Effects of Inorganic Seeds on Secondary Organic Aerosol (SOA) Formation

aerosols was dominant with low initial hydrocarbon concentration in the competition, while the reverse was true with high initial hydrocarbon concentration. This illustrates that the interplay of different compositions of real atmosphere aerosols can lead to complex synergistic effects on SOA formation.


Introduction
Atmospheric aerosol has significant influences on human health (Kaiser, 2005), visibility degradation (Cheng et al., 2011), and climate change (Satheesh and Moorthy, 2005). It was found that organic aerosols (OA) was the most abundant component of atmospheric aerosol (He et al., 2001) and more than 50% of the total OA are secondary organic aerosols (SOA) (Duan et al., 2005). SOA are produced from the oxidation of volatile organic compounds (VOCs) followed by gas-particle partitioning of the semivolatile organic products. Among the various VOCs, aromatic hydrocarbons are one type of SOA precursors which have drawn the most attention due to their abundance in the air and high SOA contribution to urban atmospheres (Lewandowski et al., 2008). Toluene and m-xylene are the two of the most abundant aromatic hydrocarbon species.
The detailed mechanism and controlling factors of SOA formation are not fully understood yet, which leads to the lower SOA level prediction from air quality models than the ambient measurements (Volkamer et al., 2006). Using smog chamber, SOA formation process can be investigated under controlled experimental conditions. Series of smog experiments have been conducted by different research groups to investigate the effects of background seed aerosols on SOA formation (Cao and Jang, 2007, Czoschke et al., 2003, Gao et al., 2004, Jang et al., 2002, Liggio and Li, 2008. Increased SOA formation and SOA yields were observed with the presence of acid seed aerosols. The effects of acidic seeds suggest that aerosol phase reactions may play an important role on SOA formation (Jang et al., 2002). Interactions between the organic and inorganic components of aerosols are important for further understanding the SOA formation process. Most research concludes that acid-catalyzed aerosol-phase reactions generate additional aerosol mass due to the production of oligomeric products with large molecular weight and extremely low volatility (Cao and Jang, 2007, Czoschke et al., 2003, Gao et al., 2004 and, therefore, enhance SOA formation. Uptake of semivolatile organic products to acidic sulfate aerosols was also found contributing to enhance SOA formation (Liggio and Li, 2008). In these studies, (NH4)2SO4 or H2SO4 seed aerosols were widely used to study the effect of particle acidity on SOA formation from both biogenic and aromatic hydrocarbons.
Atmospheric aerosols always have a very complex composition. Studying the effects of (NH4)2SO4 or H2SO4 seed aerosols did not draw the whole picture of the role that inorganic seed aerosols play in SOA formation. Metal-containing aerosols are important components of the atmosphere. Calcium and iron are the most abundant metal species in atmospheric aerosols and the average concentration of them in Beijing could be as high as about 1.2 μg m -3 and 1.1 μg/m 3 in PM2.5 (He et al., 2001) respectively. In this study, we tested the effect of different inorganic seeds on SOA formation using a smog chamber. Two aromatic hydrocarbon precursors toluene and m-xylene are used. Effects of various inorganic seeds, including neutral inorganic seed CaSO4, acidic seed (NH4)2SO4, transition metal contained inorganic seeds FeSO4 and Fe2(SO4)3, and a mixture of (NH4)2SO4 and FeSO4, were examined during m-xylene or toluene photooxidation with the presence of nitrogen oxides (NOx).

