Introduction

The World Health Organization (WHO) reports that air pollution leads to several million premature deaths a year1. The main culprit, haze pollution, is a global issue and brings a wide range of negative health impacts (such as cancer, respiratory and heart diseases) to human body2,3,4,5. Nowadays, particle pollution risks still occur in developed countries6,7 and are urgent in developing countries5,8,9. Previous study has proposed that up to 77% PM2.5 (particulate matter with an aerodynamic diameter less than 2.5 μm) was contributed by secondary aerosol formation5. Therefore, understanding the main drivers (atmospheric oxidants and oxidation pathways) for secondary aerosol formation is of significance for figuring out haze pollution10,11.

As part of the haze production, nitrous acid (HONO) photolysis contributes to more than 50% of hydroxyl radical (OH), which strongly impact the atmospheric oxidation capacity12,13,14. In the case of the North China Plain, the more serious the pollution, the greater the contribution from HONO to OH (the contribution reached up to 93% during heavily polluted episodes)15. The OH levels contributed by HONO facilitates the secondary processes during haze pollution. Therefore, HONO is the underlying controlling factor of haze, especially under severe polluted periods16,17.

Generally, models underestimate HONO, which impacts OH predictions and ultimately secondary aerosol formation (especially in highly populated regions)15,18,19. In model simulations, the most common HONO source is homogeneous reaction of OH and nitric oxide (NO)20,21. It is also known that HONO sources include direct emissions22,23,24, heterogeneous reactions of nitrogen dioxide (NO2) on wet surfaces25,26,27, photo-enhanced NO2 heterogeneous conversion28, and photolysis of atmospheric NO3- and surface adsorbed nitric acid (HNO3)29,30,31,32, etc. Also, there were some influencing factors for HONO formation mechanisms such as RH-promoted NO2 heterogeneous conversion33,34, NH3-promoted NO2 hydrolysis35, oxidation of SO2 by NO236, etc. Even though some model simulations could well simulate HONO observations, they use the single NO2 uptake coefficient to measure NO2 hydrolysis and not considered any known influencing factors such as RH15,18. Previous studies have proposed the parameterization of RH factor to NO2 heterogeneous conversion, but there is still a big difference with the observed HONO37,38. At present, the controversy for HONO formation focused on the quantification and parameterization scheme of NO2 heterogeneous reactions (the dominant source during nighttime and polluted days39). Indeed, despite the progress in the development of HONO formation mechanisms, a HONO model parameterization scheme that can be widely used to accurately predict HONO concentration remains elusive.

Here, the atmospheric concentrations of HONO and related species were simultaneously measured in urban Beijing, from October 25 to December 7, 2018. We developed the new parameterization scheme of NO2 conversion to HONO. Via observation-based box modeling involving Master Chemical Mechanism (MCM v3.3.1) simulations, we test the HONO parameterization scheme out for Beijing and it works very well. The more accurate HONO quantification here can help to accurately estimate OH and ultimately haze formation, which has implications for global and regional models everywhere.

Results and discussion

Field observations and HONO variation characteristics

Continuous measurements of HONO and related species (including environmental parameters, gaseous species, particle constituents, photolysis frequencies, volatile organic compounds, etc) were conducted in wintertime Beijing, from October 25 to December 7, 2018. The variation of meteorological parameters and chemical species concentrations during field observation period is illustrated in Fig. 1. The observed average mass concentration of PM2.5 and PM10 were 63 ± 63 μg/m3 and 102 ± 84 μg/m3, with maximum values of 275 μg/m3 and 810 μg/m3, respectively. Due to the occurrence of dust storms, the PM10 concentration was >320 μg/m3 on November 27 (00:00–06:00) and December 3 (03:00–10:00), along with high wind speed (WS), high boundary layer height (BLH), and low RH conditions. To exclude the impact of dust storms, the analysis is focused on non-storm period. During the field observation period, four pollution episodes occurred as shown in the shadow regions in Fig. 1 and Supplementary Table 3, i.e., pollution episode (PE) 1–4. Before each pollution episode, there was a gust of wind (higher than 2 m/s) from the northwest and the northeast, carrying relatively clean air masses from the less populated northern mountainous areas. During the four episodes, low wind speed (WS < 2 m/s) led to more stable meteorological dispersion conditions, with a higher RH and lower BLH compared to unpolluted periods40,41. The PM2.5 properties showed a typical periodic cycle within 4–7days42, with mass concentrations <30 μg/m3 at the beginning of each cycle, then developing to up to 275 μg/m3 within 2–5 days40. Each individual pollution episode had three successive days that exceeded the China National Ambient Air Quality Standard (CNAAQS, 75 μg/m3 for a 24 h average). RH exhibited a similar tendency with PM2.5. The average PM2.5 concentration and RH in four pollution episodes were 102 ± 71 μg/m3 and 43 ± 14%, 115 ± 81 μg/m3 and 50 ± 18%, 89 ± 62 μg/m3 and 35 ± 11%, and 95 ± 56 μg/m3 and 41 ± 17%, respectively. These four selected pollution episodes were representative of typical periodic haze pollution cycles in urban China, which were associated with stagnant meteorological conditions such as low air diffusion rates and high humidity40,43. Analysis on these pollution episodes provides important insights into the HONO formation mechanism in Beijing haze, as discussed below.

