Introduction

Atmospheric nitrous acid (HONO) can be quickly photolyzed to produce hydroxyl (OH) radicals under weak sunlight irradiation in the early morning, and thus it has long been recognized as the initiator for triggering daytime atmospheric chemistry in the lower atmosphere1. A large number of studies further reveal that daytime HONO over polluted areas also acts as the major or even dominant source for OH radicals, accounting for 20–90% of the total OH primary production2,3. However, the sources of HONO were not fully understood, e.g., there exist a large missing HONO source in daytime based on the field measurements4,5,6. To explain the missing source of daytime HONO, several potential HONO sources have been proposed, including the photo-enhanced reduction of NO2 by organic materials7,8,9, the photolysis of adsorbed nitric acid (HNO3) or particulate nitrate10,11,12,13, soil emission through biogenic production14,15, displacement of strong acids on soil surface16,17, the reaction of excited gaseous NO2 with water18,19, the photolysis of nitrophenols20,21, and the reaction of HO2·H2O complexes with NO222. Among the proposed HONO sources, the photo-enhanced reduction of NO2 by organic materials on ground surface is commonly considered as the dominant HONO source in daytime2,23. Nevertheless, there are still large gaps in daytime HONO levels between the observations and model simulations after the proposed HONO sources are taken into consideration4,24,25.

Considering that NOx (NO and NO2) is the predominant precursor for HONO formation26,27, the unknown HONO sources might be related to some additional NOx reactions. Therefore, it is necessary to further investigate the possible HONO formation from the atmospheric oxidation of NOx. In this study, a series of experiments were conducted in a smog chamber to reveal the possible HONO formation channels from photo- and dark-oxidation of NOx. Finally, the contributions of the potentially proposed mechanisms to atmospheric HONO and O3 were also assessed by the Weather Research and Forecasting coupled with chemistry (WRF-Chem) model.

Results and discussion

HONO formation from photooxidation of NOx

The variations of HONO and the key species (NO, NO2 and O3) in six gas mixtures of NOx under both dark and light conditions are illustrated in Fig. 1 and Supplementary Fig. 1, respectively. Before the irradiation, HONO concentration in each mixture was relatively stable within the one-hour duration (Fig. 1), implying the heterogeneous reactions of NOx on the inherent chamber wall made ignorable contribution to HONO production. The relatively high initial HONO concentrations (0.7–1.7 ppb) in the chamber were mainly ascribed to HONO formation from heterogeneous reactions of the standard gases (NO2 and NO) in the cylinders, e.g., HONO concentration in the chamber could be reduced by 85.6% when the same amount of NO2 standard gas was introduced into the chamber through a tube filled with and without particulate sodium hydrate. In contrast to the relatively stable HONO concentration before the irradiation, HONO levels displayed evidently increasing trends after stopping the irradiation (Fig. 1), indicating that the photochemical products formed during the irradiation were involved in the dark HONO formation. During the irradiation, HONO levels in all the mixtures initially displayed quickly decreasing trends due to HONO photolysis. Subsequently, HONO levels became relatively stable in the mixtures of NOx + air and NO2 + O2, whereas they exhibited remarkably increasing trend in the mixture of NO2 + N2 after a period of irradiation (Fig. 1a). Additionally, there was significant difference in HONO levels between the two air mixtures of 300 ppb NO2 + 200 ppb NO with and without the existence of H2O during the irradiation although their initial HONO concentrations were almost identical (Fig. 1b).

Fig. 1: The variations of HONO concentration among the six mixtures under both dark and light conditions in the smog chamber.
figure 1

a 500 ppb NO2 + 207 ppm H2O + air/O2/N2, b 300 ppb NO2 + 200 ppb NO +207 ppm H2O + air, 300 ppb NO2 + 200 ppb NO + air, 300 ppb NO2 + 207 ppm H2O + air. The yellow shadows represent the UV-light irradiation.

