Naphthalene SOA : redox activity and naphthoquinone gas – particle partitioning

Chamber secondary organic aerosol (SOA) from low-NOx photooxidation of naphthalene by hydroxyl radical was examined with respect to its redox cycling behaviour using the dithiothreitol (DTT) assay. Naphthalene SOA was highly redox-active, consuming DTT at an average rate of 118±14 pmol per minute per μg of SOA material. Measured particle-phase masses of the major previously identified redox active products, 1,2and 1,4-naphthoquinone, accounted for only 21± 3 % of the observed redox cycling activity. The redox-active 5-hydroxy-1,4-naphthoquinone was identified as a new minor product of naphthalene oxidation, and including this species in redox activity predictions increased the predicted DTT reactivity to 30 ± 5 % of observations. These results suggest that there are substantial unidentified redox-active SOA constituents beyond the small quinones that may be important toxic components of these particles. A gas-to-SOA particle partitioning coefficient was calculated to be(7.0± 2.5) × 10−4 m3μg−1 for 1,4-naphthoquinone at 25C. This value suggests that under typical warm conditions, 1,4-naphthoquinone is unlikely to contribute strongly to redox behaviour of ambient particles, although further work is needed to determine the potential impact under conditions such as low temperatures where partitioning to the particle is more favourable. Also, higher order oxidation products that likely account for a substantial fraction of the redox cycling capability of the naphthalene SOA are likely to partition much more strongly to the particle phase.


Introduction
Exposure to particulate matter (PM) is known to be a concern to human health.Oxidative stress induction is believed to be a major toxicological mechanism of inhaled particles.For example, reactive oxygen species have been observed to quickly rise in the lungs and hearts of rats after inhalation of ambient particles (Gurgueira et al., 2002), and oxidative stress mechanisms have been implicated in playing a role in asthma (Gurgueira et al., 2002;Li et al., 2003a) and other respiratory diseases (Li et al., 2008).
Inducing oxidative stress can involve either producing an excess of oxidants or depleting antioxidants within a cell.One chemical mechanism for inducing oxidative stress is through redox cycling reactions, where efficient redox cycling catalysts can accept electrons from electron-rich reducing agents; subsequently, the electrons are transferred to molecular oxygen to generate superoxide, allowing the cycle to repeat.This process has the potential to both deplete antioxidant species in the cell and generate reactive oxygen species, contributing to induction of oxidative stress.Redox cycling capacity can be measured using an acellular technique known as the dithiothreitol (DTT) assay (Kumagai et al., 2002).In the assay, a buffered aqueous extract of a PM sample is mixed with DTT, which contains two electron-rich thiol groups.The catalytic capacity of DTT to transfer these electrons to molecular oxygen can be observed by measuring the DTT rate of decay, which has been correlated to oxidative stress markers in cellular systems (Li et al., 2003b;Koike and Kobayashi, 2006).Constituents of PM that have been shown as active redox cycling catalysts include black carbon from diesel particles (Shinyashiki et al., 2009),  transition metals (Netto and Stadtman, 1996;Charrier and Anastasio, 2012), humic-like substances (Lin and Yu, 2011;Verma et al., 2012), and quinones (Kumagai et al., 2002).
Quinones are a class of organic molecules derived from aromatic species, containing two carbonyl functionalities on a larger conjugated hydrocarbon ring.It has been suggested that redox cycling by quinone species within particles can lead to sustained oxidant generation once particles deposit within the body (Squadrito et al., 2001).A general redox cycle with a quinone species (1,2-naphthoquinone) is shown in Fig. 1; the ability of quinones to act as redox cycling agents lies in the stable semiquinone radical that is formed in the process.Quinones and their biochemistry have been wellstudied and are known to be cytotoxic by both redox cycling reactions and their ability to conjugate with thiol groups (Gant et al., 1988).
