coupled thermodynamic and chemical interactions underscore the importance of elucidating the formation and depletion mechanisms of NH₄NO₃ in the atmosphere [9,11,21,22,23].

In the formation of NH₄NO₃ through the transfer of HNO_3gas to submicron particles, NH_3gas, the primary component of atmospheric alkaline gases, serves as both a key precursor and a stabilizing agent [24,25,26,27,28,29,30]. Additionally, N₂O_5gas contributes to nighttime nitrate formation through particle-phase condensation processes, during which NH_3gas further enhances the stability of NH₄NO₃ aerosols, particularly under conditions of low alkali metal ion concentrations (such as Na⁺ and K⁺) in the submicron particles [31,32,33,34,35,36]. Globally, the primary source of NH₃ is agricultural activities, although transportation and industrial emissions in urban areas also contribute to NH₃ emissions [37,38,39,40,41,42,43]. Thus, the regulatory effect of NH₃ on NH₄NO₃ demonstrates significant spatial heterogeneity and concentration dependence [12,24,44,45,46,47,48]. Observational evidence suggests that when the NH₃ concentration exceeds the stoichiometric threshold required for the complete neutralization of sulfate, its role shifts from “precursor-limited” to “thermodynamic-limited,” fundamentally altering the pathways of NH₄NO₃ formation and its phase stability [10,48,49,50,51,52,53].

In contrast to previous studies that primarily focused on the formation pathways, environmental impacts, climatic effects, and policy implications of NH₄NO₃, Shen et al. [23] provided comprehensive evidence showing that high concentrations of NH₄NO₃ during severe PM_2.5 (atmospheric particles with diameter less than 2.5 μm) air pollution events are primarily due to prolonged, rapid conversion within combustion plumes under cold, humid and stable meteorological conditions, followed by subsequent evaporation along the plume dispersion track in ambient air. Using isotopic traces, Xiao et al. [54] found that nitrate in PM_2.5 primarily originated from primary combustion in a central Chinese city. Furthermore, two independent thermodynamic modeling studies in Beijing showed that thermodynamic force did not support the formation of NH₄NO₃, even under NH_3gas-rich atmospheres [49,55]. More recently, Sun et al. [8] and Yang et al. [44] identified high concentrations of NH₄NO₃ at approximately 400 meters in altitude in northern and eastern China, consistent with emission heights and plume rise from strong industrial sources. If NH₄NO₃ aerosols in ambient air primarily originated from primary emissions and prolonged combustion plumes, NH₄NO₃ evaporation would generally be expected to occur [56,57]. This raises the question: does atmospheric NH₃ play a role in reducing NH₄NO₃ evaporation or in promoting NH₄NO₃ formation in ambient air?

This paper aims to comprehensively investigate the dual regulatory roles of ammonia in both the formation and evaporation of NH₄NO₃, using high-resolution data collected at a clean coastal site on a tropical island in China [58]. The simplicity and clarity of pollution sources at this location offer an ideal environment for isolating the effects of NH₃, thereby enhancing the robustness of the analysis. By combining field observational data analysis, thermodynamic equilibrium modeling, and source apportionment analysis, this work provides novel scientific insights into the synergistic regulation of nitrate aerosols in PM_2.5. The findings are expected to offer theoretical foundations for improved air pollution mitigation strategies and for advancing our understanding of aerosol-climate interactions.

2. Materials and Methods

The sampling site for this study is situated on the third floor of a research building within a high-tech industrial park in southeastern Hainan Province (18.328°N, 109.169°E, Figure 1a-b). This region represents a typical land-sea transition zone, positioned approximately 1.9 km west of the South China Sea coastline [58]. It is seasonally influenced by alternating southeasterly and northeasterly monsoons under the overarching control of the East Asian monsoon system. To the south of the site, a 487-meter-high hill situated 1.3 km away acts as a local topographic barrier, fostering local circulation patterns under weak synoptic forcing. Between the sampling site and the hill lies the Huaneng Nanshan Power Plant-a natural gas-fired facility located 1.2 km to the south at the hill’s base (Figure 1cd). The plant has a total installed capacity of 2 × 46 MW (new generators visible on the left side of Figure 1c) and 132 MW (original units largely obscured on the right side of Figure 1c), and serves as a major source of industrial air pollutant emissions in the area. Plumes from the facility are occasionally observed traversing the sampling location. Additionally, within a 5 km radius of the park, the surrounding landscape comprises undeveloped wastelands and agricultural land, with approximately 2.3 km² designated for future development. A major traffic road is located 50 meters east of the sampling point, with traffic volumes reaching 810 vehicles per hour during peak periods (07:00-09:00 and 17:00-19:00), 24% of which are electric vehicles (zero-emission models), classifying it as a low mobile pollution source area.