Fig. 1: Time series (local time) of meteorology parameters and chemical species from October 25 to December 7, 2018 in Beijing.
figure 1

a Wind speed (WS) colored by wind direction (WD) and boundary layer height (BLH) above ground level. b Relative humidity (RH) and temperature (T). c PM2.5 and PM10 concentrations. d NOX (NO + NO2) and HONO concentrations. e NO2 and PM2.5 nitrate (NO3-) concentrations. f SO2 and PM2.5 sulfate (SO42-) concentrations. g NH3 and PM2.5 ammonium (NH4+) concentrations. h O3 and jHONO concentrations. i CO concentrations. The gray shaded areas are four pollution episodes, named PE1, PE2, PE3, PE4.

Full size image

In each pollution episode, the concentrations of HONO, NO2, CO, and NH3 showed a gradual increase with increased concentration of PM2.5 compositions (e.g., NH4+, SO42-, NO3-), implying the potential link between these gaseous species and the formed PM2.544. SO2 increased with SO42- developed in PE1 but decreased with SO42- developed at the end stages of PE2 and PE4 (suggesting the fast conversion of SO2 to SO42-). O3 concentration generally stayed at a low level due to the weak photochemical reactivity during haze pollution, and possessed a negative relationship with PM2.5 concentration. The highest O3 concentration was observed in PE1 (daytime (8:00−16:00) average value was 18 ± 15 ppb), together with the highest HONO photolysis frequency (jHONO, daytime average value was (4.3 ± 1.7)×10-4 s-1), which reflected the highest solar radiation among four episodes. NOX (NO + NO2, the major precursor of HONO) showed similar variation trends with HONO. The diurnal variation profile of HONO, NO, and NO2 are summarized in Supplementary Fig. 1, similar to those reported by previous haze studies45,46,47. Specifically, daytime photoreactions and lifted BLH lead to low diurnal concentrations of HONO with the minimum value at noon (around 15:00). After sunset, BLH gradually declined and pollutants accumulated until the early morning, resulting in high HONO concentrations during nighttime. The observed HONO concentrations were higher in PE2 (2.83 ± 1.84 ppb) and PE4 (3.04 ± 1.71 ppb) compared to PE1 and PE3, especially at midnight and in the early morning. However, higher NO and NO2 values were observed in PE1 (38.06 ± 35.12 ppb, 35.82 ± 15.97 ppb) and PE4 (38.71 ± 24.42 ppb, 30.98 ± 6.48 ppb) compared to PE2 and PE3. Nocturnal NO concentrations were sometimes higher than NO2, which revealed the importance of freshly released plumes. Overall, the relationship between HONO and NOX is not straightforward, indicating that the formation of HONO is controlled by multiple factors.

The overview of daytime (8:00–16:00), nighttime (18:00–06:00), and daily mean concentrations of HONO, NO2, NO, and PM2.5 during four episodes investigated here and the previously reported data during winter haze events are summarized in Supplementary Table 4. During the four pollution episodes, the average values of HONO, NO, NO2, and PM2.5 were 2.52 ± 1.61 ppb, 35.14 ± 30.27 ppb, 32.23 ± 7.43 ppb, 100 ± 69 μg/m3, respectively, which were at an intermediate level compared with previous studies in China48,49,50 and other countries such as India8,9. The ratio of HONO to NOX (i.e., the conversion frequency from NOX to HONO), was around 0.03 among previously reported winter haze events in China before 201815,45,46,47,51. In this study, the average HONO/NOX value was 0.039 ± 0.021, implying a higher NOX to HONO conversion frequency level in recent years.