Quinones are minor PM constituents, but due to their aforementioned potential to contribute to adverse health effects, their sources and ambient concentrations are important to understand.Quinones have been shown to be present in ambient particles (Cho et al., 2004;Chung et al., 2006;Valavanidis et al., 2006;Tsapakis and Stephanou, 2007;Eiguren-Fernandez et al., 2008a;Ahmed et al., 2009) and contribute to particle-bound stable radical species, as measured by electron paramagnetic resonance (Dellinger et al., 2001;Pan et al., 2004;Valavanidis et al., 2006).The source of these species may be primary or secondary.Specific quinones, including the redox-active 1,2-naphthoquinone, 1,4-naphthoquinone, and 9,10-phenanthrenequinone, have been measured as products in diesel and gasoline exhaust particles (Cho et al., 2004;Valavanidis et al., 2006;Jakober et al., 2007).They may also be produced as secondary products of polycyclic aromatic hydrocarbon (PAH) oxidation.Naphthalene (Bunce et al., 1997;Chan et al., 2009;Kautzman et al., 2010;Lee and Lane, 2009;Sasaki et al., 1997), phenanthrene (Lee and Lane, 2010;Wang et al., 2007), and anthracene (Kwamena et al., 2006) all generate their corresponding quinone species (naphthoquinone, phenan-threnequinone, and anthraquinone, respectively) via reaction with gas-phase oxidants.Formation of these species from PAH precursors in the atmosphere could increase the redox activity of ambient particles during photochemical aging in locations with PAH emission sources.Indeed, photochemical production of 9,10-phenanthrenequinone has been observed as air masses age in the Los Angeles basin (Eiguren-Fernandez et al., 2008b).It has also been observed that particles collected in the afternoon in Los Angeles tend to be higher in water soluble organic carbon, organic acids and redox activity, suggesting that secondary organic aerosol formation may contribute to redox activity of ambient particles (Verma et al., 2009).
Evidence of secondary organic aerosol contribution to redox activity has also been found from controlled oxidation studies.The redox activity of the water-soluble extract of particles formed from aging diesel exhaust particles with ozone in an outdoor chamber was about a factor of two higher than particles not exposed to ozone (Li et al., 2009).Additionally, modifying the VOC mix within the chamber could alter the amount by which water-soluble redox activity increased, regardless of the presence or absence of diesel exhaust particles in the chamber (Rattanavaraha et al., 2011).Previous experiments we have performed revealed that when two-stroke gasoline engine exhaust is oxidised with ozone, the redox activity that results can be modelled with a redoxactive secondary organic aerosol that is over ten times more active per unit mass than the original primary engine particles (McWhinney et al., 2011).
While we know that PAH oxidation can generate quinone species and that quinones are efficient redox-cycling agents, very few studies so far have tried to extend this into examining our ability to predict and quantify the contribution of quinones to particle redox activity.In one such study researchers found that hydrogen peroxide generation from the DTT assay performed on particles from Fresno, California, could be accounted for by measuring the aerosol mass loadings of 9,10-phenanthrenequinone, 1,2-naphthoquinone, and 1,4-naphthoquinone (Chung et al., 2006).In another study, by assigning the DTT loss rate to individual particulate species again from the Fresno area, Charrier and Anastasio were able to predict DTT decay as arising 80 % from transition metal species and 20 % from quinone species (Charrier and Anastasio, 2012).
The aim of the current study was to examine to what extent redox behaviour can be predicted based on particulate mass loadings of 1,2-naphthoquinone, 1,4-naphthoquinone, and 9,10-phenanthrenequinone, i.e. the small, soluble quinones that are commonly associated with redox cycling.We have focussed on naphthalene secondary organic aerosol (SOA) generated by hydroxyl radical oxidation in a photoreaction chamber.Known to produce both 1,2 and 1,4 isomers of naphthoquinone (Lee and Lane, 2009), this provided a relatively simple system of redox-active aerosol generated through laboratory oxidation.The redox activity of the particles was confirmed, and the extent to which these two naphthoquinone isomers contributed to observed redox cycling was examined based on quantification by gas chromatography-mass spectrometry.During the chamber experiments, we used online measurements taken with an aerosol mass spectrometer and a proton transfer reaction mass spectrometer to determine a gas-particle partitioning coefficient to explore the likely contribution of naphthoquinone species to ambient particle redox cycling behaviour.Through this work, we hope to gain a more thorough understanding of how well we understand organic redox cycling catalysis on a molecular basis and determine how well we know the inventory of redox-active organic products.As we further examine the photochemical processes that generate redox activity in particles, we can better predict the possible influences of aerosol aging on particle toxicity.