Experimental section
The experiments were carried out in a smog chamber which was described in detail in Wu et al. (Wu et al., 2007). The 2 m 3 cuboid reactor, with a surface-to-volume ratio of 5 m -1 , was constructed with 50 μm-thick FEP-Teflon film (Toray Industries, Inc. Japan). The reactor was located in a temperature controlled room (Escpec SEWT-Z-120), with a constant temperature between 10 and 60 °C (± 0.5 °C). The reactor was irradiated by 40 black lights (GE F40T12/BLB, peak intensity at 365 nm). Based on the equilibrium concentrations of NO, NO2 and O3 in a photo-irradiation experiment of an NO2/air mixture, the NO2 photolysis rate was calculated at approximately 0.21 min -1 , using a method described by Takekawa et al. (2000Takekawa et al. ( , 2003. Prior to each experiment, the chamber was flushed for about 40 h with purified air at a flow rate of 15 L/min. In the first 20 hours, the chamber was exposed to UV light at 34 °C. In the last several hours of the flush, humid air was introduced to obtain the target relative humidity (RH). Seed aerosols were generated by atomizing salt solutions using a constant output atomizer (TSI Model 3076). To avoid hydrolysis and precipitation in the Fe2(SO4)3 salt solution, as little sulfuric acid as possible was added to the solution. What's more, for generating internally mixed seed aerosols, a mixed solution of FeSO4 and (NH4)2SO4, in which the concentration ratio of FeSO4 to (NH4)2SO4 is 1:5, was used. The generated aerosols were passed through a diffusion dryer (TSI Model 3062) to remove water and a neutralizer (TSI Model 3077) to bring the aerosols to an equilibrium charge distribution. The hydrocarbon, NO and NO2 were carried by purified dry air into the chamber. The concentrations were continuously monitored at a measurement point in the reactor until they were stable, ensuring the components in the reactor were well mixed. The experiment was then conducted for 6 hours with the black lights on.
A gas chromatograph (GC, Beifen SP-3420) equipped with a DB-5 column (30 m×0.53 mm×1.5 mm, Dikma) and flame ionization detector (FID) measured the concentration of the hydrocarbon every 15 min. NOx and O3 were monitored with an interval of 1 min by a NOx analyzer (Thermo Environmental Instruments, Model 42C) and an O3 analyzer (Thermo Environmental Instruments, Model 49C), respectively. Size distribution of particle matter (PM) was measured by a scanning mobility particle sizer (SMPS, TSI 3936) in the range of 17-1000 nm with a 6-min cycle. The volume concentration of aerosols was estimated from the measured size distribution by assuming the particles were geometrically spherical and nonporous.

Estimating the generated SOA mass (Mo)
Due to deposition of particles on the Teflon film, the measured aerosol concentration had to be corrected. Takekawa et al. (2003) developed a particle size-dependent correction method, in which the aerosol deposition rate constant (k(dp), h -1 ) is a four-parameter function of particle diameter (dp, nm), as shown in equation (1) The resulting k(dp) values for different dp (40-700 nm) were determined by monitoring the particle number decay under dark conditions at low initial concentrations (<1000 particles cm -3 ) to avoid serious coagulation. Based on more than 500 sets of k(dp) values (dp ranges from 40 to 700 nm), the optimized values of parameter a, b, c, and d were calculated to be 6.46×10 -7 , 1.78, 13.2, and -0.957, respectively. It should be noted that the estimation of deposited aerosol concentrations using this method might introduce some error (Takekawa et al., 2003) because some scatter was recognized when fitting k(dp) values into equation (1). To reduce error due to wall deposition, SOA yields were calculated when the measured particle concentration reached its maximum in the experiments because deposited aerosols were a greater proportion of the aerosol concentration change in the reactor after that time.

