Is the photochemistry activity weak during haze events ? 1 — — A novel exploration on the photoinduced heterogeneous reaction of 2 NO 2 on mineral dust 3

10 Despite the increased awareness of heterogeneous reaction on mineral dust, the knowledge of how the 11 intensity of solar irradiation influences the photochemistry activity remains a crucially important part in 12 atmospheric research. Relevant studies have not seriously discussed the photochemistry under weak sunlight 13 during haze, and thus ignored some underlying pollution and toxicity. Here, we investigated the heterogeneous 14 formation of nitrate and nitrite under various illumination conditions by laboratory experiments and field 15 observations. Observed by in-situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS), 16 water-solvated nitrate was the main surface product, followed by other species varying with illumination condition. 17 The growth of nitrate formation rate tends to be slow after the initial fast with increasing light intensity. For 18 example, the geometric uptake coefficient (γgeo) under 30.5 mW/cm (5.72×10) has exceeded the 50 % of that 19 under 160 mW/cm (1.13×10). This case can be explained by the excess NO2 adsorption under weak illumination 20 while the excess photoinduced active species under strong irradiation. Being negatively associated with nitrate 21 (R=0.748, P<0.01), nitrite acts as the intermediate and decreases with increasing light intensity via oxidation 22 pathways. Similar negative dependence appears in coarse particles collected during daytime (R=0.834, P<0.05), 23 accompanied by the positive association during nighttime (R=0.632, P<0.05), suggesting illumination a 24 substantial role in atmospheric nitrogen cycling. Overall, for the nitrate formation, the conspicuous response under 25 slight illumination offers opportunities to explain the secondary aerosol burst during haze episodes with weak 26 irradiation. Additionally, high nitrite levels accompanied by low nitrate concentrations may induce great health 27 risk which was previously neglected. Further, Monte Carlo simulation coupled with sensitivity analysis may 28 provide a new insight in the estimations of kinetics parameters for atmospheric modelling studies. 29 Atmos. Chem. Phys. Discuss., https://doi.org/10.5194/acp-2019-315 Manuscript under review for journal Atmos. Chem. Phys. Discussion started: 13 May 2019 c © Author(s) 2019. CC BY 4.0 License.

Prior to each experiment, the particles were pretreated in a stream of high-pure air (200 ml· min -1 ) for 60 min 89 to remove the adsorbed water and impurities from the surfaces ( Figure S4). Due to the overlapping bands of 90 adsorbed water (~1640 cm -1 ) and nitrogen compounds, the sample after pretreatment was exposed to humid 91 high-pure air (RH≈30%, 100 ml· min -1 ) for 20 min, after which the moisture absorption reaches saturation ( Figure  92 S5). A background spectrum was recorded after the process and then NO2 calibration gas (5.12 ml· min -1 ) was 93 added into the DRIFTS chamber with a calculated concentration of 15.33 ppm. Calibration gases with NO2 94 concentrations of 9.20 and 21.45 ppm were also involved for the concentration dependence experiments. Ten 95 light intensity levels (0.0, 0.3, 5.4, 17.5, 23.8, 30.5, 54.5, 98.5, 128.1, and 160.0 mW· cm -2 ) were referred in this 96 study. 97 Each test lasted 90 min, during which a series of spectra were recorded every 5 min. The reacted particles 98 were extracted by oscillation (5 min) with 4 ml water. The extraction solution was then passed through a 0.22 μm 99 PTFE membrane filter for ion detection. 100

Ion analysis 101
The nitrate and nitrite ions were analyzed by an ion chromatography (IC, 883 Basic, Metrohm, Switzerland), 102 which consists of an analytical column (A5-250) and a guard column. The detection was conducted by using 3.2 103 mmol· L -1 Na2CO3 and 1.0 mmol· L -1 NaHCO3 at a stable flow rate of 0.70 ml· min -1 . Multipoint calibrations were 104 performed by means of standard solutions. Good linearity of the calibration curve was obtained with R 2 >0.998. 105

Photo-electrochemical (PEC) test 106
In order to qualitatively evaluate the generation of electron-hole pairs under different light intensities, PEC wire and an Ag/AgCl electrode were employed as the counter and reference electrodes, respectively. The 111 electrolyte was 0.5 mol/L NaNO3. A xenon lamp (CEL-S500, Beijing Ceaulight Co., LTD, China) was used to 112 provide simulated sunlight. 113

