Influence of metal-mediated aerosol-phase oxidation on secondary organic aerosol formation from the ozonolysis and OH-oxidation of α-pinene

The organic component is the most abundant fraction of atmospheric submicron particles, while the formation mechanisms of secondary organic aerosol (SOA) are not fully understood. The effects of sulfate seed aerosols on SOA formation were investigated with a series of experiments carried out using a 9 m3 smog chamber. The presence of FeSO4 or Fe2(SO4)3 seed aerosols decreased SOA yields and increased oxidation levels in both ozonolysis and OH-oxidation of α-pinene compared to that in the presence of ZnSO4 or (NH4)2SO4. These findings were explained by metal-mediated aerosol-phase oxidation of organics: reactive radicals were generated on FeSO4 or Fe2(SO4)3 seed aerosols and reacted further with the organic mass. This effect would help to explain the high O/C ratios of organics in ambient particles that thus far cannot be reproduced in laboratory and model studies. In addition, the gap in the SOA yields between experiments with different seed aerosols was more significant in OH-oxidation experiments compared to ozonolysis experiments, while the gap in estimated O/C ratios was less obvious. This may have resulted from the different chemical compositions and oxidation levels of the SOA generated in the two systems, which affect the branching ratio of functionalization and fragmentation during aerosol oxidation.


Results
SOA formation from ozonolysis of α-pinene. A series of α -pinene ozonolysis experiments were carried out under similar conditions in the presence of sulfate seed aerosols close to normal environmental concentrations. The experimental details are introduced in Table S1 in Supplementary Information. After 4.5 hours of reaction, approximately 4.7 ppb of α -pinene was consumed consistently in the experiments. No scavenger for free radicals was used in these experiments. According to the time variations of ozone and α -pinene, as well as their reaction rate constant, it was estimated that about 50-60% of the consumed α -pinene reacted with ozone. The concentrations or reagent ion normalized signals (which are proportional to concentrations) for α -pinene, acetone, pinonaldehyde and pinic acid (using the m/z 71 fragment 60 ) and many other gas product fragments were not significantly different between experiments utilizing different types of sulfate seed aerosols. The differences between the average relative abundance of these gaseous products in experiments with different types of seed aerosol were less than 10%, with some examples shown in Figure S2 in the supplementary information. This suggests that the effect of seed aerosol type on gas phase reactions was not significant enough to be detected by the proton transfer reaction mass spectrometer (PTR-MS), and that the precursor concentration, oxidant levels and gas-phase wall losses between experiments were reproducible.
The organic mass concentration on the aerosols measured by aerosol mass spectrometer (AMS), due entirely to SOA formation, is plotted as a function of the reaction time in Fig. 1(a). The error bar in the figure is one standard deviation based on three repeated experiments. In this study, the wall losses of the aerosols were corrected using sulfate as a tracer for the AMS data 61 , while the data measured by the scan mobility particle sizer (SMPS) were corrected using the measured deposition rate in our previous study 59 , following the method developed by Takekawa et al. 12 . As shown in Fig. 1(a), the presence of dry (NH 4 ) 2 SO 4 , wet (NH 4 ) 2 SO 4 or ZnSO 4 seed aerosols resulted in slightly higher SOA formation than for FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols. However, these differences were found to be not significant according to the statistical analysis. The reason for this is that the initial α -pinene and ozone concentrations for the repeated experiments were not identical, resulting in non-identical α -pinene consumption and high relative standard errors (RSEs) for SOA mass. As shown in the inset pictures of Fig. 1(a), the RSEs of the SOA yields (the mass ratio of the generated SOA and consumed α -pinene) were found to be much lower than those for SOA mass. SOA yields in experiments in the presence of FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols were about 20% lower than those for (NH 4 ) 2 SO 4 or ZnSO 4 seed aerosols. These differences were statistically significant according to one-way ANOVA statistical analysis results and means comparison with the Dunn-Sidak test, the details of which can be found in the Supplementary Information. Assuming a density of 1.4 g cm −3 for SOA 62 , SMPS gave similar SOA mass concentrations to AMS. The variations in precursor concentrations and oxidant levels among experiments with different types of sulfate seed aerosols were on average less than 5%. The FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols likely accounted for the lower SOA yields in these experiments.