Calculation of SOA yields
The fractional SOA yield (Y), defined as the ratio of the generated organic aerosol concentration (Mo) to the reacted hydrocarbon concentration (ΔHC), was used to represent the aerosol formation potential of the hydrocarbon (Pandis et al., 1992). Odum et al. (1996) developed a gas/particle absorptive partitioning model to describe the phenomenon that Y largely depends on the amount of organic aerosol mass present. Equation (2) illustrates the relationship between SOA yield and organic aerosol mass concentration: In equation (2), i presents the serial number of the hydrocarbon reaction products, Ai, αi and Kom,i (m 3 μg -1 ) are the aerosol mass concentration, the stoichiometric coefficient based on mass and the normalized partitioning constant for product i respectively. If we assume that all semi-volatile products can be classified into one or two groups, equation (2) can be simplified to a one-product model (i.e., i=1) or two-product model (i.e., i=2). Parameters (α and Kom) can be obtained by fitting the experimental SOA yield data with a least square method. Since numerous compounds are actually produced by the reaction of a hydrocarbon, parameters obtained by the simplified model only represent the overall properties of all products (Odum et al., 1996). A one-product model was proved sufficiently accurate to describe the relationship between aerosol yield and mass (Henry et al., 2008, Takekawa et al., 2003, Verheggen et al., 2007. Therefore, we used a one-product model for our experimental SOA yield data to quantify of the effects of inorganic seed aerosols on SOA formation.
Comparing the temporal variation of NO and O3 during these experiments with similar initial conditions (Figure 1), the results indicate that CaSO4 and (NH4)2SO4 seed aerosols have no significant effect on gas-phase reactions. This result is consistent with the findings of Kroll et al. (2007) and Cao and Jang (2007) that (NH4)2SO4 and (NH4)2SO4/H2SO4 seed aerosols had a negligible effect on hydrocarbon oxidation.
Similarly, by comparing the temporal variation particle concentrations (Figure 2) during the experiments with identical initial conditions except for the seed aerosols, the effects of CaSO4 and (NH4)2SO4 seed aerosols on SOA formation were identified. In Figure 2, PMcorrected was calculated from the measured PM concentrations plus wall deposit loss, and PM0 was the seed aerosol concentration. The results indicate that the presence of neutral aerosols CaSO4 (16-73μg m -3 ) in the m-xylene/NOx photooxidation system have no significant effect on SOA formation. Experiments with the presence of acid aerosols (NH4)2SO4 have different particle profiles according to the concentrations of the introduced (NH4)2SO4 seed aerosol. In Figure 2, experiment Xyl-AS2 has similar particle profile with the seed-free experiment Xyl-N5, indicating that (NH4)2SO4 seed aerosols have little effect on SOA formation when the initial concentration is low. However, when with high concentration of (NH4)2SO4 seed aerosol introduced, SOA formation was enhanced (i.e. experiments Xyl-AS3 and Xyl-AS9) comparing with the seed-free experiment Xyl-N5. Comparing experiments Xyl-AS3 and Xyl-AS9, higher concentration of (NH4)2SO4 seed aerosol resulted in higher SOA concentration. Therefore, the effects of (NH4)2SO4 seed aerosol on SOA formation depend on its concentration.

Experiment
no.

HC0
(ppm) initial m-xylene concentration (HC0), initial seed aerosol mass concentration (PM0), initial seed aerosol surface concentration (PM0,S), initial NOx concentrations (NO0 and NOx,0-NO0), ratio of HC0/NOx,0, generated SOA mass (Mo), reacted hydrocarbon (ΔHC) , and SOA yield (Y) Further analysis found that the effects of (NH4)2SO4 seed aerosol on SOA yield were positively correlated with the surface concentration of (NH4)2SO4 seed aerosols. To draw the SOA yield curves shown in Figure 3, the experiments were classified into different groups (experiment Xyl-AS3 was not classified into any group since the surface concentration of (NH4)2SO4 seed aerosols in this experiment was different from others) by the surface concentration of (NH4)2SO4 seed aerosols. The regression lines for each group (there was no regression line for experiments XylCS1~2 and Xyl-AS1~3 since they had similar SOA yield with the seed-free experiments) were produced by fitting the data of generated SOA mass (Mo) and SOA yield (Y) into a one-product partition model. As indicated in Figure 3, experiments with higher surface concentration of (NH4)2SO4 seed aerosols had higher yield curves. As proposed by most research, acid-catalyzed aerosol-phase reactions (Cao and Jang, 2007, Czoschke et al., 2003, Gao et al., 2004 and uptake of semivolatile organic products to acidic sulfate aerosols enhance SOA formation (Liggio and Li, 2008). The observed SOA formation enhancement could be related to the acid catalytic effect of (NH4)2SO4 seeds on particle-phase surface heterogeneous reactions and the surface uptake of semivolatile organic products.

