The influences of ammonia on aerosol formation in the ozonolysis of styrene: roles of Criegee intermediate reactions

The influences of ammonia (NH3) on secondary organic aerosol (SOA) formation from ozonolysis of styrene have been investigated using chamber experiments and quantum chemical calculations. With the value of [O3]0/[styrene]0 ratios between 2 and 4, chamber experiments were carried out without NH3 or under different [NH3]/[styrene]0 ratios. The chamber experiments reveal that the addition of NH3 led to significant decrease of SOA yield. The overall SOA yield decreased with the [NH3]0/[styrene]0 increasing. In addition, the addition of NH3 at the beginning of the reaction or several hours after the reaction occurs had obviously different influence on the yield of SOA. Gas phase reactions of Criegee intermediates (CIs) with aldehydes and NH3 were studied in detail by theoretical methods to probe into the mechanisms behind these phenomena. The calculated results showed that 3,5-diphenyl-1,2,4-trioxolane, a secondary ozonide formed through the reactions of C6H5ĊHOO· with C6H5CHO, could make important contribution to the aerosol composition. The addition of excess NH3 may compete with aldehydes, decreasing the secondary ozonide yield to some extent and thus affect the SOA formation.

YG, 0000-0002-7826-5995 The influences of ammonia (NH 3  The chamber experiments reveal that the addition of NH 3 led to significant decrease of SOA yield. The overall SOA yield decreased with the [NH 3 ] 0 /[styrene] 0 increasing. In addition, the addition of NH 3 at the beginning of the reaction or several hours after the reaction occurs had obviously different influence on the yield of SOA. Gas phase reactions of Criegee intermediates (CIs) with aldehydes and NH 3 were studied in detail by theoretical methods to probe into the mechanisms behind these phenomena. The calculated results showed that 3,5-diphenyl-1,2,4-trioxolane, a secondary ozonide formed through the reactions of C 6 H 5Ċ HOO· with C 6 H 5 CHO, could make important contribution to the aerosol composition. The addition of excess NH 3 may compete with aldehydes, decreasing the secondary ozonide yield to some extent and thus affect the SOA formation.
2018 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited.

Introduction
Styrene is a highly reactive alkene with typical abundances ranging from 0.06 to 45 ppb in ambient atmosphere [1][2][3]. It can be emitted into atmosphere from abundant anthropogenic sources, such as adhesives, solvents, tobacco smoke and automobile exhausts [4,5]. With potential carcinogenic and mutagenic characteristics, styrene is known to be toxic to humans, and it can cause damage to the central nervous and reproductive systems if exposure to it occurs [6]. Furthermore, styrene is susceptible to reaction with ozone to form secondary organic aerosol (SOA), resulting in secondary pollution in the atmosphere [7].
Criegee mechanism, first proposed by Rudolf Criegee in 1949 [8], is widely accepted in the ozonolysis of unsaturated hydrocarbons in atmospheric chemistry. Criegee intermediates (CIs) are formed by the ring-opening reaction of a rather unstable primary ozonide (POZ) which formed directly by the 1,3-cycloaddition of O 3 across the double bond. Because of short lifespan and unavailable direct precursor, experimental studies about CI have not been carried out until recent years. Taatjes et al. first observed the simplest CI (·CH 2 OO·) directly using the tunable synchrotron photoionization combined with multiplexed mass spectrometry [9]. Thereafter several studies have been carried out to detect ·CH 2 OO· and (CH 3 ) 2Ċ OO· and study the kinetics of their unimolecular reactions using synchrotron photoionization mass spectrometry and spectroscopic methods [10][11][12][13][14]. More researches have been done using theoretical methods to investigate the reaction kinetics and mechanisms of bimolecular reactions, including the reactions of CI with NH 3 , NO x , SO 2 , (H 2 O) x , HO x and others [7,[15][16][17][18][19][20][21][22][23][24][25][26].