Oxidation level of SOA from the ozonolysis of α-pinene. The cause of the varying SOA yields in the presence of different sulfate seed aerosol types was examined through the analysis of individual organic aerosol mass spectra. The fragment with mass to charge ratio of 44 (m/z 44), arising mostly from CO 2 + , has been attributed to highly oxidized organic components 63 , such as carboxylic acids and acyl peroxides, and its ratio to the total organic signal (f 44 ) is often used to infer the degree of oxidation of SOA 26,64 . In this study, the O/C atomic ratio was estimated using the f 44 and the correlation derived by Aiken et al. 65 .
Despite small SOA mass differences between experiments, very different profiles for the f 44 and estimated O/C ratios were observed when conducting experiments with differing sulfate seed aerosols. In the first half hour of the experiment, the f 44 and estimated O/C ratios were highly uncertain, since organic aerosol concentration in this period is very low. The seed aerosols rather than the generated SOA had a large contribution to the f 44 and estimated O/C ratios in this period. The f 44 values for metal solutions and ammonium sulfate solutions are shown in Figure S5 in the Supplementary Information. As the reaction continued, the effect from seed aerosols became insignificant since the amount of generated SOA was enormous compared to the organics introduced together with seed aerosols. In the presence of (NH 4 ) 2 SO 4 , the oxidation level of SOA (as demonstrated by f 44 and O/C) consistently increased. Conversely, the oxidation level of organics decreased dramatically in the first few minutes on ZnSO 4 , FeSO 4 and Fe 2 (SO 4 ) 3 particles. As the reaction continued, the organic oxidation level on the ZnSO 4 seed aerosols decreased further, becoming similar to that of the organics on the (NH 4 ) 2 SO 4 seed aerosols. The oxidation level of organics on the FeSO 4 and Fe 2 (SO 4 ) 3 seed aerosols remained higher than that on the (NH 4 ) 2 SO 4 seed aerosols during the whole reaction. To explain the higher oxidation levels of organics on the FeSO 4 and Fe 2 (SO 4 ) 3 seed aerosols relative to the other seed aerosols, the condensed phase oxidation of organics is hypothesized to be responsible. Oxidation of condensed phase organics by gas phase radicals may also be possible and is summarized by George and Abbatt 66 . However, it is unlikely to result in significant differences for the organics partitioned to the different seed aerosols in this case because OH radical uptake on inorganics is relatively inefficient if the surface cannot be oxidized 66 . In addition, the organics generated in the gas phase should be similar among the experiments with different seed aerosols, and are unlikely to cause significant differences in the uptake of OH radical. It is further hypothesized that the iron cations in the seed aerosols are involved in the condensed phase oxidation. At 50% RH, FeSO 4 and Fe 2 (SO 4 ) 3 aerosols both demonstrated some hygroscopic growth, as shown in Figure S1 in Supplementary Information. The aerosols were in a metastable state between the crystalline and liquid phase. Consequently, there were likely aqueous-phase layers on the surface of these two seed aerosols during these experiments. Deguillaume et al. 67 provided a thorough review of the possible aqueous phase chemistry of Fe 2+ and Fe 3+ involving radicals and peroxides in the aqueous phase. Free radicals, including OH, can be formed from catalytic cycling of Fe 2+ and Fe 3+ in the aqueous phase of the ferric and ferrous sulfate aerosols. These radicals can react with the organic mass from the partitioning of gas-phase products of α -pinene oxidation on the seed aerosols, producing molecules that may contain more oxygen atoms than the original compounds. Further oxidation of SOA may also lead to some fragmentation of the condensed organics, which in the extreme case release more volatile organic compounds or even CO 2 to the gas phase. These fragmentation reactions should be responsible for the lower SOA yields in experiments in the presence of FeSO 4 and Fe 2 (SO 4 ) 3 seed aerosols relative to the other seed aerosols. Release of CO 2 to the gas phase might be an important carbon loss path from the aerosol phase, since the differences in gaseous products in experiments with different types of seed aerosol were not significant.  