The net homogeneous reaction rate (Pnet, including the reactions of OH with NO and OH with HONO as shown in Supplementary Note 3) in four episodes during 11:00–14:00 were 0.66 ppb/h, which was about 40% higher than the previous observations48,52. The daily average NO concentrations in four episodes were above 29 ppb. Recent studies also reported high NO concentrations in China48,51. Therefore, the homogeneous source (OH + NO → HONO) may play an important role in the formation of HONO. Vehicle emission is the most important direct emission source of HONO in urban areas53,54. As shown in Supplementary Fig. 2, the correlation coefficient (R2) between HONO and NOX was 0.61 in PE1, followed by 0.56 in PE3, 0.40 in PE2, and 0.29 in PE4. We note that when HONO concentrations are below 3 ppb, the R2 between HONO and NOX increased to 0.81, 0.42, 0.38, and 0.35 for PE 1–4, respectively. However, for more polluted periods during which HONO concentration was higher than 3 ppb, no significant correlation was found between HONO and NOX. As shown in Supplementary Note 2, HONO emission factor inferred by the ratio of HONO to NOX emissions was 0.00973. Direct emission factor can be corrected by multiplying the ratio of the τ(HONO)/τ(NOX) (HONO lifetime/NOX lifetime) when τ(HONO) was less than 1h55. Contributions of vehicle emitted HONO to total observed HONO (HONOemission/HONO) was higher in PE1 (28%) and PE3 (27%), followed by PE4 (21%), and PE2 (18%). The peaks of HONOemission/HONO usually appeared in morning rushing hours (around 08:00) and evening rushing hours (around 22:00). HONOemission/HONO was larger during periods with lower-level HONO concentrations, which corresponded to lower-level PM2.5 concentrations. The results implied that the contribution of vehicle emission to HONO was more significant on less polluted days, while was less important on more polluted days (i.e., with higher PM2.5 concentration). The heterogeneous NO2 conversion fraction (HONOcorr/NO2, HONOcorr = HONO – HONOemission) was used to evaluate the impact of heterogeneous NO2 conversion. As shown in Supplementary Fig. 4, HONOcorr/NO2 in PE2 (7.6%) and PE4 (7.6%) were higher than those in PE1 (3.5%) and PE3 (5.0%), implying the potential stronger capacity of heterogeneous conversion from NO2 to HONO in PE2 and PE4. In theory, the NO2 → HONO conversion can take place on both aerosol surface and ground surface. The positive correlation of HONOcorr with NO2, PM2.5, and NO2 × PM2.5 in Supplementary Fig. 5 demonstrated that NO2 heterogeneous reactions on PM2.5 surfaces were important, especially in PE2 and PE4. Other NO2 heterogeneous pathways may also influence HONO production when RH is higher than 70%. As shown in Supplementary Fig. 6 and Supplementary Discussion, higher RH (>70%) and PM2.5 (>150 μg/m3) in PEs could accelerate NO2 heterogeneous conversion, the corresponding episodes were denoted as SPEs (severe pollution episodes).

Model simulations and unknown HONO sources

The box model simulations with Master Chemical Mechanism, MCM v3.3.1 mechanisms were used to reveal the relative importance of different HONO sources in urban Beijing. The base HONO source in model simulations includes homogeneous OH + NO reaction, and the base HONO loss pathways include HONO photolysis, OH + HONO reaction, and HONO deposition51,56,57. The base model (HONObase) can only explain 8% of observed HONO in PEs (Supplementary Fig. 7), which can further underestimate OH and O3 in the atmosphere18,58. Therefore, we considered other known HONO formation mechanisms (including NO2 heterogeneous reaction on the ground surface and photo-enhanced NO2 heterogeneous reaction on the ground surface (HONOground, ppb), NO2 heterogeneous reaction on the aerosol surface and photo-enhanced NO2 heterogeneous reaction on the aerosol surface (HONOaerosol, ppb), surface adsorbed HNO3 photolysis (HONOHNO3, ppb), atmospheric NO3- photolysis (HONONO3-, ppb), and HONOemission) in the box model simulations (Table 1). Table 1 represents a general consideration for HONO sources parameterization scheme. After including these HONO sources in model simulations, the simulated HONO concentrations (HONOsim, HONObase + HONOground + HONOaerosol + HONOHNO3 + HONONO3- + HONOemission) can explain about 85% of the observed HONO in PEs but only 12% in SPEs. The unknown HONO concentrations can be shown as HONOunknown (observed HONO minus HONOsim) The contribution ratio of HONOunknown (HONOunknown/HONO) was 2%, 26%, 11%, and 15% in PE1–4, respectively. With PM2.5 concentration increased in a given pollution episode, HONOunknown/HONO also increased (Supplementary Fig. 10). Figuring out HONOunknown could provide a previously unrecognized oxidation pathway in the atmosphere through HONO photolysis. The unrecognized oxidation pathway can further explain the exceedingly high levels of particle concentrations, which were generally significantly underestimated in previous studies10,11.

Table 1 Parameters of HONO source/loss mechanisms in model simulations.
Full size table

Through linear regression analysis for HONOunknown and gaseous species concentration and environmental conditions, we found that CO, NH3, PM2.5, and RH had good correlations with unknown HONO concentrations (Fig. 2). The black and red dots in Fig. 2 represent PEs (except SPEs) and SPEs, respectively. There was no direct connection between CO and HONO production. Due to the strong correlation between CO and PM2.5 (R2 = 0.72) and the homology of CO, NH3, and NOX, HONOunknown showed a good correlation with CO. Therefore, we focused on the effect of RH, NH3, and PM2.5 on HONO production. To explore the effect of RH and PM2.5 on HONO formation, ISORROPIA-2 thermodynamic equilibrium model simulations were performed to determine aerosol liquid water content (ALWC) and aerosol pH. Generally, when RH is below 40%, ALWC is very small and the estimated aerosol pH is highly uncertain59. For that reason, we only analyzed the data with RH higher than 40%. The average values of ALWC in four episodes were 40.7, 86.6, 32.5, and 59.3 μg/m3, respectively. The average aerosol pH values in PE1–4 were 4.57, 4.39, 4.92, and 4.93, respectively, which were buffered by the conjugate acid-base pair NH4+/NH360.