Naphthalene SOA generation and collection
Chamber aerosol was generated by hydroxyl radical oxidation of naphthalene in a 1 m 3 Teflon (fluorinated ethylene propylene [FEP]) photoreaction chamber.The chamber is of a simple rectangular cuboid design with the inside of the chamber supported with a Teflon-wrapped metal frame and can be illuminated on four of the six faces with 24 ultraviolet lamps.Prior to particle generation, the chamber was flushed with several chamber volumes of zero air from an AADCO 737 Pure Air Generator.For the hydroxyl radical precursor, hydrogen peroxide was added to the chamber by bubbling zero air through a 30 % aqueous hydrogen peroxide solution (Sigma Aldrich) at a flow rate of approximately 400 cubic centimetres per minute (ccm) for one hour.Naphthalene was introduced to the chamber by vaporising a known mass of solid naphthalene (99 %, Acros Organics) in a flow of approximately 400 ccm of zero air.
Naphthalene concentrations in the chamber were monitored with an IONICON proton transfer reaction mass spectrometer (PTR-MS) equipped with a quadrupole mass spectrometer.Naphthalene and naphthoquinone concentrations were tracked by measuring the signal at mass-to-charge ratio (m/z) 129 and 159, respectively, as the singly protonated molecular ion.Once the PTR-MS naphthalene signal had stabilised in the chamber, indicating that mixing was complete, oxidation was initiated by illuminating the chamber with ultraviolet lamps with a peak emission at 310 nm (F40T12 UVB).Naphthalene concentrations were measured based on the PTR-MS signal calibrated to the known mass of naphthalene added to the chamber, assuming complete transfer and no wall loss.Hydroxyl radical concentrations are reported based on observed first order decay of naphthalene (transfer loss effects for naphthalene may affect the calibration, but this calibration does not affect any subsequent calcula-tions discussed in this study).Stable naphthalene signals in the absence of oxidant indicated that hydroxyl radical oxidation is the only significant loss mechanism for naphthalene.Particles were formed by the nucleation of oxidised organics without addition of seed particles.The resultant SOA was characterised by an Aerodyne aerosol mass spectrometer (AMS) equipped with a high-resolution time-of-flight mass spectrometer and, in some instances, a scanning mobility particle sizer (SMPS) consisting of a differential mobility analyser (model 3081) and a condensation particle counter (TSI model 3025, Shoreview, MN, USA).Temperature was not controlled or monitored in the chamber but was typically slightly above room temperature due to heat from the UV lamps, at about 25 • C.
When the AMS exhibited maximum particle loading, which was typically 90-120 min after the beginning of the experiment, the lamps were turned off to stop the oxidation reaction.Particles were subsequently collected by pulling chamber air through Zefluor filters (Pall, polytetrafluoroethylene (PTFE), 47 mm diameter and 2 µm pore size) at a flow rate of 20 litres per minute (lpm).To prevent the chamber from collapsing, a 20 lpm zero air make-up flow was added to the chamber.In some experiments, gas-phase constituents were collected during this time by pulling an additional 1 lpm of chamber air through a PTFE filter to remove particles and then through an ORBO 43 Supelpak-20 cartridge (Sigma-Aldrich) containing two XAD-2 polystyrene resin beds in series.On average, more than 98 % of 1,4-naphthoquinone was collected on the front XAD bed, indicating that analyte breakthrough was not a concern.Together, these naphthalene SOA filter samples and gas-phase XAD cartridge samples were collected for 60-90 min and stored at −20 • C until further analysis.