In addition of CI, aldehydes are also formed as co-products in the decomposition of POZ and can recombine with CI generating the more stable secondary ozonide (SOZ) intermediate. In the past few years, the SOZs were detected by several researches in the gas phase ozonolysis of simple alkenes [27][28][29]. At 730 Torr, a SOZ propene ozonide (methyl-1,2,4-trioxolane) was observed as the major product in the reaction of ·CH 2 OO· with CH 3 CHO, indicating collisional stabilization of the nascent SOZ near atmospheric pressure [30]. Analogously, Tuazon et al. proposed that C 6 H 5Ċ HOO· can combine with C 6 H 5 CHO to form a SOZ structure (3,5-diphenyl-1,2,4-trioxolane, DPSOZ) in the ozonolysis of styrene [2]. And it may undergo partial conversion into a hydroxyl-substituted ester (C 6 H 5 CH(OH)OC(O)C 6 H 5 ). This secondary ozonide has very low vapour pressure, which makes it easy to partition into the aerosol phase from gas phase and contribute a major composition in styreneozone oxidation reactions [7]. In addition, Winterhalter et al. [31] and Nguyen et al. [32] also verified the formation of internal SOZ in the reaction of β-caryophyllene with O 3 experimentally and theoretically, respectively. Therefore, in the ozonolysis of alkenes, reactions of CI with the simultaneously generated aldehydes in the system should be further studied because they may have an important influence on the distribution of products in both the gas phase and particle phase.
Jalan et al. studied the reaction mechanisms and kinetics for reactions of ·CH 2 OO· with HCHO, CH 3 CHO and CH 3 COCH 3 , in which SOZ and organic acids were formed, and the tendency to form SOZ was proposed in the order HCHO < CH 3 CHO < CH 3 COCH 3 [33]. Recently, Wei et al. investigated the detailed potential energy surface (PES) for the reaction of ·CH 2 OO· with CH 3 CHO, and proposed a slightly different pathway of SOZ isomerization [34]. All these theoretical calculations are only for the reactions between aldehydes with the simplest structures. Thus, it is of great necessity for investigating CIs reaction with aldehydes to understand the formation of SOAs in the ozonolysis of alkenes.
Furthermore, ammonia (NH 3 ) is an important alkaline constituent and plays an important role in the atmosphere [35,36]. Atmospheric NH 3 is emitted by different biogenic and anthropogenic sources, such as soil, vegetation, livestock waste, NH 3 -based fertilizer volatiles, mobile exhaust and biomass combustion emissions [35][36][37][38]. In addition to reactions with sulfuric acid and nitric acid to form ammonium sulfate and ammonium nitrate aerosols [39], NH 3 also participates in atmospheric oxidation process. For styrene-ozone system, the presence of ammonia decrease the SOA yield [7]. The authors considered that the major condensable species for this system are 3,5-diphenyl-1,2,3-trioxolane and a hydroxyl-substituted ester, and hypothesize that the presence of NH 3 may attack these two products causing their rapid decomposition. While for α-pinene ozonolysis reactions, SOA yield increased when NH 3 was added after the reaction ceased [21]. The resulting aerosol growth may be attributed to ammonium salts formed by the reaction between organic acids and NH 3 . Recently, Liu et al. investigated the influence of NH 3 on particle formation from gasoline vehicles exhausts [40]. Adding NH 3 into the reactor after 3 h photo-oxidation of these complex mixtures, the particle number concentration and mass concentrations increased rapidly, but the average carbon oxidation state of SOA remained almost unchanged. In theoretical calculations, Jørgensen & Gross investigated the reactions between NH 3 with a secondary ozonide and a hydroxyl substituted ester firstly, which formed in the ozonolysis of ethene [22]. Then the reactions between NH 3 and three simplest carbonyl oxides (H 2 COO, CH 3 HCOO and (CH 3 ) 2 COO) were studied detail [23]. The estimated reaction rates of carbonyl oxides and NH 3 range from 1.8 × 10 −13 to 5.1 × 10 −18 cm 3 molecule −1 s −1 , which is several orders of magnitude higher than between secondary ozonide and NH 3 . It may indicate that the influence of NH 3 on the formation of SOA is mainly by reacting with carbonyl oxides.