Table S2 in Supplementary Information. Similar to what was observed in the ozonolysis experiments, the time variation for gas phase precursors and products was similar between experiments, as demonstrated in Figure S3 in Supplementary Information. The average relative abundances of the gaseous products (inset of Figure S3) in experiments in the presence of FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols were similar to other experiments, with relative differences less than 15%. However, the SOA production varied significantly in the photooxidation experiments with different types of sulfate seed aerosols, as shown in Fig. 2(a). Similarly, the error bar in this figure is one standard deviation based on three repeated experiments. In experiments with FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols, the growth of organic aerosols was significantly slower than that in experiments with (NH 4 ) 2 SO 4 or ZnSO 4 seed aerosols. As a result, after 5 hours of reaction, the SOA concentrations in experiments with (NH 4 ) 2 SO 4 or ZnSO 4 seed aerosols were at least two times higher than those in experiments with the FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols. Since similar amounts of α -pinene were consumed in these experiments, SOA yields decreased by about 60% in experiments with the FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols compared to the seed-free experiment or experiments with (NH 4 ) 2 SO 4 or ZnSO 4 seed aerosols, which was much higher than the decrease in yield (about 20%) observed in ozone experiments.
Oxidation level of SOA from the OH-oxidation of α-pinene. The ratio of m/z 44 to the total organic signal (f 44 ) and O/C atomic ratio as a function of reaction time are shown in Fig. 2(b). In the experiments with (NH 4 ) 2 SO 4 seed aerosols, there was a significant increase in f 44 over time, consistent with numerous reports of SOA formation in chamber and field studies 26,30,65 . In contrast, SOA generated in experiments with the FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols were significantly more oxidized than that in experiments with the (NH 4 ) 2 SO 4 seed aerosols. It is interesting that the degree of SOA oxidation in FeSO 4 or Fe 2 (SO 4 ) 3 experiments increased in a short time, with the f 44 increasing to a plateau in about 45 minutes, as shown in both Figs 2 and S4. Such a rapid oxidation may be caused by oxidation with a high concentration of ROS in the condensed phase on the FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols. In addition to the Fenton reaction 68 , photolysis of Fe 3+ complexes 69 in the aqueous phase can also produce OH radicals under irradiation. As described by Deguillaume et al. 67 and references therein, the relative importance of Fenton reactions and the photolysis of Fe 3+ complexes in the production of OH radicals in solution remains unclear but is likely dependent upon many factors, including iron concentration and pH. In this study, the high concentration of iron ion may provide favorable conditions for the generation of free radicals. Besides, the plateau of f 44  This is statistically significant at the 0.05 level as determined using an ANOVA statistical analysis. However, for the OH-oxidation of α -pinene, SOA yields decreased more significantly between experiments, from 0.221 to 0.086 ± 0.014; a decrease of around 61%. As we noted above, ozonolysis experiments may be influenced  somewhat by OH radicals formed during the reaction. Therefore, the observed differences between the OH-oxidation experiments and the ozonolysis experiments in the effects of FeSO 4 and Fe 2 (SO 4 ) 3 seed aerosols on SOA formation are likely a lower limit of the true differences. The presence of the FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols also resulted in a statistically significant increase in oxidation level (relative to when the (NH 4 ) 2 SO 4 seed aerosols were used) in the ozonolysis of α -pinene, while this was not prominent in OH-oxidation experiments.
