Characterization of nighttime formation of particulate organic 1 nitrates based on high-resolution aerosol mass spectrometry in an 2 urban atmosphere in China 3

Kuangyou Yu1,2,*, Qiao Zhu1,*, Ke Du2, Xiao-Feng Huang1 4 1Key Laboratory for Urban Habitat Environmental Science and Technology, School of Environment and Energy, Peking 5 University Shenzhen Graduate School, Shenzhen, 518055, China. 6 2Department of Mechanical and Manufacturing Engineering, University of Calgary, Calgary, Canada . 7 * Authors have equal contribution. 8 9 Abstract. Organic nitrates are important atmospheric species that significantly affect the cycling of NOx and ozone production. 10 However, characterization of particulate organic nitrates and their sources in inorganic nitrate-abundant particles in polluted 11 atmosphere is a big challenge, and has been little performed in the literature. In this study, an Aerodyne high-resolution time12 of-flight aerosol mass spectrometer (HR-ToF-AMS) was deployed at an urban site in South China from 2015 to 2016 to 13 characterize particulate organic nitrates with high time resolution. Based on two different data processing methods, 13-21% of 14 the total measured nitrates was identified to be organic nitrates in spring, 41-64% in summer and 16%-25% in autumn; however, 15 in winter, most measured nitrates were inorganic. The good correlation between organic nitrates and fresh secondary organic 16 aerosol identified by the positive matrix factorization method at night rather than in the daytime indicated a potentially 17 important role of nighttime secondary formation. Therefore, we theoretically estimated nighttime NO3 radical concentrations 18 and SOA formation using the various VOCs measured simultaneously. Consequently, the calculated products of monoterpene 19 reacting with NO3 agreed well with the organic nitrates in terms of both concentration and variation, suggesting that the 20 biogenic VOC reactions with NO3 at night are the dominant formation pathway for particulate organic nitrates in polluted 21 atmosphere, despite of much higher abundance of anthropogenic VOCs. 22 23

peroxy radicals and NO may play a larger role than previously recognize unsaturated hydrocarbons (R4).Even though some recent studies have suggested that the formation of organic nitrates from peroxy radicals and NO may play a larger role than previously recognized, yields of organic nitrates via NO3 reacting with alkenes are generally much higher (Teng et al., 2017(Teng et al., , 2015)).
RH + OH +  2 →  2 +  2  ( 1 ) R 2 + NO →  2 ( 2 ) R 2 + NO →  +  2 ( 3 ) R =  ′ + N 3 → ( 2 )  ′ ( 4 ) Several direct methods have been developed to measure total organic nitrates (gas + particle) in the real atmosphere.For example, Rollins et al.(2012) used a thermal-dissociation laser-induced fluorescence technique (TD-LIF) to observe total organic nitrates in the United States; Sobanski et al. (2017) obtained organic nitrates in Germany using the thermal dissociation cavity ring-down spectroscopy (TD-CRDS).However, it is still difficult to identify and quantify the particle phase organic nitrates, which could contribute a large portion of secondary organic aerosol (SOA) (Rollins et al., 2012;Xu et al., 2015a;Fry et al., 2013;Ayres et al., 2015;Boyd et al., 2015;Lee et al., 2016), using these direct measurement methods.Recently, researchers have proposed some estimation methods for particle-phase organic nitrates based on aerosol mass spectrometry (AMS) with high time resolution (Farmer et al., 2010;Hao et al., 2014;Xu et al., 2015aXu et al., , 2015b)).Ng et al. (2017) reviewed the nitrate radical chemistry and the abundance of particulate organic nitrates at multiple sites of the world, but all of these sites are located in the region (i.e.United States and Europe) with relatively clean air.To our best knowledge, few studies have investigated the concentrations and formation pathways of particulate organic nitrates in the atmosphere with high anthropogenic pollutants (Xu et al., 2017), especially with high particulate inorganic nitrate, which would make identification of organic nitrates harder in aerosol mass spectrum.
South China is located in a subtropical region, where the photochemical reactions are extremely active (Zhang et al., 2008), and the biogenic VOCs and anthropogenic NOx are relatively high.To assess the evolution of particle-phase organic nitrates in a more polluted atmosphere, in this study, we deployed Aerodyne high-resolution time-of-flight aerosol mass spectrometry (HR-ToF-AMS) with other instruments over an urban site in South China from 2015 to 2016 to obtain submicron aerosols.
Then, organic nitrates and their contributions to OAs in different seasons were estimated by different methods based on the HR-ToF-AMS measurements.Furthermore, we used the estimates combined with the measured VOC data to investigate the potential pathway formation for organic nitrates in South China.