The study utilized an Aerosol Ion Monitor-Ion Chromatography system (AIM-IC) (Thermo Fisher, Waltham, MA, USA) for the continuous monitoring of both gaseous and particulate matter in the atmosphere. The detection limits for NO₃^- and NH₄⁺ were 5 × 10^-2 μg m^-3 and 4 × 10^-4 μg m^-3, respectively. The AIM-IC system was installed within the laboratory and connected to ambient air via a 2.5-meter-long stainless steel inlet tube (3.5 cm inner diameter), with the sampling probe positioned approximately 10 meters above ground level. Regular maintenance and calibration procedures were carried out throughout the observation period to ensure the accuracy and instrument stability. Data collected during two periods-November to December 2023 and June 2024 were used for analysis, capturing atmospheric composition under varying meteorological conditions. Detailed operational information for the AIM-IC system is available in our previous publications [18,58,59]. The AIM-IC system is equipped with two analytical columns, one for cations (Ion Pac CS20, 2 × 250 mm) and one for anions (Ion Pac AS 11-HC, 2 × 250 mm), along with two guard columns, one for cations (CG20, 2 × 50 mm) and one for anions (AG 11-HC, 2 × 50 mm). Additionally, the system is equipped with a PM_2.5 cyclone separator to ensure accurate sampling of fine particulates, operating at a flow rate of 3 L min^-1. The system provides hourly concentration data for reactive gases (NH₃, SO₂, HNO₃, etc.) and water-soluble ions (NH₄⁺, NO₃^-, etc.). It is important to note that the HNO₃ signal measured by AIM-IC includes not only gaseous HNO₃ but also contributions from its atmospheric precursors, such as N₂O₅, a fraction of organic nitrate, and trace amounts of NO₂ in ambient air, similar to other acidic vapor species measured by the system [59]. Therefore, HNO₃* was used instead of HNO₃ in the subsequent analysis.

Particle size spectrum data were measured simultaneously using a Scanning Mobility Particle Sizer (SMPS, Grimm, Germany). During sampling, the air was dried before entering a differential analyzer, where the number concentration size distribution from 11 to 1110 nm was obtained at a sampling flow rate of 0.3 L min^-1, with a 4-minute resolution and 112 sampling channels, providing high-precision particle size distribution data. Additionally, an OC-EC online analyzer (Model 4, Sunset, USA) was employed for the continuous monitoring of organic carbon (OC) and elemental carbon (EC) in PM_2.5. Owing to the relatively clean atmospheric conditions at the sampling site (annual mean PM_2.5 < 10 μg m^-3), OC and EC concentrations were monitored with a time resolution of 2 hours to improve signal-to-noise reliability. However, the OC-EC analyzer experienced frequent malfunctions. Data points identified as outliers due to instrument malfunction or external contamination events were excluded during quality control to ensure the accuracy and integrity of the dataset.

For model applications, this study utilized the Extended AIM Aerosol Thermodynamics Model (E-AIM, versions Model II and Model IV, available at: http://www.aim.env.uea.ac.uk/aim/aim.php) to simulate the gas-particle partitioning equilibrium of NH₄NO₃. E-AIM remains the only publicly available thermodynamic model capable of simultaneously resolving multiphase equilibrium processes involving both inorganic and organic aerosols. The core assumptions of E-AIM are as follows: (1) homogeneous internal mixing of aerosol components, and (2) thermodynamic equilibrium between aerosols and ambient gases [58,59]. In this study, both Model II and Model IV were used to calculate equilibrium partitioning among gaseous, aqueous, and solid phases. Model II is suitable across a broad range of relative humidity (0.1 < RH < 1.0) and accounts for the multiphase behavior of the NH₃-HNO₃-H₂SO₄-NH₄⁺-NO₃^--SO₄^2- (including solid salt precipitation and liquid water dissolution), resolving the distribution of species such as (NH₄)₂SO₄, NH₄NO₃, etc. Model IV extends the capabilities of Model II by incorporating Cl^- and Na⁺ to represent the influence of sea salt aerosols. Meteorological data were obtained from the China Meteorological Data Service Center (http://data.cma.cn/), and hourly PM_2.5 mass concentration data were retrieved from three national monitoring sites in Sanya. Using the NOAA HYSPLIT model (https://www.ready.noaa.gov/HYSPLIT.php), 24-hour backward trajectories were calculated at a height of 500 m, reflecting air mass transport characteristics at the top of the boundary layer during the day or above it at night, with a time resolution of 1 hour.

To apportion the source contributions of NO₃^- in PM_2.5 ([NO₃^-]_PM2.5), the study integrated the modified particle number concentration spectrum (11-700 nm, with 39 size segments reduced from the original ones) with the [NO₃^-]_PM2.5 composition matrix using the Positive Matrix Factorization (PMF) model for positive matrix decomposition. This method clarified the contributions of various source aerosols to the formation and evolution of NO₃^-. Strong new particle formation (NPF) events were excluded during the PMF analysis to minimize their dominant effect on the dataset. The study used the Q-value (Q/Qexp) as a fitting evaluation index, comparing changes in Q-values across different factor solutions and considering residual distribution and the physical significance of factor spectra to determine the optimal solution[60]. After running the model 20 times with data from 2023 and 2024, the five best factors were identified, with a Q/Q_exp ratio of 1.12.