Full size image

To explore the reasons for the distinct difference in HONO variations among the irradiated mixtures, HONO loss and formation rates in each mixture were analyzed based on the key reactions (Eqs. 1–3) that are often considered as the basis for judging the daytime unknown HONO sources in many areas6. JHONO was calculated to be about 1.0 × 10–3 s−1 according to the first-order decay law of HONO in the chamber (See Supplementary Methods and Supplementary Fig. 2), and the levels of OH radicals were estimated based on the assumption that OH radicals in the irradiated mixtures were in a steady state through Eqs. 1–428,29 (See Supplementary Methods and Supplementary Fig. 3).

$$\begin{array}{*{20}{c}} {HONO + hv \to OH + NO} & {{{{\mathrm{J}}}}_{{{{\mathrm{HONO}}}}}} \end{array}$$
(1)
$$\begin{array}{*{20}{c}} {NO + OH \to HONO} & {{{{\mathrm{k}}}}_2 = 3.3 \times 10^{ - 11}{{{\mathrm{cm}}}}^3\,{{{\mathrm{molecule}}}}^{ - 1}{{{\mathrm{s}}}}^{ - 1}} \end{array}$$
(2)
$$\begin{array}{*{20}{c}} {HONO + OH \to NO_2 + H_2O} & {{{{\mathrm{k}}}}_3 = 6.0 \times 10^{ - 12}{{{\mathrm{cm}}}}^3\,{{{\mathrm{molecule}}}}^{ - 1}{{{\mathrm{s}}}}^{ - 1}} \end{array}$$
(3)
$$\begin{array}{*{20}{c}} {NO_2 + OH \to HNO_3} & {{{{\mathrm{k}}}}_4 = 2.6 \times 10^{ - 11}{{{\mathrm{cm}}}}^3\,{{{\mathrm{molecule}}}}^{ - 1}{{{\mathrm{s}}}}^{ - 1}} \end{array}$$
(4)

As shown in Fig. 2, HONO loss rates via Eqs. 1 and 3 were evidently faster than its formation rate through Eq. 2 in all the mixtures during the irradiation. Obviously, besides Eq. 2, there existed unknown HONO sources to account for the nondecreasing HONO levels in the mixtures of NOx + air, NO2 + O2 and NO2 + N2 during the irradiation (Fig. 1a). The unknown HONO formation rates were calculated according to the following equation (Eq. 5), which showed distinct differences among the irradiated mixtures (Fig. 2).

$$HONO_{unknown} = HONO_{loss} - HONO_{formation} - \frac{{d\left[ {HONO} \right]_{meas}}}{{dt}}$$
(5)
Fig. 2: The formation and loss rates of HONO in the six mixtures of NOx during the irradiation periods.
figure 2

a 500 ppb NO2 + 207 ppm H2O +air, b 500 ppb NO2 + 207 ppm H2O + O2, c 500 ppb NO2 + 207 ppm H2O + N2, d 300 ppb NO2 + 200 ppb NO + 207 ppm H2O + air, e 300 ppb NO2 + 200 ppb NO + air, f 300 ppb NO2 +207 ppm H2O + air. Formation rate: k2[OH][NO] is the HONO formation rate through gas-phase reaction of NO with OH radicals (Eq. 2); loss rates: JHONO[HONO] is the photolysis rate of HONO (Eq. 1) and k3 [HONO][OH] is the HONO consumption rate by reaction with OH radicals (Eq. 3).

Full size image

Additionally, the higher ratio of NO to NO2 during the irradiation (Supplementary Fig. 1) significantly increased HONO formation through the reaction of NO with OH radicals, but evidently suppressed the unknown HONO formation rates (Fig. 2).