Sampling site and period
The sampling site (22.6°N,113.9°E; 20 m a.s.l) was on the roof of one academic building on the campus of the Peking University Shenzhen Graduate School (PKUSZ), which is located in the western urban area in Shenzhen (Figure 1).This site is mostly surrounded by subtropical plants without significant anthropogenic emission sources nearby, except for a local road that is ~100 m from the site.In this study, we use the statistical data from the Meteorological Bureau of Shenzhen Municipality (http://www.szmb.gov.cn/site/szmb/Esztq/index.html) as the reference data to determine the sampling periods to obtain more representative samples in different seasons during 2015-2016, as shown in Table 1.During the sampling periods, the chemical compositions and mass concentration of non-refractory PM1 were measured by an HR-ToF-AMS, and detailed descriptions of this instrument are given in the literatures (DeCarlo et al., 2006;Canagaratna et al., 2007).In summary, the HR-ToF-AMS focuses ambient particles with vacuum aerodynamic diameter smaller than 1 m into narrow beam via an aerodynamic lens, then the submicron particles are vaporized by impaction on a tungsten heated suface (~600 ˚C) and ionized by electron ionization (70eV).Only non-refractory species can be vaporized and detected.The setup and operation of the HR-ToF-AMS can be found in our previous studies (Huang et al., 2010(Huang et al., , 2012;;Zhu et al.,2016).To remove coarse particles, a PM2.5 cyclone inlet was placed on the roof of the building to introduce an air stream containing the remaining particles into a room through a copper tube with a flow rate of 10 l min −1 .Before entering the AMS, the samples are dried by a nafion dryer (MD-070-12S-4, Perma Pure Inc.) to eliminate the potential influence of relative humidity on the particle collection (Matthew et al., 2008).The ionization efficiency (IE) calibrations wer perfomed by using pure ammoniu m nitrate particles on every two weeks .The relative IEs (RIEs) for organics, nitrate and chloride were assumed to be 1.4,1.1 and 1.3, respectively.A composition-dependent collection efficiency (CE) was applied to the data based on the method of Middlebrook et al. (2012).The instrument was operated at two ion optical modes with a cycle of 4 min, including 2 min for the mass-sensitive V-mode and 2 min for the high mass resolution W -mode.The HR-ToF-AMS data analysis was performed using the software SQUIRREL (version 1.57) and PIKA (version 1.16) written in Igor Pro 6.37 (Wave Metrics Inc.)(http://cires1.colorado.edu/jimenezgroup/ToFAMSResources/ToFSoftware/ index.html).

Other co-located instruments
In addition to the HR-ToF-AMS, a suite of instruments was deployed in the same sampling site.An aethalometer (AE-31, Magee) was simultaneously used for measurements of refractory black carbon (BC) with a temporal resolution of 5 min.VOCs concentrations were measured via an automated in-situ gas chromatograph (Agilent 5977E) equipped with a mass spectrometer (Agilent 5971).Ozone and NOX was measured by a 49i ozone analyzer and a 42i nitrogen oxide analyzer (Thermo Scientific, US), respectively.