Effects of Fe2(SO4)3 and FeSO4 seed aerosols on SOA formation
A seed-free experiment and three experiments with Fe2(SO4)3 seed aerosols were carried out to investigate Fe2(SO4)3 seed aerosols on phooxidation of toluene/NOx. The four experiments had identical initial conditions except for the concentrations of the introduced Fe2(SO4)3 seed aerosol. Fe2(SO4)3 seed aerosols did not have obvious effects on SOA formation as shown in the temporal variation of PMcorrected-PM0 concentrations in Figure 4. Fe2(SO4)3 seed aerosols had no obvious effect on gas phase compounds in toluene/NOx photooxidation either. A minimal amount of acid was added to the solution to generate Fe2(SO4)3 seed aerosols. The introduced H + concentration was in the range of 0.0002-0.002 μg m -3 in the Fe2(SO4)3-introduced experiments. This is much lower than the H + concentration in the "non-acid" experiment by Cao and Jang (2007). Therefore, we presume the effect of the introduced sulfuric acid was negligible and Fe2(SO4)3 seed aerosols did not have obvious effects on SOA formation in phooxidation of toluene/NOx. We also conducted 18 irradiated toluene/NOx experiments with/without FeSO4 seed aerosols. The conditions, generated SOA mass (Mo), and SOA yield (Y) are shown in Table 2. FeSO4 seed aerosols had no obvious effect on gas phase compounds either, but significantly suppressed SOA formation. Figure 5 compares the temporal variation of particle concentrations during the 4.2 ppm toluene experiments (Exierments Tol-N3, Tol-FS1, Tol-FS3, Tol-FS8 and Tol-FS12) conducted under identical initial conditions except seed aerosol concentrations. Experiments with the presence of FeSO4 seed aerosol generated less SOA than the seed-free experiment. And experiment with a higher FeSO4 seed aerosol concentration generated less SOA than experiment with a lower FeSO4 concentration. So the inhibited effect of FeSO4 aerosols on SOA yield became stronger at higher concentrations of FeSO4 seed aerosols. At other toluene/NOx photooxidation concentrations, we also found similar temporal variation of particle concentrations. However, as indicated in Table 2 and Figure 5, SOA yields of experiments Tol-FS1 and Tol-FS3 are similar to corresponding seedfree experiments of Tol-N3. These two seed-introduced experiments (as well as Tol-FS2) were conducted at the lowest ratio of FeSO4 seed aerosol mass concentration to initial toluene mass concentration (FeSO4/toluene) and did not show obvious effect on SOA formation comparing to their corresponding seed-free experiments. In these three experiments, the mass ratios of FeSO4/toluene (assuming particle density to be 1.898 g cm -3 , density of FeSO4·7H2O, because of the lack of the information the amount of hydrate water) were calculated to be lower than 4.2×10 -4 . It is possible that most of the ferrous iron was oxidized before significant SOA mass were generated since few FeSO4 seed aerosols were introduced and high concentrations of oxidizing substances were generated during the toluene/NOx photooxidation. Besides these three experiments with lowest FeSO4/toluene mass ratio, FeSO4 seed aerosols suppressed SOA formation relative to the corresponding seed-free experiments. And in our experiments, the suppress ratio could be as high as 60%, as calculated from   We classified the experiments with FeSO4 seed aerosols introduced into three groups by FeSO4/toluene mass ratios to create SOA yield variations as a function of generated SOA mass ( Figure 6). Experiments with different FeSO4/toluene mass ratios seemed to fall into different yield curves. When FeSO4/toluene mass ratio was lower than 4.2×10 -4 , FeSO4 seed aerosols had a negligible effect and SOA yields of these experiments with FeSO4 seed aerosols coincide with the yield curve of seed-free experiments. When FeSO4/toluene mass ratio was higher than 5.1×10 -4 , the SOA yield curve indicated experiments with FeSO4 seed aerosols had lower yields than seed-free experiments. Lower yield curves from the experiments with higher FeSO4/toluene mass ratio were observed, indicating that a higher Fe/C ratio had a greater suppression effect on SOA formation from toluene/NOx photooxidation.