3. Results

3.1. Overview of Observational Results at the Coastal Site

Figure 2 illustrates time series of concentrations of major ions (including NO₃⁻, NH₄⁺, SO₄²⁻, and Na⁺) and carbonaceous components (OC and EC) in PM_2.5, along with hourly measurements of NH₃ and HNO₃* gases during the observation periods (14 November to 12 December 2023 and 7-20 June 2024). During the continuous observation in November 2023 (excluding 27-30 November), the [NO₃^-]_PM2.5 remained relatively low, fluctuating around 0.6 ± 0.3 μg m^-3, with a maximum value of only 1.7 μg m^-3. However, between 27 and 30 November, several peaks in [NO₃^-]_PM2.5 were observed, with the highest concentration reaching 3.9 μg m^-3. A weak correlation was observed between [NO₃^-]_PM2.5 and the corresponding SO₄^2- and NH₄⁺ concentrations (R² = 0.42 and 0.52, respectively; P < 0.01), whereas, SO₄^2- and NH₄⁺ showed a strongly correlated (R² = 0.91; P < 0.01). These results suggested that NO₃^- in PM_2.5 partly existed as NH₄NO₃, while SO₄^2- was primarily existed as NH₄⁺ salts. Furthermore, no correlation was found between [NO₃^-]_PM2.5 and Na⁺ concentrations, indicating that aged sea salts, such as NaNO₃, were not major contributors to the observed NO₃^- in PM_2.5. Additionally, NO₃^- in PM_2.5 and other ions exhibited significant diurnal variations, although the characteristics of these variations were inconsistent.

In contrast, the diurnal variations of the concurrently observed HNO₃*_gas were consistent clear, with peaks consistently occurring around noon and before sunset, and low values observed before sunrise (Figure 2c). Similarly, NH_3gas showed a generally consistent diurnal pattern, characterized by peak concentrations shortly after sunrise and around midnight, and minimum concentrations occurring before and after sunset. A comparison of the diurnal variations of ions in PM_2.5 and the concurrently observed gases suggested that diurnal meteorological factors (such as T, RH, wind speed, wind direction, and sunlight) may significantly influence the concentration variations of each component. The sources of ions in PM_2.5 appear to be more variable than HNO₃*_gas and NH_3gas.

Furthermore, during the one-month observation period, atmospheric NH_3gas concentrations fluctuated around 8.2 ± 3.3 μg m^-3, with a maximum of 23 μg m^-3 and a minimum of 2.5 μg m^-3. In contrast, HNO₃*_gas concentrations remained much lower, fluctuating around 0.1 ± 0.04 μg m^-3, with the maximum of 0.3 μg m^-3. Despite the fact that the formation of particulate [NO₃^-]_PM2.5 is governed by the product of NH₃ and HNO₃ concentrations[13], the markedly lower levels of HNO₃*_gas suggest that it was a more significant limiting factor for [NO₃^-]_PM2.5 formation in this environment. The carbonaceous components in PM_2.5, particularly OC, significantly exceeded the concentrations of inorganic ions, with concentration ranging from 4.7 to 12 μg m^-3 C (6.7 ± 1.3 μg m^-3 C). The corresponding EC ranged from 0.2 to 1.5 μg m^-3 C (0.4 ± 0.1 μg m^-3 C), which were slightly lower than [NO₃^-]_PM2.5. The high OC and OC/EC ratios might be related to nearby cooking emissions, as suggested by the concurrently low SO₄^2- and NO₃^- concentrations [61].

Similar to the data from November-December 2023, the [NO₃^-]_PM2.5 in June 2024 remained at a low concentration level of 0.2 ± 0.1 μg m^-3, except for June 16. During this period, the maximum concentration observed was 0.7 μg m^-3. However, on June 16, a distinct anomaly in [NO₃^-]_PM2.5 was observed, with a peak of 1.8 μg m^-3, significantly higher than the concentrations on the preceding and following days. During the two-week observation period in June, [NO₃^-]_PM2.5 exhibited noticeable diurnal variation, but the timing and width of the daily peaks were inconsistencies, suggesting the complexity of its influencing factors. A weak correlation was observed between [NO₃^-]_PM2.5 and the corresponding SO₄^2- and NH₄⁺ concentrations (R² = 0.46 and 0.51, respectively; P < 0.01), whereas SO₄^2- and NH₄⁺ were strongly correlated (R² = 0.88, P < 0.01). These results suggest that NO₃^- in PM_2.5 partly existed in the form of NH₄NO₃, while SO₄^2- predominantly existed as NH₄⁺ salts. A significant correlation was found between NO₃^- and Na⁺ in PM_2.5 (R² = 0.64, P < 0.01), indicating that aged sea salts, such as NaNO₃, may have contributed to the formation of NO₃^- in PM_2.5.