Organic nitrate estimation
In this study, we use two methods to estimate the organic nitrates based on AMS organic data, following the same a nalysis approach in Xu et al. (2015b).The first method for estimating organic nitrates is based on the NO + /NO2 + ratio (NOX + ratio) in the HR-AMS spectrum.Due to their very different NOX + ratios (RON and RNH4NO3) (Farmer et al., 2010;Boyd et al., 2015;Fry et al., 2008;Bruns et al., 2010), the NO2 and NO concentrations for the organic nitrates ( 2, and   ) can be quantified with the HR-AMS data via Eqs.( 1) and ( 2), respectively (Farmer et al., 2010): where Robs is the NOX + ratio from the observation.The value of RON is difficult to determine because it varies between instruments and precursor volatile organic compounds (VOCs).However, RON/RNH4NO3 has been assumed instrumentindepennt (Fry et al., 2013).In this study, we use RON/RNH4NO3 estimation range (from 2.08 to 3.99) from the literature (Farmer et al., 2010;Boyd et al., 2015;Bruns et al., 2010;Sato et al., 2010) to determine RON due to the variation of precursor VOCs .
It is important to note that if a large percentage of organic nitrates value are negative using this method, it is because the values of Robs is smaller than RNH4NO3, further indicating the inorganic nitrates contributes near all to the total nitrates.
The second method is based on the positive matrix factorization (PMF) analysis.In addition to the PMF of the organic mass spectra (Zhang et al., 2011;Ng et al., 2010;Huang et al., 2013), the same analysis of the HR organic mass spectra, combined with NO + and NO2 + , ions was performed to identify the relative contributions of organ ic and inorganic nitrates (Hao et al., 2014;Xu et al., 2015b).In this study, the detailed PMF analysis procedure can be found in our previous publications (Huang et al., 2010;Zhu et al., 2016;He et al., 2011).For each season, three organic factors and one inorganic factor are identified: a hydrocarbon-like OA (HOA), a more-oxidized oxygenated OA (MO-OOA), a less-oxidized oxygenated OA (LO-OOA) and a nitrate inorganic aerosol (NIA).In this method, the NO + and NO2 + ions are distributed among different organic aerosol factors and NIAs; the concentration of the nitrate functionality in the organic nitrates (NO3,org) is equal to the sum of NO + and NO2 + in the organic nitrates (i.e., NO + org and NO + 2,org) via Eqs.( 3) and (4), respectively (Xu et al., 2015b): where [OA factor]i represents the mass concentration of the ith OA factor and   and  2, represent the mass fractions of NO + and NO2 + , respectively.

Nitrate radical estimation
In this section, the approach of nitrates radical estimation is similar to Xu et al. (2015a).The average concentration of VOCs and the reaction rate coefficients of NO3 + VOCs at 25 ˚C are listed in Table S1.NO3 is the product of NO2+O3, and its losses where  ℎ represents the rate of heterogeneous uptake,  is the uptake coefficient, ν represents the molecular speed, and SA represents the surface area of the particles.By using the upper-limit values of  = 0.04 (Saunders et al., 2003), ν = 240 m 2 cm -3 1 and SA = 220 m 2 cm -3 , we calculate  25 ,ℎ to be approximately 1760 s.In addition, the N2O5 lifetime, with respect to the reaction with H2O, is (Crowley et al., 2011): 2 represents the reaction rate of N2O5 and H2O, and [  2  ] represents the water concentration (unit of molecule cm -3 ); the daily maximu m [  2  ] is 5.5*10 11 molecule cm -3 at 6:00 during the sampling period, and the calculated value is 1.4*10 10 s.
Then, we estimate the NO3 lifetime by only considering the reaction with VOCs ( 3, ): The average lifetime of NO3 is approximately 14.08 s.Based on the estimation of the N2O5 and NO3 lifetimes above, we can conclude that the influence of N2O5 could be ignored when estimating the NO3 concentration and, due to the high reactivity of NO3 (14.08 s), the steady-state NO3 can be predicted: where  3 is calculated from the solar zenith angles and NO3 photolysis rates (Saunders et al., 2003) and, in this study, the typical value of  3 is 0.12 s -1 during the daytime. 1 is 3.52*10 -17 cm 3 molecule -1 s -1 , and  2 is 2.7*10 -11 cm 3 molecule -1 s -1 according to the Master Chemical Mechanism model (http://mcm.leeds.ac.uk/MCM/; under 25 ˚C).The average concentrations of O3 and NO2 are 6.82 and 19.38 ppb, respectively.