In this article, the ozonolysis of styrene was studied both experimentally and theoretically to explore the formation of SOA. Different from many researches investigating the contribution of NH 3 to atmospheric nucleation [41,42], NH 3 was added to investigate its influence on the formation of SOA in the reaction. As mentioned above, the reactions of CIs with aldehydes as well as NH 3 in the system need to be considered. A series of experiments under different initial conditions were carried out in a Teflon chamber, while the mechanism and the PESs for reactions between CIs and aldehydes and NH 3 involved were studied in detail by means of theoretical calculations. Finally, the influence of NH 3 on the formation of SOA was discussed by comparing the competitive reactions of NH 3 and aldehydes with CIs.

Experimental materials and methods
Experiments were carried out in a simulation chamber with the volume of the FEP-Teflon reactor about 6 m 3 (1.7 m × 1.7 m × 2.1 m, surface/volume ∼ 3.30 m −1 ). The enclosure wall of the chamber is filled with thermally isolated material, and the inner side of the wall is covered by reflective stainless steel to achieve uniform light intensity. The reactor is fixed on top and bottom frames, and the top frame can be moved vertically, so the FEP reactor is collapsible. All the connection parts of the reactor are made of Teflon or stainless steel without O-rings in order to avoid the evaporation of volatile organic compounds into the reactor. To minimize the influence of wall effect, the sampling tubes are stretched to the middle of the reactor. Prior to each experiment, the reactor was cleaned by reducing the reactor volume to less than 10% of its original volume and refilling it to its maximum volume with purified air at least five times.
The AADCO pure air generator (Model 737, USA) is used to purify the ambient air, and then the purified air is used as the background and carrier gas. In order to further purify the air, compressed air from the AADCO generator is passed through two consecutive scrubbers filled with activated carbon and silica gel respectively. Residual hydrocarbons, NO, NO 2 , O 3 (less than 1 ppbv), NH 3 (less than 1 ppbv) and particles (less than 10 particles cm −3 ) are almost undetectable after passing the purification system. In a typical experiment, a known volume of the styrene (99%, Sigma-Aldrich) was injected into a temperature-controlled glass bulb vaporizer (maintained at around 373 K in this work) by micro syringe and flushed into the reactor by purified air. Ozone was generated by an adjustable ozone generator (COM-AD-01, ANSEROS, Germany). Ammonia standard gas (50 ppmv in N 2 ) was provided by National Institute of Metrology, China, and added into the reactor through a mass flow controller. The temperature and relative humidity (RH) are monitored by a commercial temperature and humidity sensor (HC2-C05, Rotronic, China). The hydrocarbon concentration is detected by a gas chromatography equipped with flame ionization detector (GC-FID, Agilent Technologies, USA). O 3 concentration is measured by an O 3 analyser (Model 49i, Thermo Scientific, USA). NO, NO 2 , NO x and NH 3 mixing ratios are monitored using an NH 3 analyser (Model 17i, Thermo Scientific, USA). A scanning mobility particle sizer (SMPS 3936, TSI, USA), which consists of a differential mobility analyser (DMA, TSI model 3080) and a condensation particle counter (CPC, TSI model 3775), is used to measure the SOA particle concentrations and size distribution as a function of the reaction time. All experiments were carried out at room temperature and dry conditions (RH < 5%).