The diurnal variation of the concurrently observed NH_3gas concentrations also exhibited strong fluctuations, but the shape and timing of the peaks varied considerably. During the June observation period, the maximum and minimum NH₃ concentrations ranged were 13 μg m^-3 and 0.5 μg m^-3, respectively. Concentrations below 1 μg m^-3 predominantly occurred between midnight and sunrise. The diurnal variation of HNO₃*_gas generally showed an inverse pattern to that of NH_3gas, except on 16 June. HNO₃*_gas concentrations fluctuated around 0.06 ± 0.06 μg m^-3, with the maximum of 0.6 μg m^-3. However, nighttime concentrations remained below 0.1 μg m^-3. On 16 June, between 06:00 and 12:00, HNO₃*_gas concentrations exceeded 2 μg m^-3 during four separate intervals. The impact of high concentrations of HNO₃*_gas on [NO₃^-]_PM2.5 will be discussed later. Throughout the observation period, the carbonaceous components in PM_2.5 exhibited relatively low concentrations. Specifically, OC ranged from 0.5 to 2.7 μg m^-3 C (1.1 ± 0.4 μg m^-3 C), and EC ranged from 0.01 to 0.4 μg m^-3 C (0.1 ± 0.1 μg m^-3 C). These values were among the lowest compared with recent measurements in the U.S. [62]. This reduction in OC and EC concentrations can be attributed to the influence of the southwest monsoon, which brought air masses primarily from the ocean.

3.2. Thermodynamic Equilibrium Simulations for the Four Cases

To understand the causes for NO₃^- peaks in PM_2.5, thermodynamic equilibrium simulations were conducted to examine the gas-aerosol partitioning states. Figure 3 shows a comparison between the thermodynamic equilibrium simulations and observations for HNO_3gas, NH_3gas, and HCl_gas. During 27-29 November 2023, the simulated concentrations of HNO_3gas deviated from the 1:1 line in over 90% of the time, indicating that the NO₃^- in PM_2.5 has not reached equilibrium with HNO_3gas and was undergoing volatilization into the gas phase. The degree of deviation was independent on the corresponding [NO₃^-]_PM2.5, as indicated by the color scale in Figure 3. However, the simulation did not account for the presence of K⁺, which may bind with a fraction of NO₃^-. K⁺ had an appreciable contribution in PM_2.5 as analyzed in Section 3.3-3.6. This omission could partially explain the positive biased simulated HNO_3gas. Nevertheless, given the elevated levels of atmospheric NH₃ observed during this period, the potential impact of this error is expected to be minimal. Moreover, on 27 and 29 November, approximately 10% of the data points aligned with the 1:1 line, suggesting that NO₃^- in PM_2.5 and gaseous HNO₃ reached equilibrium during these periods. Conversely, no negative deviations from the 1:1 line were observed, providing strong evidence that particulate nitrate formation did not occur. Recall that the observed HNO₃* includes not only gaseous HNO₃ but also N₂O₅, organic nitrate and minor hydrolyzed NO₂. This may introduce a positive bias in the observed HNO_3gas, which could theoretically result in a negative deviation in the simulation. This effect is likely more pronounced at night, when N₂O₅ concentrations may exceed those of HNO₃. However, no such negative deviations were observed. Additionally, when NaCl is not considered in the model, a more pronounced positive bias in simulated HNO_3gas was observed on 28 and 29 November, especially those cases under conditions with higher [NO₃^-]_PM2.5 (Figure S1). This suggests that a portion of HNO_3gas may undergo heterogeneous reactions with NaCl to form particulate NaNO₃. However, the difference was negligible on 27 November.

On 16 June 2024, the modeled concentrations of HNO_3gas closely aligned with the 1:1 line for case with higher-[NO₃^-]_PM2.5, and were positively deviated from the 1:1 line to some extent for other cases. Again, when NaCl was excluded from the model, a more pronounced overestimation of HNO_3gas was observed including some higher-[NO₃^-]_PM2.5 cases. Thus, it can be inferred that 1) NO₃^- associated with NH₄⁺ in PM_2.5 was undergoing volatilization to HNO_3gas; 2) NO₃^- associated with Na+ was achieved the gas-aerosol equilibrium on that day.

The simulated equilibrium NH_3gas concentration slightly exceeded the observed value, primarily because the observed NH_3gas concentration is sufficiently high, meaning the gas-particle equilibrium has minimal impact. The simulated HCl_gas concentration shows a mix of both positive and negative deviations from the 1:1 line. These inconsistencies may stem from the presence of sea salt and ammonium salts in externally mixed forms, which are treated as internally mixed in the thermodynamic simulation. Fresh sea salt tends to produce a positive deviation, whereas aged sea salt may lead to a negative one. Additionally, the observed HCl*_gas includes contributions not only from HCl but also from Cl₂ and NO₂Cl [63], potentially introducing a positive measurement bias that could account for negative deviations in the simulated HCl concentrations.

From a theoretical perspective, the elevated atmospheric temperatures typical of tropical regions facilitate the attainment of gas-particle equilibrium for NH₄NO₃, with equilibrium timescales on the order of minutes [64,65]. The simulation results for 16 June also corroborated this behavior. However, in particles produced by fresh combustion, NH₄NO₃ may be encapsulated by organic material [66,67,68] , significantly prolonging the time required to reach equilibrium.