As the smog chamber is made of inert Teflon film without conducting any experiments in the presence of organic compounds, the photo-enhanced reduction of NO2 by organic materials7,8,9 could be excluded for the unknown HONO sources in the irradiated mixtures. Although the heterogeneous NO2 conversion on the Teflon surface in the presence of H2O could also contribute to HONO30,31, its reaction rate was too slow to explain the unknown HONO sources due to no photo-enhancement in the kinetics of the reaction32,33. The possible HONO formation through reaction of HO2·H2O complexes with NO222 could be also excluded because HO2 formation rates through reaction of OH with O3 (See Supplementary Methods and Supplementary Fig. 4) were at least 2 orders of magnitude slower than the unknown HONO formation rates in the irradiated mixtures (Fig. 2). The quantum yield for ground state oxygen atom (O3P) and NO from photolysis of NO2 is unity under <400 nm ultraviolet irradiation34, and thus HONO formation through the reaction of excited NO2 with water18,19 might be negligible in the irradiated mixtures because the black-light lamps used for the irradiation mainly emit 330–400 nm ultraviolet with the central wavelength of 365 nm (Supplementary Fig. 5). Considering the evident increase of HNO3 levels formed through Eq. 4 in the mixtures of NOx during the irradiation (Supplementary Fig. 6), the unknown HONO formation in the irradiated mixtures was suspected to be related to the reactions involving in HNO3. The above speculation could be supported by the significantly linear correlation (R2 = 0.99) between the average unknown HONO formation rates and the HNO3 formation rates for the five mixtures with the same H2O concentration (~207 ppm) (Fig. 3). It should be noted that the data point for the mixture of NO2 + NO without the existence of H2O was found to evidently deviate from the linear correlation (Fig. 3), implying that H2O is also involved in HONO formation through the reactions associated with HNO3. In addition, less HNO3 formation could be expected for the irradiated mixtures with increasing the NO/NO2 ratio (Supplementary Fig. 1) due to the competition reactions of NO and NO2 with OH radicals, resulting in evident decrease of the unknown HONO formation rates (Fig. 2).

Fig. 3: The correlation between the average unknown HONO and HNO3 formation rates.
figure 3

Note that: 1. 500 ppb NO2 + 207 ppm H2O + N2; 2. 300 ppb NO2 + 200 ppb NO + air; 3. 300 ppb NO2 + 200 ppb NO + 207 ppm H2O + air; 4. 300 ppb NO2 + 207 ppm H2O + air; 5. 500 ppb NO2 + 207 ppm H2O + air; 6. 500 ppb NO2 + 207 ppm H2O + O2.

Full size image

Based on the above photooxidation experiments in the presence of H2O, the unknown HONO formation rate (ppb s−1) associated with HNO3 formation rate (ppb s−1) (Fig. 3) could be overall expressed as:

$$\frac{{d[HONO]}}{{dt}} = 0.53 \times \frac{{d[HNO_3]}}{{dt}} + 1.2 \times 10^{ - 4}\,(ppb\,s^{ - 1})$$
(6)

It should be noted that the interception of 1.2 × 10–4 in Fig. 3 was the average unknown HONO formation rate (ppb s−1) when the average HNO3 formation rate was zero, indicating that there still existed the unknown HONO sources other than the reactions related to HNO3, e.g., the potential contribution from the gas-phase reaction of NO with OH radicals produced from blank Teflon chambers under ultraviolet irradiation35 to HONO.

Although photolysis of adsorbed HNO3 (HNO3(ad)) (Eq. 7) has been proposed to be an important source for atmospheric HONO, it might be negligible in the irradiated mixtures because it usually occurs under the irradiation with wavelengths less than 320 nm11. Considering the relatively fast increase of HONO levels for the irradiated mixtures with relatively high NO concentrations just after turning off the UV lamps (Fig. 1), the possible reaction of NO with HNO3 was suspected to contribute to HONO formation. Therefore, the possible HONO formation from the reaction of NO with HNO3(ad) was further investigated in the chamber by using long path absorption photometer (LOPAP) for HONO measurement under dark condition. As shown in Fig. 4a, b, the HONO concentration quickly increased from zero to about 50 ppt after adding 100 ppb NO into the chamber containing the purified air, whereas HONO concentration kept zero when the same amount of NO was introduced into the chamber containing the air mixture in the presence of 6 ppm HNO3 and 200 ppm H2O. The high HNO3 concentration could greatly reduce HONO absorption efficiency in the stripping coil of the LOPAP due to decrease of pH in the absorption solution, which might mask the possible HONO formation in the chamber. As expected, HONO concentration sharply increased from zero to more than 3 ppb when the same amount of NO was introduced into the chamber that has been exposed to 6 ppm HNO3 and 200 ppm H2O for one week and then cleaned with the purified air for four times (Fig. 4c, the green data points). The HONO concentration could still increase from zero to ~500 ppt for the second introduction of NO after the chamber being cleaned again (Fig. 4c, the blue data points), which was at least a factor of 5 higher than that of the chamber experiment with NO addition before introducing HNO3 (Fig. 4a). Because the gas phase reaction of NO with HNO3 is extremely slow36 and a certain amount of HNO3 might be absorbed on the chamber wall after the exposure of the high HNO3 concentration due to its viscosity, the HONO formation might be from the heterogeneous reaction of NO with HNO3(ad) (Eq. 8) on the chamber wall.