These differences in mass and SOA oxidation level may be explained by the competition between fragmentation and functionalization, whereby fragmentation and functionalization lead to a decrease of SOA mass and increase in oxidation level, respectively 30,32,33 . The branching ratio between fragmentation and functionalization is uncertain, but was found to increase as O/C rises 7,30,70 . The SOA was less oxidized in ozonolysis experiments than that in OH-oxidation experiments, as demonstrated in Figs 3 and S4. The lower oxidation level of the products might lead to a lower branching ratio between fragmentation and functionalization during the oxidation in the condensed phase with FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols, resulting in more-oxidized SOA and slightly lower SOA yield in ozonolysis experiments with FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols compared to ozonolysis experiments with (NH 4 ) 2 SO 4 seed aerosols. In contrast, the high branching ratio between fragmentation and functionalization resulted in much less SOA mass and a little higher O/C in OH-oxidation experiments with FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols relative to that in OH-oxidation experiments with (NH 4 ) 2 SO 4 seed aerosols.
Another possible reason for the differences in the change of oxidation level and SOA yields between ozonolysis and OH-oxidation experiments is a different amount of OH radicals in the condensed phase. With abundant peroxides in the oxidation of α -pinene 71,72 and similar conditions of seed aerosol and RH, OH radicals generated in the condensed phase were assumed to be similar between ozonolysis and OH-oxidation experiments. The differences caused by FeSO 4 or Fe 2 (SO 4 ) 3 between ozonolysis experiments and OH-oxidation experiments were likely to be due to differences in the initial chemical composition of the condensed SOA before it was partially fragmented by OH radical in the aerosol phase. Recently, iron-carboxylate complex photolysis was reported to be an important sink for carboxylic acids 73 . Thus, the presence of light may also contribute to the decrease in SOA yield in the OH-oxidation experiments in the presence of FeSO 4 or Fe 2 (SO 4 ) 3 seed aerosols with highly oxidized SOA 53 .

Discussion
The results suggest that iron-containing sulfate seed aerosols, i.e. FeSO 4 or Fe 2 (SO 4 ) 3 , could have a substantial impact on the formation and properties of SOA. The effects of FeSO 4 seed aerosol on SOA in OH-oxidation experiments were similar to those in previous studies 25,59 , but the decrease percentages of SOA mass were higher than in previous studies (60% vs. 8-34%) with similar concentrations of FeSO 4 seed aerosols. A detailed comparison of this study with other studies is introduced in Table S6 in the Supplementary information. The higher percentage decrease in this study occurred because most of the SOA were generated on the surface of FeSO 4 seed aerosols and were involved in the metal-mediated aerosol-phase oxidation in this study, while in previous studies the new particle formation was not controlled. What's more, FeSO 4 and Fe 2 (SO 4 ) 3 showed very similar effects on SOA mass and the O/C ratio in this study, indicating a cyclic oxidation-reduction of the iron. The generation of ROS in the aerosol phase can lead to rapid and efficient oxidation of SOA, resulting in oxidation level increase and aerosol mass loss. UV irradiation is not necessary for this effect since it was observed in both dark ozonolysis experiments and HONO photo-oxidation experiments. Most of the experiments were repeated three times to quantify the experimental errors, and this effect was found to be statistically significant. So this effect may need to be considered in simulations of SOA formation, and would help to explain the high O/C of organics in ambient particles that thus far have not been able to be reproduced in laboratory and model studies. In addition, these effects were found to be related to the initial properties of the SOA. The aerosol mass loss in the oxidation is more pronounced for highly oxidized SOA compared to SOA with lower oxidation level, due to the presence of aerosol-phase ROS. The concentrations of ROS within atmospheric particles and their effects on different families of SOA are critical for the improved characterization of aerosol-phase processes of atmospheric organic aerosol.