3.3. Case 1: Synergistic Effects of Primary Emissions and Meteorological Conditions on the Rapid In-Plume Formation and Subsequent Volatilization of Fresh Ammonium Nitrate

The case analysis is further conducted to characterize the NO₃^- peaks. Case 1 is based on observations obtained on 27 November 2023. Figure 4a-e displayed the concentrations of NO₃^-, NH₄⁺, and Na⁺ in PM_2.5, along with the concurrent atmospheric NH_3gas concentrations, particle number size distributions, and changes in PM_2.5 mass concentration at three national control sites in Sanya. They also included the correlation between the normalized [NO₃^-]_PM2.5 and the sum of particle number concentrations larger than 100 nm (N_>100) (as defined in the caption of Figure 4), as well as 24-hour air mass back trajectories for that day. Figure S2ab illustrated the correlations between NH₄⁺ and NO₃^-, and between NH₄⁺ and the equivalent concentrations of (NO₃^- + SO₄^2-) in PM_2.5. The results in Figure 4a revealed a bimodal distribution of [NO₃^-]_PM2.5, with a lower peak at 09:00 and a higher peak at 21:00. PM_2.5 mass concentrations at the three national control sites in Sanya also exhibited a similar bimodal distribution, except for a trough observed at the Junyue Sea-Beach station, which may be related to localized showers. This suggested that the increase in [NO₃^-]_PM2.5 may have been a widespread phenomenon. However, the calculated air mass back trajectories indicated that 1) the sampling site was neither downwind nor upwind of the national control sites, and 2) the trajectories nearly overlapped during both peak and the non-peak period (Figure 4e). Therefore, it can be inferred that the atypical diurnal variations observed on that day were more likely due to local atmospheric processes rather than regional pollutant transport.

The concentrations of NH₄⁺ and Na⁺ in PM_2.5 exhibited unidirectional increases and decreases in their diurnal variations, respectively, and did not show a statistically significant correlation with NO₃^- (Figure S2 and S3a). This indicates that sea salt and atmospheric NH_3gas were not the primary controlling factors for the increase in [NO₃^-]_PM2.5. However, atmospheric NH_3gas also displayed a bimodal distribution and a weak statistically significant correlation with NO₃^- (R² = 0.32, P < 0.01). A more detailed analysis revealed that changes in the concentration of NH_3gas concentrations lagged behind those of [NO₃^-]_PM2.5 (Figure 4ab). Additionally, the strong correlation between the equivalent concentration of NH₄⁺ and that of (NO₃^- + SO₄^2-) (Figure S2) suggested that NH₄⁺ in PM_2.5 predominantly existed as inorganic ammonium salts. Thus, the delayed peaks of atmospheric NH_3gas are more likely attributed to the volatilization of NH₄NO₃ [8,23,56], supporting by the thermodynamic equilibrium stimulation analysis presented in Section 3.2 . In this specific case, atmospheric NH_3gas likely played a role in restricting its volatilization as gaseous (HNO₃ + NH₃).

The particle number size distribution presented in Figure 4c indicated that, during periods of elevated [NO₃^-]_PM2.5, the median accumulation mode particle diameter was notably smaller than during periods with lower [NO₃^-]_PM2.5. Additionally, the normalized [NO₃^-]_PM2.5 exhibited a strong positive correlation with the normalized N_>100, with a regression slope of 0.7. This strongly suggests that NH₄NO₃ primarily originated from primary sources [23,54,63,64]. The slope of less than unity also confirmed that NH₄NO₃ was undergoing volatilization (Figure 4d). Furthermore, the high-resolution particle number size distributions further demonstrated that, during the NO₃^- peak periods, the number concentration spectra fluctuated dramatically (Figure 4c), a typical feature indicative of the influence of nearby pollution sources.

3.4. Case 2: Biomass-Burning-Derived Potassium Salts and Cloud-Processing of KNO₃ and NH₄NO₃ by Mixing with Additional Anthropocentric Sources, Followed by NH₄NO₃ Volatilization

The second case occurred on 28 November 2023. On this day, the diurnal variation of [NO₃^-]_PM2.5 exhibited a distinct pattern compared to that on 27 November (Figure 4 and Figure 5). Between 00:00 and 20:00, [NO₃^-]_PM2.5 remained relatively stable at 1.3 ± 0.2 µg m^-3 (Figure 4a), followed by a statistically significant increase to 1.5 - 3.0 µg m^-3 (P < 0.05). Notably, no corresponding increase in PM_2.5 mass concentration was recorded at the three national control sites (Figure 5b). This suggested that the increase in [NO₃^-]_PM2.5 was a localized phenomenon rather than a regionally distributed event. Simultaneous particle number size distributions revealed that, during the [NO₃^-]_PM2.5 increase, the median diameter of the accumulation mode particles decreased rather than increased (Figure 5c). Furthermore, a strong positive correlation was identified between the normalized [NO₃^-]_PM2.5 and the normalized N_>100 (Figure 5d), with the regression slope of 2.2, significantly exceeded unity. These findings suggested that [NO₃^-]_PM2.5 was influenced by primary sources; however, primary emissions alone cannot fully account for the observed increase.