DTT assay
The DTT analysis used was adapted from the procedure described by Cho et al. (2005), and was slightly modified from our previous work (McWhinney et al., 2011).Naphthalene SOA particle filter samples were divided into quarters and one quarter was extracted by mixing with 10 mL of phosphate buffer (0.1 M, pH 7.4).DTT (Sigma Aldrich) was added to 1 mL aliquots of the aqueous filter extracts to give a final volume of 1.25 mL and a DTT concentration of 0.1 mM.The redox reaction was performed at room temperature and quenched at 15 min intervals over the course of one hour by taking 0.25 mL aliquots of the DTT mixture and adding 1 mL of 0.25 mM 5,5'-dithiobis-(2-nitrobenzoic acid) (DTNB, Alfa Aesar).DTT was quantified by measuring the absorbance of the resulting product at 412 nm with a UV-Vis spectrophotometer.Redox activity is expressed as the time rate-of-change of DTT concentrations in M min −1 or as a particle mass normalised DTT decay rate in pmol min −1 µg −1 .DTT analysis was carried out on the filter extract without filtration of insoluble matter, as assay results

Quinone analysis
Quinones were quantified by gas chromatography-mass spectrometry (GC-MS) using the derivatisation method from Cho et al. (2004).A one-quarter portion of the filter was extracted directly with dichloromethane (approximately 0.5 mL) by sonication for 15 min in a glass centrifuge tube.Menadione (2-methyl-1,4-naphthoquinone, Sigma Aldrich) was added to each extract as an internal standard.The quinone species were derivatised using reductive acetylation, wherein the quinone is reduced to the hydroquinone functionality in the presence of zinc and esterified in the presence of acetic anhydride (see Fig. 2).The derivatisation was performed by adding 0.5 mL of acetic anhydride (ACP Chemicals) and approximately 100 mg of powdered zinc (Sigma Aldrich) to the dichloromethane extract.The mixture was heated to 80 • C and mixed every five minutes.After 15 min, the reaction was cooled, an additional 100 mg of zinc was added, and the mixture was heated for another 15 min.After cooling, the remaining acetic anhydride was quenched with 1 mL of deionised water and the diacetylated quinones were extracted with 3 mL of pentane.After shaking, the mixture was centrifuged and the top organic layer was removed, dried under a stream of nitrogen, and reconstituted in 100 µL of dichloromethane.The dichloromethane mixture was analysed by GC-MS using the parameters in Table 1.Calibration standards of 1,2-naphthoquinone, 1,4-naphthoquinone, and 5-hydroxy-1,4-naphthoquinone (Sigma Aldrich) were prepared using the same procedure.The calibration was performed in each case based on the fully deacetylated fragment ion.Limits of quantification were operationally defined as the lowest standard concentration used for calibration, corresponding to quinone masses of 0.2 µg per full filter for naphthalene SOA samples.

Gas-particle partitioning coefficient calculation
The gas-particle partitioning coefficient of 1,4naphthoquinone was determined using the equation , 1994) where c p and c g are the concentrations of 1,4-naphthoquinone in the particle phase and the gas phase, respectively, and M p is the mass loading of the particles in the chamber.The concentration of 1,4-naphthoquinone in the gas phase was calculated from the PTR-MS signal at m/z 159 ([M + H] + ), and the concentration in the particle phase was determined from the AMS signal at m/z 158 (M + ).The PTR-MS response was calibrated by comparison with the quantified 1,4naphthoquinone from the XAD cartridge samples assuming full recovery of 1,4-naphthoquinone, while AMS signals were calibrated with the quantified 1,4-naphthoquinone from filter samples (r 2 = 0.9 for PTR-MS calibration and r 2 = 0.6 for AMS calibration).The particulate mass loading was taken as the total organics measured by the AMS.To correct for particle bounce due to the low relative humidities used in the experiments, a collection efficiency of 0.5 was used for all AMS data.To verify the suitability of this choice, AMS mass loadings were compared to mass loadings calculated from SMPS-measured particle volumes using a particle density of 1.55 g cm −3 (Chan et al., 2009); good agreement between the two values was obtained using 0.5 as the AMS collection efficiency.

Redox activity prediction for naphthalene SOA
We are interested in determining how well we can predict the redox activity of secondary organic aerosol particles using molecular-level information on the activities and loadings of known quinones.To start with as simple a system as possible by using only one SOA precursor, naphthalene was chosen for its ability to form aerosol in relatively high yield -particularly under low NO x conditions (Chan et al., 2009) -which was advantageous for the small size of our photoreaction chamber.The mass spectrum of the SOA generated in the chamber was fairly invariable from day-to-day and consisted mostly of hydrocarbon (C x H y ) and oxidised hydrocarbon (C x H y O z ) fragments; elemental composition information is given in Table 2, as obtained using established methods (De-Carlo et al., 2006;Aiken et al., 2007), and the average AMS mass spectrum from two representative experiments is shown in Fig. 3.There was an unexpected small fraction of the particle mass made up of C x H y N type fragments potentially as a result of amine contamination in the chamber, which was previously used for the photooxidation of ethanolamine.These peaks make up approximately only 3 % of the organic mass, on average, for an overall N : C ratio of approximately 0.03.It is currently unknown what types of species would be contributing these fragments to the overall mass spectrum.SOA generated from naphthalene oxidation was expected to exhibit high redox activity, based on knowledge that both 1,2-and 1,4-naphthoquinone isomers could be found in the particle phase as naphthalene oxidation products (Lee and Lane, 2009).The particles generated from naphthalene oxidation were highly redox active and reflective of the reproducible composition; DTT decay rates were reasonably invariant with an average DTT decay per unit mass of particle of 118 ± 14 pmol min −1 µg −1 (n = 10; individual experimental results are summarised in Table 3).As a comparison, we previously estimated the activity of the two-stroke engine exhaust SOA to be 8.6 ± 2.0 pmol min −1 µg −1 (McWhinney et al., 2011).For prediction of the SOA activity, we focussed on the 1,2-and 1,4-naphthoquinone isomers, which as mentioned in the Introduction, along with 9,10-phenanthrenequinone, are commonly believed to be the most important water soluble quinone electron transfer agents (e.g.Chung et al., 2006;Charrier and Anastasio, 2012).As mentioned in the Experimental section, the quinones were extracted directly from the filter using dichloromethane rather than measuring the water-soluble species, so it is assumed that the efficiency of quinone extraction was equal for phosphate buffer and dichloromethane.