Results of the organic nitrate estimation
Table 2 shows the concentrations of nitrate functionality in organic nitrates (i.e., NO3.org) and their contributions to the total measured nitrate, which is estimated by the NO + /NO2 + ratio method and PMF method.Note that the small difference between the average Robs and RNH4NO3 in winter leads to a large portion of negative data (Table 1), which suggests that a very limited amount of organic nitrates contribute to the total nitrate, as discussed above; thus, we only discuss the organic nitrates in spring, summer and autumn.For the NO + /NO2 + ratio method, two calculated RON values for each season based on the RON/RNH4NO3 estimation range (from 2.08 to 3.99) are applied to provide the upper and lower bounds of the estimated NO3.org mass concentration.For the PMF method, the NOX + ions were assigned to different PMF factors (Figure S1) due to the different physicochemical properties of the nitrate components.The NIAs are dominated by NO + and NO2 + but also contain some organic fragments, such as CO2 + and C2H3O + , which agrees with the literature (Hao et al., 2014;Xu et al., 2015b;Sun et al., 2012), This indicates that the NIA factor experiences potential interference from organics .In addition, the NO + /NO2 + ratio in NIAs is higher than that in pure NH4NO3, which supports the underestimation of NO3.org concentrations with this method.This also explains why the concentration of NO3.org estimated using the PMF method is always close to the lower estimation of NO3.org via the NO + /NO2 + ratio method in Table 2. To verify the reliability of the estimated results, each NO3.org concentration time series calculated by these two methods is shown in Figure 2, and the correlation coefficient (R) for each season is adequate (0.82 for spring, 0.82 for summer and 0.77 for autumn), indicating that similar results are achieved by using the NO + /NO2 + ratio method and PMF method.To summarize, we chose a reliable estimation range of NO3.org for each season: 0.12 to 0.19 µg m -3 for spring, 0.34 to 0.53 µg m -3 for summer and 0.21 to 0.33 µg m -3 for autumn.Furthermore, we found that organic nitrates contribute 9-21% to OAs in spring, 11-25% in summer and 9-20% in autumn based on the reliable estimation range of the NO3.org concentration and the assumption that the average molecular weight of organic nitrates ranges from 200 g mol -1 to 300 g mol -1 (Rollins et al., 2012).The level of NO3.org is highest during the summer in Shenzhen, which is consistent with the seasonal variations in literatures (Ng et al., 2017).In addition, it is seen that the difference in the NO3.org mass concentration between South China and the southeastern USA is small, even though the levels of anthropogenic emission species (such as BC and NOX) are much higher in South China than those in southeastern USA (Xu et al., 2015b); this implies that aerosol ONs might not be closely related to anthropogenic emissions.

Correlation between organic nitrates and LO-OOA
In this study, we also performed a PMF analysis on both organic spectra solely to investigate OA source apportionment.The same organic factors were identified as those in the PMF analysis on organics , combined with NOX + ions, including HOAs, LO-OOAs and MO-OOAs.For the total daily data, organic nitrates were better-correlated with LO-OOAs than with any other factor.Then, we further found a noticeably improved correlation between the LO-OOAs and organic nitrates at night (20:00-6:00) and a decreased correlation during the daytime (7:00-19:00) in Figure 3.The biggest improvement is seen in summer, where the correlation coefficient (R) increases from 0.77 for the whole day to 0.91 at night.Considering the relatively high BVOC emissions during the summer in Shenzhen (Zhu et al., 2012), the summer LO-OOAs may be closely related to the oxidation of BVOCs, especially at nighttime.