Theoretical methods
The reactions between CI and aldehydes as well as NH 3 involved in this work were studied using theoretical methods. Both the spin-unrestricted and spin-restricted form of the B3LYP functional was carried out on the transition states found in the reactions of CIs with formic acid in our previous study [24]. The results show that the UB3LYP and B3LYP energies, frequencies and geometrical parameters are identical. Therefore, all stationary points in the PESs were optimized using the B3LYP density functional method [43] and the 6-311G++(2d,2p) basis set. The harmonic vibrational frequencies were also calculated at the same level to characterize all stationary points as either minima or transition states. In addition, intrinsic reaction coordinate (IRC) calculations were performed for each transition state to confirm the connections between the expected reactants and products. The relative energies were obtained in high-level ab initio method CBS-QB3 [44,45]. In CBS-QB3 method, geometries are optimized and frequencies are calculated at the level of B3LYP/6-311G(d,p). Second, energy calculations at MP2/ 6-311+G(2df,2p) level are done and CBS extrapolation is calculated. Third, MP4(SDQ)/6-311G(d,p) and QCISD(T)/6-311G(d,p) single point energies are computed. Finally, two empirical correction terms: effect of absolute overlap integral and spin contamination, are considered in the overall energy estimate. All quantum chemical calculations in this work were performed with Gaussian 09 software package [46].

Results and discussion
3.1. Influence of NH 3

on SOA yield
As listed in table 1, the experiments performed can be classified into three scenarios according to the corresponding experimental conditions. In group A, ozonolysis of styrene was studied under various [O 3 ] 0 /[styrene] 0 ratios without NH 3 . In group B and C, the effects of NH 3 on SOA formation were studied. The difference is that, in group B, NH 3 was added at the beginning of the reaction under different [NH 3 ] 0 /[styrene] 0 ratios, while in group C, excess NH 3 was added after the reaction. All the experiments were carried out with the concentration of ozone in excess of styrene ([O 3 ] 0 /[styrene] 0 ratio range from 2 to 4), and styrene is consumed completely in each experiment. An average density of 1.2 g cm −3 was used to convert the total aerosol volume measured by DMA to total mass for aerosol formation from styrene ozonolysis.
According to the partitioning theory originally outlined by Pankow [47] and Odum et al. [48], SOA yield is defined as the ratio of the amount of SOA formed to the amount of hydrocarbon consumed, is the mass of organic aerosol formed by the oxidation of HC (μg m −3 ). The wall loss of volatile hydrocarbon is negligible for FEP-Teflon chamber, and the method of wall loss correction for aerosol has been described in our previous work by Hu et al. [49]. Briefly, particle wall loss can be described as where N(d p ) is the concentration of particles and k dep (d p ) is the deposition rate coefficient for particles with diameter d p [50]. The relationship between k dep and d p can be determined by optimization of four parameters (a, b, c and d) with the experimental data [51].
Parameters a, b, c and d were optimized to be 0.0094, 0.60639, 1449.97666 and 2.45854 respectively for this chamber, and then N(d p ) can be corrected and the suspended aerosol mass concentration M 0 can be calculated from the wall-loss corrected volume concentration.
After correcting with wall loss, the aerosol yields for all three group experiments are displayed in figure 1. Compared to the reactions without NH 3 , the addition of NH 3 led to significant decrease in SOA yield. The different initial concentration of NH 3 resulted in different SOA yield. Moreover, the addition of NH 3 before and after the reaction had obviously different influence on the yield of SOA. When NH 3 was added before the reaction (group B), the final SOA yield decreased significantly with the increase of [NH 3 ] 0 /[styrene] 0 ratio; while when NH 3 was added after the reaction (group C), the yield of SOA was clearly higher than that under a similar [NH 3 ]/[styrene] 0 ratio in group B, but it was still lower than that without NH 3 (group A). Figure 2 shows the changes of SOA concentration in the reaction of styrene and ozone when NH 3 was added 5 h after reaction starts. The solid black triangles and circles represent the number concentration of aerosol with and without wall loss correction, respectively. Similarly, the hollow red triangles and circles represent the volume concentration of aerosol with and without wall loss correction. As shown in figure 2, with NH 3 added, both the number and volume concentrations of the SOA are decreased significantly. Na et al. also found an obvious decrease for the volume concentration of SOA when NH 3 was added in the reaction of styrene with O 3 [7], which was similar to our result. For the number concentration, it showed a slight increase in their work, which differs from the present result, probably because of the high excess ozone used in our experiment.