One thing we should point out is that the initial concentrations of reactants in this study were higher than those present in the atmosphere. This served to generate high enough SOA concentrations to reduce the experimental uncertainty. The organic aerosol mass loading was comparable to that in the ambient atmosphere, and only one hydrocarbon was used in the experiments. According to a comparison of this study with previous studies, with details in Table S6 in the Supplementary Information, the suppressing effect of FeSO 4 on SOA mass was more significant with lower SOA mass loading. The concentrations of iron in the experiments were calculated and listed in Table S3 in the Supplementary Information. Although the total mass concentrations of iron in this study were lower than in our previous studies, and were comparable to that in the atmosphere in polluted cities, the water-soluble iron concentrations and the iron concentrations in the particle phase were higher than those under normal ambient conditions. The high water-soluble iron concentration is likely to cause an overestimation of the effect of iron sulfate on SOA formation, while the high iron concentration in the particle phase might cause an underestimation of the effect. This is discussed in detail in the Supplementary Information. The influence of metal-mediated aerosol-phase oxidation on secondary organic aerosol formation would be most significant with a low SOA loading and a high concentration of highly dispersed water-soluble iron.

Methods
Chamber facility. Experiments were carried out in a 9 m 3 cylindrical reactor, constructed from Teflon film and irradiated by 24 Sylvania black-light lamps (365 nm). A detailed description of the chamber has been published elsewhere 74 . The chamber was connected to a proton transfer reaction mass spectrometer (PTR-MS, IONICON Analytik) and an O 3 monitor (2B Technologies) to measure gas-phase organic compounds and ozone. A negative-ion proton-transfer chemical-ionization mass spectrometer (NI-PT-CIMS) was used to measure the initial concentration of HONO in each experiment to create similar gas-phase oxidation environments among the Scientific RepoRts | 7:40311 | DOI: 10.1038/srep40311 OH-oxidation experiments. The NI-PT-CIMS has been described in detail elsewhere 75 . Aerosols in the chamber were measured with a scan mobility particle sizer (SMPS, TSI Model 3080) and a compact time-of-flight aerosol mass spectrometer (AMS, Aerodyne Research, Inc. C-ToF-AMS) 76 to obtain the information regarding the size distribution and chemical composition of the aerosols.
Seed aerosol and oxidant generation. Sulfate seed aerosols were generated by atomizing 1 g L −1 of zinc sulfate, ferrous sulfate and ferric sulfate solutions using a constant output atomizer (TSI Model 3076). The generated droplets were passed through a diffusion dryer (TSI Model 3062), where the RH was below 10% at the exit, to obtain dry seed aerosols. A differential mobility analyzer (DMA, TSI Model 3081) was used to size select the seed aerosols (72 nm). Wet seed aerosols were obtained by directly introducing droplets into the DMA for size selection (79 nm). According to the SMPS measurements, most of the dry or wet seed aerosols (approximately 75% in number) had a diameter of 76 ± 5 nm at 50% RH in the chamber. O 3 was generated from zero air in a UV ozone generator (OG-1, PCI Ozone Corp.). OH radical was generated from photolysis of HONO, while HONO was generated by passing a HCl-containing gas stream through a tube containing NaNO 2 salt granules, as described by Roberts et al. 77 . Experimental procedure. In the ozonolysis experiments, the seed aerosols were introduced followed by O 3 . Gaseous α -pinene was generated by injection of liquid α -pinene into a vaporizer where a large flow of zero air carried the α -pinene vapor into the chamber to begin the reaction. In the OH-oxidation experiments, the seed aerosols were also introduced first, followed by HONO and α -pinene in a stream of zero air. Turning on the UV lights initiated the formation of OH radicals and was considered the reaction starting point. Due to the limited volume for sampling in the chamber and the ongoing particle deposition during the reaction, the experiments were stopped after approximately 5 hours of reaction. After 5 hours of α -pinene oxidation in both the ozonolysis and OH-oxidation experiments, a plateau in the concentration of organic aerosols was apparent. Due to the low concentrations of α -pinene and oxidants, as well as the sufficient surface area provided by the seed aerosols, no obvious homogenous nucleation was observed during this study.