Uptake coefficient estimation 114
The reactive uptake coefficient, γ, is defined as the ratio of the reactive gas-surface collision rate (d (4) 120 Where slope represents the growth rate of the nitrate peaks, f is the conversion factor, is the particle 121 reactive surface area, NO 2 is the mean velocity of NO2 molecule, [NO2] is the NO2 concentration, R is the gas 122 constant, T is the temperature, 2 is molecular weight of NO2 (Table S1). 123 The conversion factor (f) is obtained from a calibration plot with the amount of were calculated and then normalized to 100%. On this basis, the contribution of each input variable to the output 134 can be assessed. Three input variables are included for γgeo: slope, f, and As. For γBET, the As is further divided into 135 mass and SBET as discussed above. reproducible results were obtained with the coefficient of determination (R 2 ) greater than 0.990. The bands in the 153 spectra are quite rich, indicating various products as summarized in Table S3.
Illumination has impacts on either product species or the production. The final DRIFTS spectra grow in 189 intensity as the illumination becomes stronger. Raman measurements also indicate the drastic enhancement caused 190 by sunlight, evident by the higher nitrate peak after illumination compared to that after dark process (Section S9) 191   135%, 189%, 158%, 148%, 103%, 39%, and 16% higher than the corresponding theoretical ones, respectively. 215 This 'fast-slow' uptrend seems to be of great importance as it shows that the γ-values measured at designed 216 irradiation intensity may not be extrapolated in a linear way to those relevant to the atmosphere. The balance 217 between PAS formation and NO2 adsorption is responsible for the uneven illumination effect, which will be 218 carefully discussed in the mechanism section. 219 To distinguish the contributions of each variable to the output, sensitivity analysis is performed on the basis of 220 the simulated data. Slope and f contribute most to the total variance of γBET and γgeo, while SBET and m for γBET, and 221 Ageo for γgeo contribute little (Section S10). Accordingly  Figure 3 presents the association between atmospheric nitrate and nitrite varying with particle mode and 245 sampling period. Significant positive correlation can be found during nighttime in coarse mode (Figure 3a). 246

Nitrogen redox 244
However, there is no case indicating high nitrite and nitrate levels during daytime, and the dependence seems to be 247 negative (Figure 3b)  Atmospheric nitrate and nitrite from diverse periods exhibit analogous size distribution: greatest in coarse 259 mode, followed by droplet mode and condensation mode ( Figure S9). Yet, except the large mass fraction in coarse 260 mode, nitrite presents extra peak under 1.8 μm, indicating reaction pathways differing from nitrate formation 261 (Moore et al., 2004). That is, nitrate is difficult to accumulate by aqueous reactions or homogeneous processes 262 while nitrite seems to be easy, which results in the lower correlation coefficients for small size particles ( Figure  263 3d-i). Since the main reaction pathways (R.1, 3, 4) still take place in aqueous media, and some other oxidants (e.g. 264 H2O2, O3, and Fe 3+ ) would replace the promoting role of semiconductor components in mineral dust under 265 illumination (R.2, 5) (Hems et al., 2017; Hou et al., 2017; Xue et al., 2016), the correlation in droplet mode 266 appears to be obvious with merely lower coefficients. Furthermore, both ions exhibit great mass fractions (>50%) 267 in coarse mode, making the associations for full-size particles similar with those for coarse aerosols (Figure 3j-l).  ., 2015). However, the nitrate formation on mineral dust is found to be more dependent on weak 322 sunlight, indicating that photochemistry processes are still crucial in heavy haze. Since the NO2 concentrations in 323 the troposphere are much lower than the simulated levels, authentic dust may be close to achieving its highest 324 uptake capacity in the presence of faint sunlight (Figure 4b). Hence, photoinduced reaction on mineral dust may 325 contribute greatly to secondary aerosols during extreme haze events. 326 Nitrate pollution has got much concern recently, while little attention has been paid to the nitrite burst 327 accompanied by low nitrate concentration. Nitrite may induce adverse health risk for its close association with 328 various cancer cases (Zhang et al., 2018). Compared to the polluted aerosols with high nitrate level, the 329 carcinogenic aerosols with great nitrite concentration may be more harmful to human health. As an intermediate in