Competitive reactions of Criegee intermediate with aldehydes and NH 3
As mentioned in the introduction, CIs and aldehydes generated in alkene ozonolysis reactions can further react with each other, producing a more stable SOZ which may have an important contribution to the aerosol phase. After the addition of NH 3 , the reactions between CIs and NH 3 may compete with the reactions of CIs with aldehydes, which could affect the formation of SOA. In the following, the reactions of CIs with aldehydes and NH 3 were studied in detail by theoretical methods to investigate whether the competition of these reactions affect the formation of SOA.  There are three possible reaction paths for the isomerization and unimolecular decomposition of DPSOZ. The first reaction path is to form benzoic acid and benzaldehyde ([C 6 H 5 COOH + C 6 H 5 CHO]) via the transition state (TS DPD1 ) with a barrier of −4.6 kcal mol −1 . TS DPD1 involves the break of the central O-O and C-O bond in DPSOZ accompanied by an H-shift process simultaneously. The second path is to generate phenyl formate (a product tentatively identified by Tuazon et al. [2]) and benzaldehyde ([HCOOC 6 H 5 + C 6 H 5 CHO]) products via TS DPD2 , which has the lowest energy saddle point (−6.8 kcal mol −1 ) and lies 2.2 kcal mol −1 below the TS DPD1 . In this process, the O-O and the central C-O bond in DPSOZ break and one O atom shifts to the benzene ring simultaneously. The third isomerization pathway is more complicated, which involves three transition states and two intermediates (DPSOZ → TS DPI1 → HPMB1 → TS DPI2 → HPMB2 → TS DPI3 ). The reaction barrier of this process is 4.2 kcal mol −1 below the bimolecular reactants. Regardless of reaction barriers, C 6 H 5 CHO acts as a bridge for the isomerization of C 6 H 5Ċ HOO· to HCOOC 6 H 5 and C 6 H 5 COOH.
Both the reaction of C 6 H 5Ċ HOO· + HCHO and ·CH 2 OO· + C 6 H 5 CHO generate the same structure of secondary ozonide PSOZ (3-phenyl-1,2,4-trioxolane). The detailed reaction pathways are shown in figure 6. Both of them begin with the formation of a pre-reactive complex before the transition state. Similar to DPSOZ, the formation of PSOZ is highly exoergic and followed by decomposition under three different pathways. That is, (i) generate formic acid and benzaldehyde ([HCOOH + C 6 H 5 CHO]) through the transition state TS PD1 ; (ii) generate phenyl formate and formaldehyde ([HCOOC 6 H 5 + HCHO]) through the transition state TS PD2 ; (iii) generate benzoic acid and formaldehyde ([C 6 H 5 COOH + HCHO]) through a more complicated pathway (PSOZ → TS PI1 → HMB1 → TS PI2 → HMB2 → TS PI3 ). It is worth noting that the energy of transition state TS PI1 is lower than TS PD1 by 5.3 kcal mol −1 , indicating that the reaction pathway involving the reaction state TS PI1 is expected to make a major contribution.
The       3-methyl-1,2,4-trioxolane to products by Wei et al. [34] (pathway A in the reference). The saddle point TS I1 is 7.0 kcal mol −1 lower than TS D in energy, indicating that the latter pathway makes a major contribution in these competition channels.