On 28 November, the concentrations of NH₄⁺ and Na⁺ in PM_2.5 exhibited diurnal variations similar to those observed on 27 November, although with substantially larger fluctuations. Although no statistically significant correlation was identified between [NO₃^-]_PM2.5 and NH₄⁺ concentrations in PM_2.5 (P > 0.05, Figure S2c), NH₄⁺ showed a strong correlation with the equivalent concentrations of (NO₃^- + SO₄^2-), with a regression slope close to unity. Based on this, using 20:00-23:00 as the analysis window (t₀, t₁, t₂, t₃), the net increases in NH₄⁺ concentrations at t₁-t₃ relative to t₀ can be calculated by the difference. Assuming NH₄⁺ was fully associated with (NO₃^- + SO₄^2-) at a 1:1 equivalent concentration ratio, the corresponding net increase in NO₃^- attributed to NH₄NO₃ at t₁-t₃ was calculated to be 0.02 µg m^-3, 0.04 µg m^-3, and 0.05 µg m^-3, respectively. These contributions account for 3.0%, 2.4%, and 4.2% of the total net increase in [NO₃^-]_PM2.5 during the same periods. Recall that thermodynamic equilibrium stimulation analysis indicated the volatilization of NH₄NO₃ into gas phase. The significant increase in [NO₃^-]_PM2.5 between 21:00 and 23:00 was likely related to the formation of other nitrates, such as KNO₃ and NaNO₃. However, the minimal increase in Na⁺ concentration in PM_2.5 during this period (Figure 5a) suggested that NaNO₃ may accounted for only a small fraction. In contrast, a marked increase in K⁺ concentration was observed in PM_2.5 between 21:00 and 23:00 (Figure S4), indicating that biomass burning emissions may be a major contributor [69,70]. Concurrently, atmospheric NH₃ also increased (Figure 5b), which may similarly be attributed to biomass burning emissions [33]. Other anthropogenic or natural sources of NH₃ during nighttime are considered less likely [18,33,71]. It is noteworthy that, compared with 27 November, the nucleation mode in the particle number size spectrum nearly disappeared during the NO₃^- increase, suggesting that the biomass burning aerosols may have undergone atmospheric processing, such as cloud modification under non-precipitating conditions [60]. Therefore, the significant increase in [NO₃^-]_PM2.5 during 21:00-23:00 was more likely driven by secondary formation processes within non-precipitating clouds, rather than direct primary emissions. Interestingly, a secondary increase in K⁺ was also observed between 01:00 and 03:00 on the same day (Figure S4), accompanied by a pronounced rise in the particle number concentration within the accumulation mode (Figure 5c). Surprisingly, the increase in concentrations of NO₃^- and atmospheric NH_3gas during this period were almost negligible. Therefore, primary biomass burning aerosols and non-precipitation modification alone were insufficient to explain the increase in [NO₃^-]_PM2.5 between 21:00 and 23:00, leaving additional anthropogenic‌ sources as co-contributors. Importantly, no significant shift was observed in the air mass back trajectories before and after the [NO₃^-]_PM2.5 increase (Figure 5e).

3.5. Case 3: Transition from Regional Transport to Local Combustion Sources Induces Strong Phase Instability and NH₄NO₃ Volatilization

On 29 November 2023, [NO₃^-]_PM2.5 exhibited two distinct periods: from 00:00 to 18:00 and after 18:00. During the first period, the concentration changes of NO₃^- and NH₄⁺ were consistent with the trends observed in PM_2.5 concentrations at the national monitoring stations (Figures 6ab), suggesting a common regional source. In contrast, after 18:00, [NO₃^-]_PM2.5 rapidly increased from 1.4 µg m^-3 to a peak of 3.9 µg m^-3, without a corresponding rise in PM_2.5 concentrations at the national stations. Furthermore, the correlation between [NO₃^-]_PM2.5 and NH₄⁺ concentrations over the full day (R² = 0.34) was significantly lower than that during the first period (R² = 0.92) (Figure S2e). These results point to the hypothesis that the increase in [NO₃^-]_PM2.5 after 18:00 was primarily attributable to local sources. Moreover, the air mass trajectories during the period of elevated [NO₃^-]_PM2.5 were in between two slightly separately trajectories before 18:00 (Figure 6e).

However, whether the equivalent concentrations were calculated using data from 00:00-23:00 or from 00:00-18:00, the results consistently showed that [NH₄⁺] = 0.9 * ([SO₄^2-] + [NO₃^-]) + 0.02, with an R² of 0.99 and P < 0.01 (Figure S1f). This suggested that NO₃^- in PM_2.5 both before and after 18:00 primarily existed in the form of NH₄NO₃. Consequently, the observed decrease in [NO₃^-]_PM2.5 between 00:00 and 18:00, along with the declining proportion of NH₄NO₃ among ammonium salts, may be attributed to the volatilization of NH₄NO₃, consistent with thermodynamic simulations aforementioned. However, the amount of NH_3gas released through this process was considerably smaller than the total reduction in ambient NH_3gas concentrations. The concurrent decline in atmospheric NH_3gas levels may have further promoted the volatilization of NH₄NO₃ (Figure 6b).