Fig. 2: Relationship between HONOunknown and main influencing factors in PEs.
figure 2

The influencing factors include (a) RH, (b) PM2.5, (c) NH3, and (d) CO. Red scatters are data with RH > 70% and PM2.5 >150 μg/m3, named SPEs.

Full size image

During haze events, higher RH (>40%) provides higher ALWC and improves the NO2 uptake coefficient on aerosol surfaces61,62. Zheng et al.10 proposed that the formation rate of sulfate and nitrate generated by heterogeneous reactions to form HONO increased with the increase of RH, including the NO2 heterogeneous reaction on the aerosol surfaces. Previous study suggested that NH3 concentrations of higher than 10 ppb would be sufficient to the new particle formation in the highly polluted atmosphere63. In this study, NH3 and HONOunknown had clearly positive correlations when NH3 was higher than 11 ppb (corresponded RH was higher than 40%), which indicated the effectiveness of NH3 and RH promoted HONO formation. Therefore, the enhanced NO2 heterogeneous reaction on the aerosol surface under higher RH conditions may be regulated by NH3 concentrations. Li et al. have proposed that NH3 can promote NO2 hydrolysis at the air-water interface to produce HONO, which may be an important process at polluted sites where NH3 is rich35. The role of NH3 is to stabilize the reaction system and promote HONO formation. Therefore, the enhanced NO2 heterogeneous reaction on the aerosol surfaces when RH > 40% can be explained by the effect of NH3 on the aerosol. NH3, as an alkaline gas in the atmosphere, plays an important role in aerosol pH. High aerosol pH increases the solubility of acidic gas and thus increases the reaction rate64,65. Ding et al.66 proposed a relationship between NH3 and aerosol pH: variation of NH3 in a low concentration range has a significant effect on aerosol pH, while the variation of NH3 in a high concentration has little effect on aerosol pH. Hence, when RH was higher than 40%, the enhancement effect on NO2 heterogeneous conversion on the aerosol surfaces could be expressed as NH3×k/(NH3 + 10). k is the coefficient measuring the combined effect of NH3 and RH on HONO formation. Assuming that the unexplained HONO concentrations come from the enhanced NO2 heterogeneous conversion, we try to calculate the potential factor between unknown HONO production rate and RH, NH3. Then, the k of 1.93 can better measure the HONO concentrations when \(40 \% \,<\, \text{RH}\le 70 \%\).

In SPEs, higher slopes were found between HONOunknown and CO, NH3, PM2.5, RH compared with other periods. Furthermore, the slope between HONOunknown and NO2 was higher than in the other periods (Supplementary Fig. 8), suggesting a more rapid conversion rate from NO2 to HONO. The PM2.5 constituents have a strong hygroscopic capacity, aerosols can continuously absorb ambient water until the aerosol particles are in equilibrium with ambient air. Indeed, ALWC increased exponentially with RH (Supplementary Fig. 9)67. When RH is higher than 70%, PM2.5 particles were in a deliquesced state. Based on the high gas-to-particle conversion under high RH, the NO2 heterogeneous conversion frequency on the aerosol surfaces would be enhanced in SPEs68,69. NO2 hydrolysis could be accelerated sharper when RH > 70% than that when \(40 \% \,<\, \text{RH}\le 70 \%\)10,37.

Additionally, the higher intercept between SO42- and NO3- in SPEs relative to other periods suggested the higher conversion ratio from SO2 to SO42- (Supplementary Fig. 8). A recent study showed that HONO reacts with SO2 to form SO42- at pH values greater than 4.5 during cold winter haze days70. Given that the average aerosol pH in SPEs was 4.28, HONO-SO2 chemistry is unlikely to play an important role in the formation of SO42-. Furthermore, Liu et al.36 have performed a series of experiments with deliquesced particles at pH values of 4–5 with experimental conditions similar to our observation conditions, which indicated that (1) SO2 oxidation by HONO was negligible, (2) the NO2 and SO2 reaction at NH3 buffered pH values of 4–5 on aerosol surfaces have atmospheric importance under haze conditions, which was mainly contributed by \({\text{NO}}_{2}+{\text{SO}}_{3}^{2-}\) reaction, (3) the increased RH and NH3 concentrations significantly increased NO2 and SO2 reaction rate. In our study, apart from the more rapid NO2 to HONO conversion frequency and SO2 to SO42- conversion frequency in SPEs, there was also a change in the relationship between NO2 and SO42- (Supplementary Fig. 8). The increased SO42- concentrations in SPEs may arise from the oxidation of SO2 by NO2 on the deliquesced aerosol particles, along with the production of HONO36. pH values of 4–5 were predicted for ambient aerosols during haze events by thermodynamic models36,71,72, which was similar to the pH in SPEs.