$$HNO_{3(ad)} + hv\,({\uplambda} \,< \,320nm) \to HONO$$
(7)
$${{{\mathrm{NO}}}} + HNO_{3(ad)}\mathop { \to }\limits^{Surface} HONO + NO_2$$
(8)
Fig. 4: The variations of HONO concentration for the mixtures of NO and HNO3 under dark condition in the smog chamber.
figure 4

a 100 ppb NO + air, b 6 ppm HNO3 + 200 ppm H2O + 100 ppb NO + air; c <6 ppm HNO3(ad) + <200 ppm H2O + 200 ppb NO + air. The black arrows indicate the time for introducing the reactants into the chamber.

Full size image

Based on the quick increase of HONO concentration just after addition of NO for the chamber experiments, the heterogeneous reaction of NO with HNO3(ad) might be very fast. The HNO3(ad) on the chamber wall might be in multiple layers and only the HNO3(ad) at the uppermost layer could be involved in HONO formation, resulting in the stable HONO level after its pulse increase owing to the quick termination of the heterogeneous reaction with depleting the HNO3(ad) at the uppermost layer. The HNO3(ad) at the inner layers could be exposed again to the surface layer after cleaning the chamber by the purified air, explaining the significant increase of HONO level for the second experiment with NO addition. As for the irradiated gas mixtures of NOx, the continuous HNO3 formation through the photochemical reaction of NO2 with OH could guarantee that the fresh HNO3(ad) on the wall surface was always exposed to NO, resulting in the high HONO yield (~53%) from the HNO3 conversion through the heterogeneous reaction (Eq. 8).

HONO formation through the heterogeneous reaction of NO with HNO3(ad) on particles has also been proposed to account for significant overestimation of [HNO3]/[NOx] ratio by photochemical models in comparison with measurements performed in the free troposphere37 and over the boundary layer of polluted urban atmospheres38. Quick increase of HONO levels was also observed after exposing NO to the HNO3(ad) on soot by a previous study in a low-pressure flow reactor (Knudsen cell)39. Based on a small quantity of HNO3(ad) consumption in the presence of NO, the heterogeneous reaction (Eq. 8) was considered as a slow reaction by the above study39. However, the heterogeneous reaction rate might be largely underestimated because of the multiple layers of HNO3(ad) on the soot.

HONO formation from dark-oxidation of NOx

The dark-oxidation of NOx by O3 can also produce HNO3 through Eqs. 9–1229, and thus HONO formation is also expected to occur during the dark-oxidation process. To verify the conjecture, a series of chamber experiments under dark condition were conducted by continuously introducing the NO standard gas (403 ppm in N2) at a fixed flow rate of 20 mL min−1 (2.7 ppb min−1 equally) into air mixtures with different initial O3 concentrations (0, ~100, ~220, and ~280 ppb). As shown in Fig. 5a, HONO levels were almost identical in the air mixtures with initial O3 concentrations of 0 ppb and ~100 ppb during the entire period, whereas they significantly increased in the air mixtures with initial O3 concentrations of ~220 ppb and ~280 ppb after introducing NO.