Effects of mixed (NH4)2SO4 and FeSO4 aerosols on SOA formation
Atmospheric aerosol is often a mixture of different components. We tested the effect of internal mixed (NH4)2SO4 and FeSO4 seed aerosols on SOA formation in m-xylene/NOx photooxidaiton. The experimental conditions, generated SOA mass (Mo), and SOA yield (Y) are shown in Table 3. To generate internal mixed (NH4)2SO4 and FeSO4 aerosols, a mixed solution of (NH4)2SO4 and FeSO4, in which the mass concentration ratio of (NH4)2SO4 to FeSO4 was 5:1, was used in the atomizer. So the approximately 60 μm 3 cm -3 seed aerosols in the three experiments with mixed (NH4)2SO4 and FeSO4 seed aerosols (Xyl-FA1~3) contained about 10 μm 3 cm -3 FeSO4 seed aerosols and 50 μm 3 cm -3 (NH4)2SO4 seed aerosols.
As mentioned above, neither (NH4)2SO4 seed aerosols nor FeSO4 seed aerosols had obvious effects on gas phase compounds. And in the experiments in this section, we found that mixed (NH4)2SO4 and FeSO4 seed aerosols had no obvious effect on gas phase compounds either.
In Figure 7, after wall deposition correction and deduction of seed aerosols, temporal variation of particle concentrations in experiments conducted under identical initial conditions except seed aerosol concentrations (the initial concentration of m-xylene is 1.1ppm, 2.1ppm and 3.2 ppm in picture a, b and c, respectively) were compared.  Table 3. Experimental conditions and results in toluene photooxidation: initial toluene concentration (HC0), initial FeSO4 seed aerosol concentration (PM0), initial NOx concentrations (NO0 and NOx,0-NO0), ratio of HC0/NOx,0, generated SOA mass (Mo), reacted hydrocarbon (ΔHC) and SOA yield (Y)

Experiment
As indicated in Figure 7(a), comparing with the seed-free experiment Xyl-N7, both experiment Xyl-AS10 and experiment Xyl-FA1 had higher particle concentrations while experiment Xyl-FS1 had lower particle concentrations. So, in 1.1ppm m-xylene photooxidation, the presence of (NH4)2SO4 aerosols and mixed aerosols (mixed (NH4)2SO4 and FeSO4) both increased SOA formation, while the presence of FeSO4 suppressed SOA formation. In Figure 7(b) and Figure  7(c), the effects of single (NH4)2SO4 seed aerosols (promotion effect) and single FeSO4 seed aerosols (suppression effect) on SOA formation were consistent with Figure 7(a). However, the mixed aerosols seemed to have different effects on SOA formation in photooxidation systems with different initial concentrations of m-xylene. In Figure 7(b), experiment Xyl-FA2 had similar temporal variation of particle concentrations with its corresponding seed-free experiment Xyl-N8, and in Figure 7(c), experiment Xyl-FA3 had lower temporal variation of particle concentrations than its corresponding seed-free experiment Xyl-N9. It must be noted that the seed aerosols in experiments Xyl-FA1~3 had similar concentrations and components. So, aerosols at the same mixing ratio of (NH4)2SO4 and FeSO4 could either enhance or suppress SOA formation depending on the experimental conditions. It seemed that the promotion effect of (NH4)2SO4 aerosols and the suppression effect of FeSO4 aerosols competed when both of them existed. And the promotion effect of (NH4)2SO4 aerosols was dominant with low initial hydrocarbon concentration in the competition, while the reverse was true with high initial hydrocarbon concentration. This illustrates that the interplay of different compositions of real atmosphere aerosols can lead to complex synergistic effects on SOA formation. According to the composition of the seed aerosols, experiments with inorganic seed aerosols introduced were classified into three groups. In Figure 8, SOA yield (Y) variations as a function of generated SOA mass (Mo) from m-xylene/NOx photooxidation were plotted. The regression lines for each group were produced by fitting the data of generated SOA mass (Mo) and SOA yield (Y) into a one-product partition model. As indicated in Figure 8, experiments with the presence of (NH4)2SO4 had a higher SOA yield curve than the seed-free experiments, while experiments with the presence of FeSO4 seed aerosols had a lower one, indicating the presence of (NH4)2SO4 and FeSO4 seed aerosols increased and decreased SOA yield, respectively. For the experiments with mixed seed aerosols, their SOA yield curve was similar to or a little higher than the seed-free experiments when the SOA mass load was low, but their SOA yield curve was lower than the seed-free experiments when the SOA mass load was high.