$$2{\text{NO}}_{2}+{\text{SO}}_{3}^{2-}+{\text{H}}_{2}\text{O}\to 2\text{HONO}+{\text{SO}}_{4}^{2-}$$
(1)

As shown in Supplementary Fig. 8, because NO2 hydrolysis and \({\text{NO}}_{2}+{\text{SO}}_{3}^{2-}\) reaction were both NO2 heterogeneous reactions, positive correlation was found between HONOunknown and NO2 in SPEs. However, SO42-/NO3- was part of the NO2 reaction product, leading to the not positively correlated relationship between SO42-/NO3- and NO2 in SPEs. Summing up, the NO2 heterogeneous conversion (including NO2 hydrolysis and \({\text{NO}}_{2}+{\text{SO}}_{3}^{2-}\) reaction) should be considered in SPEs. Previous studies have suggested that the NO2 uptake coefficient can be increased fourfold under high humidity (RH > 70%) due to the RH enhancement effects10,37. Additionally, NH3 can promote NO2 hydrolysis35. Previous field observations, laboratory experiments, and molecular dynamics simulations have proposed that NH3 can accelerate NO2 and SO2 reaction73,74,75. The enhancement effect for NO2 hydrolysis and \({\text{NO}}_{2}+{\text{SO}}_{3}^{2-}\) reaction conversion rate under SPEs should be considered because: (1) the enhanced ALWC accelerated the NO2 heterogeneous conversion to form HONO on the aerosol surface; (2) RH of higher than around 70% corresponded to the PM2.5 particles in a deliquesced state; (3) the heterogeneous reactions of NO2 and SO2 on the deliquesced aerosol particles with pH values of 4–5 accelerated HONO formation. Even though the relative contribution of NO2 and SO2 reaction and NO2 hydrolysis to HONO formation may be different, the inner reason was the effect of aerosol pH and deliquesced particles to increase the uptake of NO2 on the aerosol surface. Therefore, under seriously polluted conditions (RH > 70% and PM2.5 > 150 μg/m3), an enhancement factor of 4 multiplied by \(\frac{{\text{NH}}_{3}\,\times\, 1\text{.}93}{{\text{NH}}_{3}\,+\,10}\) was used to quantify the impact of NH3, RH on the NO2 uptake coefficient on the aerosol surface.

Overall, according to the differences between the model results and the observations, the correlation analysis between unknown HONO and other species, and experimental and theoretical evidence from the previous studies35,36,73,75, the enhancement factor (fNH3, RH, as follows) for NO2 uptake coefficient on the aerosol surface (divided by RH and PM2.5 concentrations) was included in model simulations.

$${f}_{{{\rm{NH}}}_{3},{\rm{RH}}}=\left\{\begin{array}{ll}1, & {\rm{RH}}\le 40 \% \\ \frac{{{\rm{NH}}}_{3}\times 1.93}{{{\rm{NH}}}_{3}+10}, & 40 \% \,<\, {\rm{RH}}\le 70 \% \\ \frac{{{\rm{NH}}}_{3}\times 1.93}{{{\rm{NH}}}_{3}+10}\times 4, & {\rm{RH}} \,>\, 70 \% ,{{\rm{PM}}}_{2.5} \,>\, 150{{{\upmu}}{\rm{g}}}/{{{\rm{m}}}^{3}}\end{array}\right.$$
(2)

Under typical winter haze in Beijing when \(40 \% \,<\, \text{RH}\le 70 \%\), RH and NH3 can promote the NO2 hydrolysis at the air-water interface to produce HONO. As pollution developed when RH > 70%, the aerosols grew fast and ALWC dramatically increased because of the strong hygroscopic capacity of water-soluble ions. The increased ALWC and NH3 can further increase gaseous species (such as NO2) uptake and increase gas-to-particle conversion, which derived both NO2 hydrolysis and NO2 heterogeneous conversion with \({\text{SO}}_{3}^{2-}\) at aerosol pH 4–5. By increasing the NO2 uptake coefficient on the aerosol surface by the factor of fNH3, RH without adjustment, model performance on HONO simulations in PE1–4 was nearly approach to the level of observed HONO (Supplementary Fig. 10). In PE1–4, the final simulated HONO explained observed HONO of about 107%, 97%, 94%, and 97%, respectively. fNH3, RH explained HONO of about 13% in PEs (41% in SPEs). fNH3, RH had a small impact on the period with RH ≤ 70%, but a large impact on the SPEs period. fNH3, RH was an important factor to measure the HONO concentrations during severe polluted episodes.