$$\begin{array}{*{20}{c}} {O_3 + NO \to NO_2 + O_2} & {{{{\mathrm{k}}}}_6 = 1.8 \times 10^{ - 14}{{{\mathrm{cm}}}}^3\,{{{\mathrm{molecule}}}}^{ - 1}{{{\mathrm{s}}}}^{ - 1}} \end{array}$$
(9)
$$\begin{array}{*{20}{c}} {O_3 + NO_2 \to NO_3 + O_2} & {{{{\mathrm{k}}}}_7 = 3.5 \times 10^{ - 17}{{{\mathrm{cm}}}}^3\,{{{\mathrm{molecule}}}}^{ - 1}{{{\mathrm{s}}}}^{ - 1}} \end{array}$$
(10)
$$\begin{array}{*{20}{c}} {NO_3 + NO_2 \to N_2O_5} & {{{{\mathrm{k}}}}_8 = 1.9 \times 10^{ - 12}{{{\mathrm{cm}}}}^3\,{{{\mathrm{molecule}}}}^{ - 1}{{{\mathrm{s}}}}^{ - 1}} \end{array}$$
(11)
$$N_2O_5 + H_2O_{(ad)} \to HNO_3$$
(12)
$$\begin{array}{*{20}{c}} {N_2O_5 \to NO_3 + NO_2} & {{{{\mathrm{k}}}}_9 = 0.07{{{\mathrm{s}}}}^{ - 1}} \end{array}$$
(13)
$$\begin{array}{*{20}{c}} {NO_3 + NO\mathop { \to }\limits^{M(N_{2,}O_2)} NO_2} & {{{{\mathrm{k}}}}_{10} = 2.6 \times 10^{ - 11}{{{\mathrm{cm}}}}^3\,{{{\mathrm{molecule}}}}^{ - 1}{{{\mathrm{s}}}}^{ - 1}} \end{array}$$
(14)
Fig. 5: HONO formation from the reaction of NO3 with NO under the dark condition.
figure 5

a, b The HONO and N2O5 concentrations for the mixtures with different initial O3 concentrations and ~207 ppm H2O, respectively; c the HONO and N2O5 concentrations for the mixture with the initial ~260 ppb N2O5 concentration and ~828 ppm H2O (See Supplementary Methods); note that the NO standard gas was added continuously into the chamber with the flow rate of 20 mL/min from 60 to 180 min for (a), (b) and (c); d pathways for the reaction of NO3 with H2O to produce HONO with and without the existence of NO as well as the optimized geometries of the reactant complex (RC), transition state (TS) and product complex (PC). Blue, red and white circles represent N, O and H atoms, respectively.

Full size image

Although dark reactions (Eqs. 9–14) can occur in all the air mixtures, the reaction extent largely depends on the initial O3 concentrations. For the air mixture with initial O3 concentration of ~100 ppb, O3 was mainly consumed by reaction with NO (Eq. 9), which suppressed NO3 formation from the reaction of O3 with NO2 (Supplementary Figs. 7 and 8), resulting in negligible N2O5 accumulation (Fig. 5b). As for the air mixtures with higher initial O3 concentrations, the reaction of O3 with NO2 (Eq. 10) was significantly accelerated because of their relatively high concentrations during introduction of NO (Supplementary Fig. 7c and d), leading to remarkable formation of NO3 (Supplementary Fig. 8) and evident accumulation of N2O5 (Fig. 5b). The quick decrease of N2O5 levels after their peak values in the two mixtures with high initial O3 concentrations were attributed to the reaction of NO with NO3 which suppressed N2O5 formation rate through Eq. 11 and accelerated N2O5 consumption rate through Eq. 13. More HNO3 could be formed with increasing N2O5 levels in the chamber because of the heterogeneous reaction of N2O5 with the absorbed H2O on the chamber wall (Eq. 12). The evidently faster increase of HONO levels with increasing the initial O3 concentrations (Fig. 5a) further indicated that HONO formation was possibly associated with the forementioned heterogeneous reaction involving in HNO3. Compared with the period after depletion of N2O5, the increase of HONO level was evidently faster during the period when obvious N2O5 was present (Fig. 5a, b), implying that other reactions besides Eq. 8 might also make contribution to HONO formation.