Additionally, the diurnal variations of Na⁺ and NO₃^- in PM_2.5 were nearly opposite (Figure S3c), indicating that NO₃^- associated with NaNO₃ in PM_2.5 was likely negligible. The particle number size distributions demonstrated that after 18:00, the particle number concentration was dominated by the smaller-sized Aitken mode (Figure 6c). The median Aitken mode diameters showed significant variation at minute-scale resolution and showed an overall increasing trend, although it remained significantly smaller than the accumulation mode diameter dominant before 06:00 that day. According to particle emission fingerprints from different sources reported in the literature [72], the particles observed after 18:00 were most likely emitted from combustion sources rather than formed via ambient nucleation processes [73].

Further analysis of the normalized [NO₃^-]_PM2.5 and N_>100 revealed a strong correlation during the period from 00:00 to 18:00, with a regression slope of 1.5. Once Again, primary emissions alone could not fully account for the observed [NO₃^-]_PM2.5. However, it is likely that secondary NO₃^- formation occurred earlier during air mass transport to the observation site, considering that NH₄NO₃ appeared to have predominantly volatilized from the particle phase. This will be further confirmed in the following thermodynamic simulations. In contrast, after 18:00, the correlation between normalized [NO₃^-]_PM2.5 and N_>100 weakened significantly, and the regression slope increased to 8.1. The slope value alone implied that NO₃^- in PM_2.5 was mainly derived from secondary sources, although the observed HNO₃*_gas concentration remained relatively low (0.13-0.21 µg m^-3). Therefore, the pathway of large-scale N₂O₅ production, which could promote NH₄NO₃ formation, is ruled out. On the contrary, NH₄NO₃ may have formed rapidly during the cooling of combustion plumes, followed by dilution and NH₃-limited volatilization [23,54].

3.6. Case 4: Primary Emissions and Marine Aerosol Interactions Drive Diurnal Sea-Salt-Restricted Nitrate Formation Under Southeast Monsoon Influence

Figure 7a-h showed a unique case observed under the influence of southeast monsoon. During this period, the nearby natural gas-fired power plant was intermittently located directly upwind of the sampling site. On 16 June 2024, the concentrations of NO₃^-, Na⁺, and NH₄⁺ in PM_2.5 exhibited distinct temporal patterns before and after 12:00 (Figure 7a). After 12:00, [NO₃^-]_PM2.5 exhibited a bell-shaped temporal that closely resembled the pattern of Na⁺ in PM_2.5. The strong correlation between [NO₃^-]_PM2.5 and Na⁺ in PM_2.5 supported the heterogenous formation of NaNO₃ through the reaction of HNO_3gas with sea-salt aerosols at the time [74]. In contrast, the observed low concentrations of HNO_3gas and the thermodynamic simulations didn’t support the formation of NH₄NO₃ during the moments. Note that the diurnal variation of Na⁺ concentrations in PM_2.5 on that day was reasonably consistent with the tide height (Figure S5).

Before 12:00 on that day, the concentrations of NO₃^-, Na⁺, and NH₄⁺ all increased synchronously, exhibiting a bimodal distribution with peaks at 07:00 and 09:00. Three peaks of HNO₃*_gas were observed before 12:00 ranged from 2.1 to 2.9 µg m^-3, which were approximately an order of magnitude higher than other values recorded throughout the study. However, low concentrations of OC were recorded at only 1.3-1.6 µg m^-3 during the peak periods. Although the concentration of EC increased to 0.7 μg m^-3 between 08:00 and 10:00, the ratio of OC/EC was only 2.1, a typical value for primary particulate carbonaceous emissions [75]. Moreover, PM_2.5 levels at three national sites did not show a proportional rise during these peak times, indicating no large-scale influx of air pollutants. Therefore, the large sawtooth-pattern increases in [HNO₃*_gas] and [NO₃^-]_PM2.5 were more likely associated with primary emissions from the nearby natural gas-fired power plant, rather than the secondary formation of HNO₃*_gas in the atmosphere followed by the heterogeneous reaction with sea-salt aerosols [74]. In this case, the amount of sea-salt aerosols appeared to be a controlling factor for [NO₃^-]_PM2.5, considering that no corresponding peak in [NO₃^-]_PM2.5 occurred concurrently with the third peak of HNO₃*_gas in the absence of [Na⁺]_PM2.5 peak. The trough of [Na⁺]_PM2.5 at 08:00 was likely due to temporary wet scavenging when low cloud crossed over the hill (Figure S5).

Unlike those on 27-29 November 2023, there was no significant correlation between the normalized [NO₃^-]_PM2.5 and N_>100 with P > 0.05 on 16 June. The particle number size distributions showed an increase concentrations of accumulation mode particles after 08:00, which coincided with the increase in EC and NH₄⁺ in PM_2.5. These results suggested that the strongest emissions from the nearby power plant likely occurred between 08:00 and 12:00, which were apparently consistent with these larger values of HNO₃*_gas.

3.7. Source Apportionment of NO₃^- in PM_2.5 During the Four Cases

The analysis of observation cases from 2023 and 2024 highlights the critical role of both primary and secondary sources in partitioning the gas-particle distribution of nitrate. Therefore, this study quantitatively evaluated the contributions of aerosols from various sources to nitrate using the PMF model (Figure 8). In November, NO₃^- in PM_2.5 was predominantly attributed to Factor 2 and Factor 5 only. Factor 5 contributed between 52% and 73%, while Factor 2 contributed between 26% and 46%. By contrast, on 16 June, Factor 1 made a limited contribution (4.6%), whereas the contribution from Factor 5 increased to 79%, and Factor 2 decreased to 16%.