Both 1,2-and 1,4-naphthoquinone were detected in the particle phase at mass loadings of a similar magnitude (see Table 3).Each run of the DTT assay included positive controls containing only 1,2-or 1,4-naphthoquinone in a Measured reported is based on dilution of the extract used in the assay for a particular sample and will not scale directly with reported quinone or PM masses.
phosphate buffer to determine the mass-dependent DTT decay rates for these species for any given set of assay runs.The activities of the positive controls were used to calculate the expected DTT decay of each SOA filter extract based on the measured naphthoquinones in the dichloromethane extract.
The predicted values based on 1,2-and 1,4-naphthoquinone particulate concentrations are plotted against the measured redox activity in Fig. 4 assuming the redox activities are additive, as was reported for quinone species in the supporting information of Charrier and Anastasio (2012).There is a strong correlation between the predicted and observed DTT decay rates (r 2 = 0.87).However, predicted DTT decay rates fall considerably below those observed.After correction for recovery (measured for 1,2-and 1,4-naphthoquinone as 83 ± 10 % and 87 ± 8 %, respectively), predicted values accounted for 21 ± 3 % of the measured values.
Based on the measured loadings of the two naphthoquinone isomers on the filters, we are considerably underestimating the observed redox activity of the aqueous filter extracts.One possible explanation for this discrepancy could be poor extraction of the quinone species using the dichloromethane extraction method compared to the aqueous extraction method.Considering the high solubility of naphthoquinone in dichloromethane and the aforementioned reasonable recoveries from spiked filters, this is not believed to be the cause of the discrepancy.Rather, we postulate that there are other species in the naphthalene SOA exhibiting redox activity.

Redox activity of naphthoquinone analogues
The rate of quinone redox cycling is determined primarily by the one-electron redox potential of the various species.For example, it has been previously shown that the logarithm of the rate constant of the reaction between a reducing agent, ascorbate, and a given quinone species is linearly related to the one-electron reduction potential of the conversion of the quinone to its corresponding semiquinone radical anion (Roginsky et al., 1999).Therefore, quinones with low one-electron reduction potentials are likely to exhibit slower redox cycling rates than those with higher potentials.With 1,4-and 1,2-naphthoquinones having one-electron reduction potentials of −140 mV and −89 mV, respectively (Wardman, 1989), we have looked for plausible products of the reaction of naphthalene and hydroxyl radical with a similar reduction potential that have not been previously identified in product studies.Specifically, the presence of 5-hydroxy-1,4-naphthoquinone and 5,8-dihydroxy-1,4-naphthoquinone -with reduction potentials of −93 mV and −110 mV, respectively (Wardman, 1989) -were examined due to the plausibility of these compounds as third and fourth generation oxidation products of naphthalene.
The 5-hydroxy-1,4-naphthoquinone product (but not 5,8dihydroxy-1,4-naphthoquinone) was identified and quantified in the naphthalene SOA particles by GC-MS, confirmed by retention time (Fig. 5) and fragmentation pattern.To the best of our knowledge, this is the first time this product had been identified in secondary organic aerosol produced from the oxidation of naphthalene.Like 1,4-naphthoquinone, 5hydroxyl-1,4-naphthoquinone has a considerable signal at the molecular ion in the electron impact mass spectrum (NIST Mass Spec Data Centre, 2013).We compared the AMS time series data for the first generation product 2formylcinnamaldehyde through the m/z 131 [M−CHO] + (C 9 H 7 O + ) fragment (NIST Mass Spec Data Centre, 2013), the second generation product 1,4-naphthoquinone molecular ion at m/z 158 (C 10 H 6 O + 2 ), and the expected m/z 174 signal for the molecular ion of 5-hydroxy-1,4naphthoquinone (C 10 H 6 O + 3 ) from high-resolution data (i.e.excluding other ions at each nominal mass).The temporal trends, shown in Fig. 6, are supportive of the production of first generation 2-formylcinnamaldehyde, second generation 1,4-naphthoquinone, and third generation 5-hydroxy-1,4-naphthoquinone in the particle phase, with each rising and falling in concentration progressively later in the experiment.