The reaction of NO3 with NO was suspected to produce HONO because NO3 concentration was relatively high during the period in the presence of N2O5 (Supplementary Fig. 7). To certify the suspicion, dark experiments were further conducted by introducing NO into air mixtures of N2O5 (See Supplementary Methods). The variations of HONO and N2O5 levels in the air mixture are shown in Fig. 5c and those of NO, NO2 and O3 concentrations are illustrated in Supplementary Fig. 9. As shown in Fig. 5c, after introducing NO into the chamber, the abrupt increase of HONO level was well accompanied with quick decrease of N2O5 concentration. Additionally, the increase of HONO level displayed a turning point when N2O5 concentration approached zero. Due to small O3 concentration in the air mixture of N2O5 before introducing NO (Supplementary Fig. 9), the introduced NO mainly reacted with NO3 leading to fast decrease of N2O5 and quick increase of NO2 through Eqs. 13–14. Based on the significantly linear correlation (R2 = 0.997) between the increment of HONO and the decrease of N2O5 with the slope of 0.02 (Supplementary Fig. 10), HONO yield of ~2% could be obtained due to NO3 consumption.

To account for the above phenomenon, the reaction (Eq. 15) was proposed.

$$NO_3 + NO + H_2O \to HNO_3 + HONO$$
(15)

The proposed reaction (Eq. 15) was verified to be feasible through Density Functional Theory (DFT) calculations (See Supplementary Methods), which is barrierless in comparison with the extremely higher energy barrier for the reaction of NO3 with H2O (Fig. 5d).

Atmospheric implication

Atmospheric NOx could also experience the same oxidation processes as the cases of the simulation in the chamber, and hence the unknown HONO sources identified through the chamber experiments are also expected to occur in the real atmosphere. To assess the contribution of atmospheric NOx oxidation to HONO, atmospheric HONO levels at two sampling sites of a rural area (Station of Rural Environment, Research Center for Eco-Environmental Sciences, SRE-RCEES40,41,42) and Beijing city (Institute of Atmospheric Physics, IAP43) in the NCP were simulated by the WRF-Chem model with incorporation of the two HONO formation pathways identified in this study (See Supplementary Methods).

For the scenarios BASE (only considering the gas-phase HONO formation) and E (BASE plus direct HONO emissions), the simulated HONO levels were about one order of magnitude lower than the observed values (Fig. 6), which was in agreement with previous modeling results44,45. The simulation was significantly improved after further considering the major highlighted HONO sources for the scenario EPLA (E plus the photolysis of particulate nitrate, and the (photo-enhanced) heterogeneous reactions of NO2 on land and aerosol surfaces), but the simulated daytime HONO levels were only about half of the observed values. The scenario ELAN1 (E plus the heterogeneous reactions of NO2 on land and aerosol surfaces, and the NOx photooxidation in daytime) could well reproduce the daytime HONO levels even if the photo-enhanced heterogeneous reactions of NO2 and the photolysis of particulate nitrate were not taken into consideration, implying the significant role of the NOx photooxidation in HONO formation over the polluted areas. Additionally, the HONO formation from the reaction of NO3 with NO also made great contribution to nighttime HONO, as shown in Fig. 6 for the scenario ELAN1N2 (ELAN1 plus NO oxidation by NO3 in nighttime).

Fig. 6: Time series and average diurnal variation of simulated HONO from different scenarios versus the observed HONO at SRE-RCEES and IAP of the NCP.
figure 6

a, b Time series and average diurnal variation of simulated and observed HONO at SRE-RCEES, respectively; c, d time series and average diurnal variation of simulated and observed HONO at IAP, respectively. The different scenarios contain BASE, E, EPLA, ELAN1 and ELAN1N2 (see Supplementary Methods and Supplementary Table 1).

Full size image

Many field studies reported that the daytime missing HONO sources showed significant correlations with the product of NO2 photolysis rate coefficient (JNO2) and NO2 concentration (CNO2), and thus photo-enhanced reductions of NO2 on the surfaces of aerosols and ground were proposed to be the major source of daytime HONO6,46. As the diurnal variation of JNO2 is in line with OH concentration (COH) at ground level47, the daytime missing HONO sources could also be expected to significantly correlate with the product of k4 × COH × CNO2, namely HNO3 formation rates. The photo-enhanced conversion of NO2 to HONO identified in the flow tube experiments7,8 might also be partially viewed as the contribution from the forementioned heterogeneous reaction associated with HNO3 because HNO3 could be formed through reactions of NOx with HOx (OH + HO2) radicals that are recently found to be produced by irradiating atmospheric particles48,49 and organic mixtures50.