Factor 1 exhibited a unimodal particle size distribution, with a median diameter centered approximately 40 nm. This source factor displayed distinct peaks during both morning and evening rush hours, suggesting a potential association with non-electric vehicular emissions [72,73,76,77]. The relatively weaker morning peak may be attributed to the widespread use of electric motorcycles for commuting in tropical regions. In contrast, Factor 2 presented a complex trimodal size distribution, with the dominant mode characterized by a median particle diameter exceeding 200 nm. Despite this, its total number concentration was the lowest among all five factors, approximately an order of magnitude lower than those of Factors 1, 3, and 4, indicating that it likely represents a regional background source, particularly of marine origin [72,78]. Factor 3 demonstrated a bimodal distribution, with prominent peaks at approximately 70 nm and a minor mode below 20 nm. This factor was likely attributed to emissions from Chinese-style cooking, where high-temperature oil use leads to substantial particle generation, particularly between 18:00 and 20:00. However, its contribution to NO₃^- in PM_2.5 was found to be negligible. Factor 4 was dominated by nucleation-mode particles, indicative of short atmospheric nucleation times. Its occurrence was more frequent during nighttime, implying that these events were likely influenced by transient coastal atmospheric conditions [78,79]. Correspondingly, this factor also exhibited minimal contribution to NO₃^- levels in PM_2.5. Lastly, Factor 5 displayed a bimodal particle size distribution and was primarily associated with emissions from a nearby natural gas-fired power plant (~1 km distance), with additional influence from regional background sources. Factor 5 exhibited a distinct trough in the particle size distribution around 100 nm, a feature resembling the Hoppel-effect modified distribution under an environmental supersaturation of 0.2%. This suggests that the gas-fired power plant plume, when emitted into an environment with a substantial temperature contrast, may promote the formation of microdroplets, thereby leading to an aerosol size distribution akin to that observed in non-precipitating cloud processing. Consequently, it is anticipated that Factor 5 will contribute a higher mass concentration of nitrate in winter compared to summer (Figure 8).

As shown in Figure 1, the gas-fired power plant emits substantial amounts of water vapor droplets, which facilitate the formation of liquid-phase (NH₄⁺ + NO₃^-). The residence time of the vapor droplets, ranging from several tens of seconds to a few minutes, is primarily influenced by ambient temperature and atmospheric humidity; lower temperatures and higher humidity extending their residence time [23]. This may account for the elevated NO₃^- concentrations observed in PM_2.5 during the humid November nights. It is evident that the (NH₄⁺ + NO₃^-) formed rapidly through water vapor condensation and subsequent flue gas dilution is likely to volatilize from the particulate to the gas phase. Compared to Factor 2, Factor 5 appears to represent a more unstable component, as approximately 20% of high-concentration NO₃^- cases could not be resolved in the PMF analysis (Figure S6). In these instances, the PMF model clearly underestimates NO₃^-, possibly because the NO₃^- captured by the analysis reflects aged and relatively stable NH₄NO₃. In contrast, high-concentration NO₃^- in PM_2.5 is relatively fresh and in the NH₃-restricted volatilization process of NH₄NO₃. It is worth noting that a slight difference in the extracted profile of each factor was observed when NO₃^- concentrations were used to replace those in the final particle size bin (Figure S7), which may be attributed to subtle differences in their respective sources.

4. Conclusions

This study systematically investigates the dual regulatory effects of atmospheric NH₃ on the formation and evaporation of NH₄NO₃ at a clean tropical coastal site. By integrating high-resolution field measurements, E-AIM modeling, and PNF source apportionment analysis, we demonstrate that NH₃ plays a role in suppressing the volatilization of NH₄NO₃ under varying atmospheric conditions rather than acting solely as a precursor promoting its formation.

Multiple case studies revealed that NH₄NO₃ in PM_2.5 originated from both primary emissions-particularly from local combustion plumes and in-cloud processing of KNO₃ and/or NH₄NO₃, especially under the influence of the northeast monsoon. Under NH₃-rich but HNO₃-limited conditions, particle-phase NH₄NO₃ tended to volatilize as confirmed by thermodynamic modeling and various supporting lines of evidence. Source apportionment using the PMF model indicated that regional background transport and emissions from a nearby natural gas power plant were the dominant contributors to nitrate in PM_2.5. Under the southwest monsoon, the clean marine air mass was not conductive to NH₄NO₃ formation even when perturbed by plumes from local gas-fired power plants. In the case, the sea-salt-aerosol-restricted reactions between HNO₃*_gas and sea-salt aerosols dominantly contributed to NO₃^- in PM_2.5.

These findings challenge the conventional understanding of nitrate formation as purely precursor-limited, highlighting the importance of simultaneously accounting for formation and volatilization dynamics. This study refines the understanding of NH₄NO₃ behavior in ammonia-rich marine boundary layers and has important implications for optimizing future nitrate control strategies in coastal and urban environments.