Nighttime formation of organic nitrates in particles
In this section, we will further investigate the potential formation pathway of organic nitrates at night according to the analysis of particulate organic nitrates in southeastern USA in Xu et al. (2015a).Since on-line VOCs measurement with an automated in situ gas-chromatography mass spectrometer (GC-MS) was only performed during the spring campaign, the followin g theoretical analysis is just applied to the dataset for the spring case.According to Figure 4, the concentrations of anthropogenic VOC species, such as propane and toluene, were dozens of times higher than those of biogenic VOCs at night.However, based on the estimation method in Section 2.3, two biogenic VOCs, i.e., limonene and -pinene, were identified as the key VOC precursors, accounting for approximately 90% of the NO 3 lost due to the reactions with VOCs.Thus, the nighttime SOA production from limonene and -pinene with NO 3 was further calculated (Text S1).The results showed that the estimated mass concentration range (0.15-1.29 µg m -3 ) for SOAs agrees well with the range (0.38-0.87 µg m -3 ) of organic nitrate concentrations at night.And noted that SOA yields from limonene + NO 3 is much higher than -pinene + NO 3 (Hallquist et al., 1999;Fry et al., 2011Fry et al., , 2014;;Spittler et al., 2006;Boyd et al., 2017).The diurnal patterns of the mass concentrations in the the lower PBL and rush hour traffic (He et al., 2011).Compared to BC, the organic nitrate concentration shows a much more distinct variation trend, with low mass loading in the daytime and high mass loading at night, and it has two unique rapidgrowth processes (19:00-22:00 and 3:00-6:00) after sunset, which cannot be explained by the PBL variation.Especially, the second increase in organic nitrates from 3:00-6:00 clearly suggests significant nighttime SOA formation.Compared to the bulk SOA, which shows an increase in the daytime related to photochemical formation while a steady level after midnight, the concentration of organic nitrates increases from 3:00-6:00 also clearly indicates different formation mechanism.However, the SOA of our theoretical calculation based on BVOCs and NO 3 indeed shows two similar increasing processes, well explainin g the observed trends.Since the similar nighttime formation trend for organic nitrates from biogenic emissions was also seen in a forest environment in Finland (Yan et al., 2016), the results in this study shows that the BVOC+ NO 3 chemistry is also potentially critical for the formation of nighttime organic nitrates even in a polluted urban atmosphere.

Conclusions
An Aerodyne HR-ToF-AMS was deployed in urban Shenzhen in South China for one month per season during 2015-2016 to characterize particulate organic nitrates with high time resolution.We find that the mass fraction of organic nitrates in the total measured nitrates is substantial during warm seasons , including spring (13-21%), summer (41-64%) and autumn (16-25%), while the contribution is negligible during winter in South China.The comparison analysis between organic nitrates and each OA factor for different periods of the day shows good correlations (R=0.77 in spring, 0.91 in summer and 0.72 in autumn) between organic nitrates and the LO-OOA factor at nighttime.Based on this, we further investigate the potential pathway for the formation of nighttime organic nitrates, and the results suggest that the BVOCs+NO 3 chemistry plays a key role in the formation of nighttime organic nitrates and SOAs, and limonene is the most important precursors of VOCs for this type of reaction.Overall, we infer that, even in polluted urban atmosphere with high abundance of anthropogenic pollutants, particlephase organic nitrates are possibly mostly derived from the formation from biogenic VOCs, and may be a good surrogate for studying the effects of biogenic emissions on air pollution and the climate.

Figure 1 .
Figure 1.The location of the sampling site.
Figure 2. (a) Time series of NO3.org concentrations estimated by the NO + /NO2 + ratio method and PMF method for each season, (b) scatter plots of NO3.org1_ratio and NO3,org_PMF.
Figure 3. Scatter plots of NO3.org1_ratio and LO-OOA for each season for the whole day (a), at nighttime (b) and during the day (c).
. Phys.Discuss., https://doi.org/10.5194/acp-2018-1009Manuscript under review for journal Atmos.Chem.Phys.Discussion started: 8 October 2018 c Author(s) 2018.CC BY 4.0 License.measured BC, SOA (LO-OOA+MO-OOA) resolved by PMF, organic nitrate functionality (NO3.org1_ratio) and SOAs from the BVOC+ NO 3 in the low SOA yield are shown in Figure 5.The concentration of BC is low during the daytime, with a planetary boundary layer (PBL) height that is increased, while the concentration increases in the early evening due to the effects of b oth

Figure 4 .
Figure 4.The mean concentrations of VOCs and the corresponding NO3 reactivity at night during the spring campaign.

Figure 5 .
Figure 5.Diurnal trends of measured BC, SOA resolved by PMF, organic nitrate functionality (NO3.org1_ratio), and SOA fro m BVOC+ NO 3 in a low SOA yield.

Table 1 .
Meteorological conditions, PM1 species concentrations and relevant parameters in the estimation of organic nitrates for different seasons in Shenzhen.

Table 2 .
A summary of organic nitrate estimations via the NO + /NO2 + ratio method and PMF method a NO3.org for upper bounds is denoted as NO3.org1_ratio, and NO3.org for lower bounds is denoted as NO3.org2_ratio b NO3.org estimated using the PM F method is denoted as NO3.org_PMF