In PEs, the contributions of sources to HONO formation are expressed in Fig. 3 and Supplementary Table 6. The base model with only one HONO production pathway of homogeneous reaction of NO and OH can contribute to 12% diurnal HONO and 6% nocturnal HONO. In urban sites with large NOX emissions, vehicle emissions were always a considerable nocturnal HONO source. The contribution from the photolysis of nitrate to diurnal HONO formation was about 2% during haze events. The photolysis of HNO3 and NO3- was negligible to HONO formation. Ye31 proposed that jNO3 (s–1) varies from 6.2 × 10−6 to 5 × 10−4. Here, we simulated jNO3 (8.3 × 10−5) by the factor of 5, a variation between −0.6% and 3.2% was proposed. The reasonable variation range implied the possibility of jNO3 change in our observed polluted days. Even so, NO3- photolysis was not that important in HONO formation during severely polluted days for the lack of solar radiation. HONO formation mostly came from NO2 heterogeneous reactions. Heterogeneous reactions on the aerosol surface (including the effect of fNH3, RH, subsequent expressions of heterogeneous reactions on the aerosol surface always include the fNH3, RH factor) and ground surface contributed to HONO of 68%, even contributed 91% in SPEs. After adding the “NO2 heterogeneous reactions on aerosol surface and ground surface” mechanism to the base model simulations, about 76% of daily HONO and most HONO at noontime can be explained. The parameterization of NO2 heterogeneous reactions was crucial to further quantify diurnal OH concentrations in model simulations during pollution episodes.

Fig. 3: Diurnal variations of the HONO sources during four pollution episodes.
figure 3

In (a) PE1, (b) PE2, (c) PE3, and (d) PE4, the light blue area represents the base model simulation (HONObase). The added HONO sources were the dark blue area (HONO concentrations from vehicle emission, HONOemission), light green area (HONO concentrations from NO2 heterogeneous reaction on the aerosol surface and photo-ienhanced mechanisms, HONOaerosol), dark green area (HONO concentrations from NO2 heterogeneous reaction on the ground surface and photo-enhanced mechanisms, HONOground), dark red area (HONO concentrations from photolysis of atmospheric particulate NO3-, HONONO3-). In addition, orange area was the HONO production from the fNH3, RH factor on the aerosol surface (HONOaerosol, NH3, RH).

Full size image

Environmental implications on haze pollution

Beijing is the typical megacity worldwide with condensed populations and buildings. Severe haze pollution generally occurs in autumn and winter. The pollution features include unfavorable diffusion and high humidity, which imply the potential capacity of local chemical conversions. RH and NH3 concentrations can affect the properties of atmospheric aerosol (such as ALWC and aerosol pH) and atmospheric multiphase chemistry60. NH3 is important in the atmosphere, which contributes to the formation of fine particles. NH3-promoted NO2 hydrolysis at the air-water surface, such as on the aerosol particles, is an important HONO formation pathway35. Abundant NH3 concentrations and high RH in China cause aerosol pH values around 4–5, which were favorable for SO2 oxidation by NO2 and further HONO production36,65. Therefore, HONO concentrations increased partly caused by the developed haze and abundant NH3 concentrations. In recent years, NH3 emissions gradually increased or keep stable together with decreased NO2 and SO2 emissions76,77. In Beijing wintertime, NH3 shows an increase of 59% on average from 2009 to 201678 and keeps stable till 201879. Therefore, the role of NH3 and HONO was more important than before.

After constraining the measured HONO concentrations to model, the OH concentrations were obtained (the following are simply expressed as OH concentrations). Diurnal average OH concentrations during 11:00–13:00 in PE1–4 were 1.3 × 106 cm−3, 1.0 × 106 cm−3, 0.9 × 106 cm−3, 0.9 × 106 cm−3, respectively, which were comparable to the levels measured in the winter polluted period of 2017 at a distance of one kilometer from our observation site of 1.5 ± 0.9 × 106 cm−3 during 11:00–13:0080 and in the other wintertime campaigns in urban areas81,82,83,84. HONO was the important species contributing to OH production in PEs (Supplementary Fig. 13)14,15. The reason is that the very low solar radiation during severely polluted days leads to low photolysis frequencies and low ozone concentrations85. We have summarized the contributions from HONO to OH in previous studies in Supplementary Table 7. Calculated by different methods (such as the general data analysis, Box-MCM, Box-RACM, WRF-Chem, CMAQ) in different types of locations, the contributions from HONO to OH are always significant, especially during higher polluted periods. The particulate matter growth during severe pollution episodes was generally underestimated, indicating the missing oxidation processes and/or oxidants in the current model simulations10,11. The OH concentrations from the base model included the contributions from O3, HCHO, HONO (only one production pathway from ON and NO reaction), etc. The base model simulation explained 50% OH concentrations in PEs and 22% in SPEs. The added HONO formation pathways provided OH via HONO photolysis. After adding the HONO formation mechanisms of NO2 heterogeneous reactions on the surfaces (HONOhetero = HONOaerosol + HONOground + HONOaerosol, NH3, RH) in the model simulation, OH concentrations can be explained at 98% in PEs (Supplementary Fig. 11). Notably, HONOhetero contributed 48% OH concentrations in PEs and 78% OH concentrations during SPEs. With the development of haze pollution, the contribution from NO2 heterogeneous conversion to HONO increased then the contribution to OH increased (Fig. 4). Take PE2 as an example, during 11:00 − 13:00 on the first day with PM2.5 concentrations were 7 μg/m3, HONOhetero contributed to HONO and OH of about 80% and 20%, respectively; during 11:00 − 13:00 in the last day with PM2.5 concentrations were 252 μg/m3, HONOhetero contributed to HONO and OH of about 93% and 79%, respectively. On the last day of each episode during 11:00 − 13:00, HONO and OH mostly came from the base model and HONOhetero. Hence, only considering the HONOhetero in the base model simulations, OH concentrations can be well explained during pollution episodes and severe pollution episodes.