Hypothesis for inorganic seed aerosols' effects
In our experiment, we observed that FeSO4 seed aerosols suppressed SOA formation while Fe2(SO4)3 seed aerosols had no effect on SOA formation. It appears that the inhibiting effect of Fe(II) involves its strong reducing properties. Hydrocarbon precursors are oxidized by OH·, NO3·, etc. During the gas phase reaction, the oxidized products usually have a lower saturation vapor pressure and, as a result, condense to the aerosol phase. When these oxidized condensable compounds (CCs) containing carbonyl, hydroxyl, and carboxyl groups (Gao et al., 2004, Hamilton et al., 2005 contact ferrous iron in the aerosol phase, they may react to produce ferric iron and less condensable compounds (LCCs) or incondensable compounds (ICs). The ferrous iron may stop some CCs from being further oxidized and forming low-volatility products (Hallquist et al., 2009), including oligomers (Gao et al., 2004). The experimental results also showed that the presence of neutral CaSO4 seed aerosols seed aerosols have no significant effect on photooxidation of aromatic hydrocarbons, while the presence of acid (NH4)2SO4 seed aerosols can significantly enhance SOA generation and SOA yield. A possible mechanism is shown in Figure 9. Oligomerization is one important step during SOA formation (Nguyen et al., 2011). As proposed by (Kroll et al., 2007), the effect of (NH4)2SO4 seed aerosols may be attributed to acid catalyzed particle-phase reactions, forming high molecular weight, low-volatility products (e.g. oligomers). These processes may deplete the semivolatile CCs in the particle phase, and enhance SOA formation by shifting the gas-particle equilibrium, which is shown in Figure 9, and, therefore force more CCs condense to aerosol phase. Since (NH4)2SO4 and FeSO4 seed aerosols may both influence the semivolatile CCs, there is a competition for CCs to form higher-volatility products (LCCs or ICs) or low-volatility products (e.g. oligomers).

Figure 9.
Hypothesized mechanism for inorganic seed aerosols' effects on SOA formation: ferrous iron Fe (II) reduces or decompose some condensable compounds (CCs), which are oligomer precursors, interrupting oligomerization and generating high volatility products (LCCs or ICs); while acid seed aerosols catalyze aerosol-phase reactions, generating oligomeric products

Conclusion
Effects of various inorganic seeds, including neutral inorganic seed CaSO4, acidic seed (NH4)2SO4, transition metal contained inorganic seeds FeSO4 and Fe2(SO4)3, and a mixture of (NH4)2SO4 and FeSO4, were examined during m-xylene or toluene photooxidation. Our results indicate that the presence of CaSO4 seed aerosols and Fe2(SO4)3 seed aerosols have no effect on photooxidation of aromatic hydrocarbons, while the presence of (NH4)2SO4 seed aerosols and FeSO4 seed aerosols have no effect on gas-phase reactions, but can significantly influence SOA generation and SOA yields. (NH4)2SO4 seed aerosols enhance SOA formation and increase SOA yield due to acid catalytic effect of (NH4)2SO4 seeds on particle-phase surface heterogeneous reactions. While FeSO4 seed aerosols suppress SOA formation and decrease SOA yield possibly due to the reduction of some oligomer precursor CCs. These results reveal that many inorganic seeds are not inert during photooxidation process and can significantly influence SOA formation. These observed effects can be incorporated into air quality models to improve their accuracy in predicting SOA and fine particle concentrations.