Compared with the scenario BASE, the simulated O3 levels by the scenario ELAN1N2 had noticeable enhancements (~38.4%) and were in good agreement with the observations at multiple monitoring stations across the NCP (Supplementary Fig. 11), implying the additional HONO sources over the NCP could play an important role in the regional O3 formation. Therefore, HONO formation from atmospheric oxidation of NOx should be taken into consideration for formulating O3 control strategies in polluted areas.

Methods

Smog chamber experiments

The experiments were conducted in a 3-m3 collapsible Teflon environmental chamber at 25 ± 1 °C, which was irradiated by 63 black-light lamps with a central wavelength of 365 nm (Supplementary Fig. 5). Detailed information for the chamber was described in our previous studies51,52. The buffer gas (ultrapure N2, ultrapure O2 or synthetic air with the purity of ≥ 99.999 %) was introduced into the chamber at a flow rate of 70 L/min through a mass flow controller (Beijing Sevenstar Electronic Technology Co. Ltd., China). To achieve the target concentrations for each mixture, NOx obtained from the standard gases (584 ppm NO in N2 and 591 ppm NO2 in N2) and O3 generated by an electrical discharge generator were directly introduced into the chamber with glass syringes. Additionally, water vapor was introduced into the chamber through bubbling a bubbler containing ultrapure water by 3 L/min N2. Before each experiment, the chamber was flushed at least 3 times with the ultrapure N2 to avoid possible interfering substances in the chamber.

NOx (NO and NO2) and O3 were measured by the NOx and O3 analyzers (Model 42i and 49i, Thermo-fisher Scientific Inc, USA), respectively. HONO, HNO3 and N2O5 were measured based on a wet chemical method53. In brief, they were absorbed by a stripping coil with 25 μM sodium carbonate solution. The gas and liquid flow rates were set to be 2 L min−1 and 0.25 mL min−1, respectively. After sampling, NO2− and NO3− in the absorption solution were measured by an ion chromatography54,55 (IC, WAYEAL IC6200, China). The wet chemical method for HONO and HNO3 measurement has been systematically evaluated and widely used in the previous studies11,53,56. It should be mentioned that the measured NO3− in the absorption solution represented the sum of HNO3 and N2O5 levels. Considering that N2O5 formation can be obviously suppressed when NO levels began to accumulate in the chamber, the measured NO3− in the absorption solution with the existence of excess NO should be attributed to HNO3 which had been formed from hydrolysis of N2O5 on the chamber wall. N2O5 concentrations before introduction of NO were roughly calculated by subtracting HNO3 concentration from the sum NO3- concentration in the absorption solution collected from the chamber.

WRF-Chem configuration

The WRF-Chem model (ver. 4.0.3)57 was adopted to explore the relative contributions of various HONO sources to atmospheric HONO. Carbon Bond Mechanism version-Z (CBMZ) scheme58 coupled with the 4-bin sectional Model for Simulating Aerosol Interactions and Chemistry (MOSAIC) scheme59 were used as the gas and aerosol chemical mechanisms in the model, respectively. The anthropogenic emission data were obtained from the Multi-resolution Emission Inventory for China with base year of 2017 (MEIC-2017) with a resolution of 0.25° × 0.25° (http://www.meicmodel.org/)60. The biogenic emissions were estimated online by the Model of Emissions of Gases and Aerosols from Nature version 2.1 (MEGAN v2.1)61.

The major HONO sources, including the direct emission, the heterogeneous reactions and the photo-enhanced reductions of NO2 on aerosol and ground surfaces, and the photolysis of particulate nitrate, were incorporated into the emission module and MOSAIC chemical mechanism in WRF-Chem model, respectively. The new HONO sources identified by this study were added into MOSAIC chemical mechanism and CBMZ chemical mechanism, respectively. Other WRF-Chem configurations adopted in this study were listed in Supplementary Table 2. The simulations were performed from 1st to 22nd June 2017 with 7-day spin-up period on a domain covering the North China Plain (NCP), centered at 39.5 °N, 114.8 °E, with 9 km horizontal resolution, 223 × 202 grid cells, and 30 vertical levels from the ground level of 17 m to the maximum pressure of 50 hPa. For more details, see Supplementary Methods.