Fig. 4: The 11:00–13:00 average concentrations of OH, HONO, and PM2.5 in polluted episodes.
figure 4

In (a) PE1, (b) PE2, (c) PE3, and (d) PE4, the dark color bar represents the HONO and OH concentrations change due to NO2 heterogeneous reactions. The light color bar represents the other HONO and OH concentrations.

Full size image

The OH concentrations increased by HONOhetero provided higher oxidizing capacity surroundings and activated chemical reactions in model simulations. Nitrogen oxidation ratio (NOR, NO3-/(NO3- + NO2)) and sulfur oxidation ratio (SOR, SO42-/(SO42- + SO2)) were 0.74 and 0.98, 0.43 and 0.74, 0.14 and 0.47 in SPEs, PEs, and clean days (PM2.5 < 75 μg/m3), respectively. The high value of NOR and SOR during pollution episodes indicated the important secondary transformation.

NO3- can be produced by the gas-to-particle partition from HNO3. The HNO3 production rate mainly comes from OH + NO2 and N2O5 hydrolysis (P(HNO3) = P(HNO3)OH + P(HNO3)N2O5, Supplementary Note 5), which represents the upper limit of particle NO3- formation. In PEs, P(HNO3)OH and P(HNO3)N2O5 were 0.68 ppb/h and 0.11 ppb/h in the daytime, and 0.06 ppb/h and 0.58 ppb/h in the nighttime. Nocturnal P(HNO3)N2O5 was much higher in PE2 (0.94 ppb/h), PE3 (0.44 ppb/h), and PE4 (0.58 ppb/h) than in PE1 (0.25 ppb/h), which confirmed the importance of heterogeneous reactions in PE2–4. P(HNO3)OH dominated the diurnal HNO3 formation during pollution episodes. HONOhetero can contribute to 0.9 μg/(m3 h) NO3- production at most in PEs.

OH can also accelerate the SO42- formation from SO2 and OH gas-phase reaction. Harris et al. suggested that gas-phase oxidation is less important than aqueous-phase oxidation for SO42- formation, gas-phase SO42- formation accounted for 24% of observed SO42−86. Based on the proportion, HONOhetero can contribute to 1.3 μg/m3 SO42- production at most by the gas-phase oxidation in PEs. Additionally, a recent study proposed that the SO42- formation rate from NO2 and SO2 heterogeneous reaction on aerosols under winter haze conditions at aerosol pH of 5 was comparable to the levels of the missing sulfate source of 3 μg/(m3 h)10,36.

The accurate measurement or quantitative constraints of HONO were the key points to the understanding of the atmospheric oxidation capacity in model simulations during polluted periods. Our study proposed a new approach to elucidate the HONO budget on typical wintertime polluted days in the Chinese megacity, Beijing. Despite the large uncertainties (sensitivity tests in Supplementary Table 8 and Supplementary Discussion), the mechanisms were consistent with our field observation during four typical haze episodes. Besides, we validated these NO2 heterogeneous conversion mechanisms (including the enhancement factor of fNH3, RH) without adjustment in other wintertime polluted periods at the same site. Field observations were from January 2 to 20, 2019, and October 16 to 20, 2019, including four pollution episodes with PM2.5 concentrations varied from 2 μg/m3 to 574 μg/m3 (Supplementary Table 9). After adding HONOhetero in model simulations, 32% OH concentrations during the observed periods and 89% OH concentrations during the severe pollution episodes were explained (RH > 70% and PM2.5 >150 μg/m3). And the OH concentrations simulated by adding parameterized NO2 heterogeneous reactions were comparable to that by constraining observed HONO concentrations. Results demonstrated the validity and the potentially wide applicability of these NO2 heterogeneous conversion mechanisms in urban Beijing during wintertime haze periods. Furthermore, NO2, NH3, and RH data can be obtained from emission inventory and the China Meteorological Administration’s Satellite Data Broadcast System. Hence, fNH3, RH promoted NO2 heterogeneous conversion mechanisms can be widely used in atmospheric model simulations, especially on highly polluted days.