As is anticipated from a probable third generation oxidation product of naphthalene, the yield of 5-hydroxy-1,4naphthoquinone is somewhat lower compared to the second generation 1,2-and 1,4-naphthoquinones.The molecule was not detected in gas-phase XAD samples from the chamber and particle-phase loadings were on average about 41 % of 1,4-naphthoquinone and 55 % of 1,2naphthoquinone after correction using an average recovery of 40 ± 12 % from spike and recovery measurements.While 5-hydroxy-1,4-naphthoquinone is a minor contributor in terms of aerosol mass, the redox activity of 5-hydroxy-1,4-naphthoquinone is higher than that of the other two quinone species (about 7.8 pmol min −1 ng −1 compared to 2.1 and 5.7 pmol min −1 ng −1 for 1,4-and 1,2-naphthoquinone, respectively).The 5-hydroxy-1,4-naphthoquinone accounts for, on average, 33 % of the predicted redox activity in the particles (Fig. 7).Including 5-hydroxy-1,4-naphthoquinone in the predicted DTT decay rates, closure improves to 30 ± 5 % in Fig. 4 (versus 20 ± 3 % for 3 and 4 August filter samples using only 1,2-and 1,4-naphthoquinone to predict redox activity).The recovery of 5-hydroxy-1,4-naphthoquinone was low and variable compared to 1,2-and 1,4-naphthoquinone; it is possible that this may have resulted in an overestimation in particulate mass loadings for this species (and subsequently the relative contribution of the molecule to the measured redox cycling rates).As such, the numbers reported for this species above may be considered as an upper limit.Regardless, this product is significant in terms of overall redox cycling.If the redox activity of the particles is calculated without correcting for different recoveries, 5-hydroxy-1,4naphthoquinone is still expected to account for about 19 % of the predicted redox cycling on average.It is not implausible that other functionalised naphthoquinone species contribute small amounts to the unaccounted redox cycling activity.It is beyond the scope of the current project to identify remaining redox-active species, but we note that there are very likely as-of-yet unidentified redox-active organic compounds that could contribute to redox cycling in particulate matter in addition to the three small quinone species that have been the subject of previous focus.

Air-particle partitioning of 1,4-naphthoquinone
The degree to which the redox-active naphthoquinone products influence the toxicity of particles depends on the partitioning behaviour of the compounds.The drawback of chamber photooxidation studies is that particle mass loadings are typically many times higher than those observed under ambient conditions, which tends to favour partitioning to the particle phase.In the current experiments, particle mass loadings were on the order of several hundreds of micrograms per cubic metre, and so whether or not 1,2-or 1,4-naphthoquinone partition to the particle phase in significant enough quantities to be of concern to health under typical ambient conditions is an important question.Gas-particle partitioning coefficients allow some estimation of how important naphthoquinone photooxidation is to the health impact of particulate matter, particularly in urban areas where anthropogenic emissions of naphthalene are important.Only one study has previously reported a gas-particle partitioning coefficient for 1,4-naphthoquinone, which was in diesel exhaust particles (Jakober et al., 2007).We aim to verify the partitioning behaviour of 1,4-naphthoquinone in a system with only organic carbon aerosol and without potential interference from adsorption into the black carbon found in diesel particles.
The gas-and particle-phase concentrations of 1,4naphthoquinone were obtained from the PTR-MS and AMS measurements, respectively.In order to do so, a number of assumptions needed to be made.First, all AMS signal at m/z 158 was assumed to come from the C 10 H 6 O + 2 ion, which is the molecular ion of naphthoquinone.As the ionisation efficiency calibration was more reliable for the lower-resolution V-mode data from the AMS (using a shorter V-shaped ion time-of-flight region), these data were used to calculate the total masses of naphthoquinone and total SOA mass loading.Higher resolution W-mode data from the AMS (using a longer W-shaped ion time-of-flight region) confirmed that, on average, 90 % of the signal at m/z 158 came from the C 10 H 6 O + 2 ion, allowing the unit-mass resolution data from m/z to be used with an error of approximately 10 %.