Criegee intermediate reactions with NH 3
The profiles of PES for reactions of C 6 H 5Ċ HOO· and ·CH 2 OO· with NH 3 are shown in figure 8. The structures of all the pre-reaction complexes, transition states, intermediates and stabilized products in the reactions are shown in figure 9. At the beginning of these reactions, there is a rapid pre-equilibrium between the reactants and the pre-reaction complex.  Figure 9. The structures of pre-reaction complexes, transition states, intermediates, and stabilized products in the reactions of C 6 H 5Ċ HOO· + NH 3 and ·CH 2 OO· + NH 3 . All these structures are optimized at the B3LYP/6-311++G(2d,2p) level.
barrier, the energy released by CI + aldehydes → SOZ may help it partly decompose. Formed without energy barrier and with low vapour pressure, DPSOZ (3,5-diphenyl-1,2,4-trioxolane) would make a major contribution to the aerosol composition. This conclusion is consistent with the previous experimental results [2,7]. However, unlike the previous conclusion that hydroxyl(phenyl)methyl benzoate (C 6 H 5 CH(OH)OC(O)C 6 H 5 ) may be one of the products, our theoretical calculations show that these intermediates (HPMB1 and HPMB2) are easily decomposed into C 6 H 5 COOH and C 6 H 5 CHO.
When NH 3 was added, CIs show a tendency to react with NH 3 to generate hydroperoxide alkylamine (HPMA and HMA), the saturated vapour pressure of which is not so low as DPSOZ and therefore not so prone to enter into the particle phase as DPSOZ. The rate constants for bimolecular reactions are calculated using conventional transition state theory (TST) (see electronic supplementary material). The results show that the ratio between the reaction rates for C 6 H 5Ċ HOO· reacting with C 6 H 5 CHO and NH 3 is 10 2 -10 5 . This means that, if the reaction between C 6 H 5Ċ HOO· and NH 3 plays a leading role compared to that of C 6 H 5 CHO, then the concentration of NH 3 should be 2 to 5 orders of magnitude larger than the concentration of C 6 H 5 CHO. In the experiments B1-B3, the amount of NH 3 added in this work was far more than that of aldehydes produced in the reaction. The addition of excess NH 3 would significantly consume the amount of CI, resulting in a marked decrease in the yield of the final SOA. However, in the experiments B4-B6, even under a smaller [NH 3 ] 0 /[styrene] 0 ratio, the SOA yield was still found to be decreased, indicating that the above competitive reaction mechanism is not the only way of the influence. There may be some other reaction mechanisms which can also reduce the SOA yield after the addition of NH 3 , like the nucleation effect or the organic amine formation. And this needs to be investigated in the further research.

Conclusion
This work has systematically investigated the effect of NH 3 on SOA yields from the ozonolysis of styrene using experimental and theoretical methods. Chamber experiments were carried out without NH 3 or under different [NH 3 ] 0 /[styrene] 0 ratios. The geometry optimization of the stationary points along the reaction pathways were calculated at B3LYP/6-311G++(2d,2p) level, and single point energies have been refined by CBS-QB3 theory. The following conclusions could be drawn from the present work. (i) The addition of NH 3 could lead to a decrease of SOA yield in the ozonolysis of styrene. And the higher the initial concentration of NH 3 is, the lower the final SOA yield observed. (ii) The addition of NH 3 at the beginning of the reaction or several hours after the reaction occurs has obviously different influence on the yield of SOA. (iii) Quantum chemical calculations reveal that a secondary ozonide 3,5-diphenyl-1,2,4-trioxolane (DPSOZ), formed through the reactions of the Criegee intermediate C 6 H 5Ċ HOO· with C 6 H 5 CHO, could make important contribution to the aerosol composition. (iv) The addition of excess NH 3 would significantly consume the amount of Criegee intermediate, which may decrease the secondary ozonide yield and thus decrease the SOA formation. (v) There may be some other mechanism for the influence of NH 3 (e.g. the nucleation effect or organic amine formation) on SOA formation, which may be dominant under low-NH 3 conditions. The findings herein indicate that the reactions of CIs with NH 3 may play a non-negligible role in locations where the concentration of NH 3 is relatively high, which could influence the aerosol formation.