In this work, the important role of NO2 heterogeneous reactions in HONO formation was highlighted. Higher HONO produced more OH, and further more secondary inorganic aerosols formation. Reactions on PM2.5 surfaces would accelerate again. All the above can create a phenomenon where haze pollution and HONO production mutually reinforces generation. To mitigate atmospheric pollution: ALWC is worthy of notice as the media of multiphase reactions; NH3 emission control is proposed for its neutralization effect87,88,89. In urban cities, NH3 and NOX are largely released from combustion processes (power plants, vehicle emissions, coal-burning, etc)90. The combustion emission standard of NH3 should be further optimized. In addition to the haze pollution in Beijing, the HONO formation mechanisms in this study are possibly widespread, accelerating the atmospheric nitrogen cycle and sulfur cycle. To better comprehend atmospheric pollution and human health, it’s essential to measure or parameterize HONO.

Methods

Experimental site and instrumentation

Field observation of HONO and other major air pollutants was conducted on the third floor of No.2 building at Institute of Chemistry, Chinese Academy of Sciences (ICCAS, 39°59′22.68″N, 116°19′21.58″E) in Beijing, which was a typical urban site and described in the previous studies in detail45. The instruments used in this study are listed in Supplementary Table 1. In this place, HONO was measured by a custom-made HONO analyzer, which has been used during previous observation periods45,91. The instrument’s principle was consistent with the long path absorption photometer (LOPAP). The detection limit was 200 ppt. Response time was 13 min. Intercomparisons of commercial LOPAP and custom-made HONO analyzer proved the accuracy and reliability of our instrument92. In addition, a commercially available monitor for gases and aerosols in ambient air (MARGA, 1 S, Metrohm Analytical B.V.) with a PM2.5 inlet was used to quantify the water-soluble ions in PM2.5 (Cl-, NO3-, SO42-, NH4+, Na+, K+, Mg2+, Ca2+) and precursor gases (NH3, SO2, HNO3, and HCl). Meteorological parameters consisting of air temperature (T), relative humidity (RH), wind speed (WS), and wind direction (WD) were measured by a Vaisala weather transmitter (WXT520). The mixing ratios of NO, NO2, and NOX (NO + NO2) were determined using a NOX analyzer (Thermo Scientific, Model 42i) with the detection limit of 1 ppb. The chemiluminescence techniques would overestimate NO2 concentrations by the interference of NOy, so we corrected NO2 concentrations as shown in Supplementary Note 1 and used it for discussion.

Another observation field was the Institute of Atmospheric Physics, Chinese Academy of Sciences (IAP, 39°58′54.91″N, 116°23′4.79″E), about 5 km southeast of the ICCAS site, which had similar surroundings of condensed population and heavy traffic. Aerosol surface area (Saero) was measured by Scanning Mobility Particle Sizer (SMPS) with particle number size distributions from 14.5 nm to 710.5 nm. And photolysis frequencies (jO1D, jHONO, and jNO2) were measured by photolysis spectrometers (PFS-100, Focused Photonics Inc.). The detailed information on Saero and photolysis frequencies was described in previous studies93,94. Boundary layer height (BLH) was measured by a single-lens ceilometer, Vaisala CL51, Finland. Volatile organic compounds (VOCs) concentrations were gained from an automatic gas chromatograph system with a mass spectrometer and a flame ionization detector (GC-MS/FID)95.

PM2.5, CO, and O3 were data from Wanliu monitoring station (got from the Beijing Municipal Biology Environment Bureau station), about 3 km southwest of the ICCAS measurement site. This site had similar surroundings to the above two sites.

Model description

Aerosol liquid water content (ALWC) and aerosol pH were derived by ISORROPIA-2 thermodynamic equilibrium model. Inorganic ions in PM2.5 and gaseous precursors measured by the online ion chromatography system MARGA were constrained in the model. In the model, the forward mode was used for aerosols in the metastable conditions, and only data when 40% ≤RH < 95% were analyzed59. Comparison between observed and simulated inorganic ions in PM2.5 and gaseous precursors showed consistency, which suggested the accuracy of model simulation59,66. Previous studies used the same method (ISORROPIA-2 model) to calculate the aerosol pH value during polluted days, and the calculated aerosol pH value (4.0–5.0)66,71,96,97 was similar to ours.

HONO budget was explored by box model with MCM (master chemical mechanism version 3.3.1) mechanism. Box model is a zero-dimensional model, which is constrained by the profile of observed meteorological parameters and atmospheric pollutants concentrations. MCM is the most representative specific chemical mechanism. Each model species corresponds to an atmospheric compound and describes a series of chemical reactions in the atmosphere in detail. The summary of the constrained species in box model in this study is shown in Supplementary Table 2. The simulated interval was 1 h. Here, we only concentrated on pollution episodes with stagnant meteorological conditions. The most common source of HONO in model simulations was OH and NO homogeneous reaction. To improve the understanding of HONO, some HONO formation mechanisms were added to the simulation as exhibited in Table 1.