The second assumption was that the AMS was sensitive to the 1,4 isomer but not the 1,2 isomer of naphthoquinone.This is based on the thermal instability of 1,2naphthoquinone.Thermal analysis of quinones revealed that while 1,4-naphthoquinone exhibits thermal stability up to 600 • C, 1,2-naphthoquinone begins decomposing at temperatures higher than 100 • C, and the thermal instability was verified by very poor reproducibility in GC-MS analysis (Sousa et al., 2012).The authors attributed the thermal instability to the carbonyl groups of the quinones being in the ortho position to each other, as 9,10-phenanthrenequinone also exhibited thermal instability.We postulate that 1,2naphthoquinone was only recently identified in naphthalene SOA (Lee and Lane, 2009) due to the tendency of the 1,2 isomer to degrade under the GC-MS conditions normally used to separate and measure quinone species, particularly at high injector temperatures.As such, we are only able to detect 1,2naphthoquinone by GC-MS when using the reductive acetylation derivatisation technique.
With the approximately 600 • C temperature of the AMS vapouriser, 1,2-naphthoquinone is not expected to be observable by the AMS.To verify this, each of the naphthoquinone isomers were separately mixed with a high molecular weight oil, bis(2-ethylhexyl) sebacate (BES).The mixtures were heated to homogeneously nucleate particles, which were subsequently sampled under the high-resolution mode of the AMS.Signals at m/z 158 from BES without quinone and with the 1,4 and 1,2 isomers are shown in Fig. 8.As hypothesised, the BES with 1,4-naphthoquinone exhibited signal at a nominal mass of 158 corresponding to the C 10 H 6 O + 2 molecular ion, which is partially obscured by signal from BES (from a fragment of C 9 H 18 O + 2 ).A signal for this ion was not observed for 1,2-naphthoquinone.Based on these results, we believe 1,2-naphthoquinone is likely to thermally decompose upon making contact with the AMS vapouriser and that the signal from the naphthoquinone molecular ion can be attributed to the 1,4 isomer.Reasonable correlation between the AMS m/z 158 signal with 1,4-naphthoquinone filter measurements (r 2 = 0.6) was obtained.The third assumption made was that the PTR-MS signal primarily arises from gas-phase 1,4-naphthoquinone.The reliability of this assumption depends on the relative amounts of 1,2-naphthoquinone and 1,4-naphthoquinone in the gas phase.Previous studies have found that in the gas phase, concentrations of the 1,2 isomer are lower than for the 1,4 isomer in both chamber (Lee and Lane, 2009) and ambient (Eiguren-Fernandez et al., 2008a) particles.This observation held true with our chamber experiments, with the ratio of the 1,2 to 1,4 naphthoquinone isomers being 0.015±0.010 in the gas phase.Based on this observation, attributing the PTR-MS signal to 1,4-naphthoquinone is a reasonable assump- Representative experimental data for a naphthalene generation experiment.AMS organics (green) and m/z 158 (orange) were corrected for a collection efficiency of 0.5.The PTR-MS signal for protonated naphthoquinone at m/z 159 (blue) was normalized to the reagent ion count.The gas-particle portioning coefficient K p (black triangles) was calculated for each time point using AMS and PTR-MS data calibrated to filter and XAD measurements of 1,4naphthoquinone, respectively.
tion.Calibration using five experiments with collected XAD cartridges yielded a strong correlation between the PTR-MS data and the 1,4-naphthoquinone extracted from the XAD resin (r 2 = 0.90).Finally, some evidence of equilibrium must be exhibited to properly calculate a partitioning coefficient assuming partitioning into a bulk phase.The low relative humidity in the chamber is likely to result in particles that are not liquid, but equilibrium can be established over appreciable periods of time if the particles are at least semi-solid with fast enough diffusion in the bulk.Indeed, during the course of an experiment, the calculated values for the gas-particle partition coefficient were not constant, but tended towards a stable value near the time when the lights are turned off in the chamber, as illustrated in Fig. 9.The stable value indicated that the system likely reached equilibrium by the end of a typical 90 min experiment.Using a time scale of diffusion of τ = d 2 /(4π D), a time scale of 90 min corresponds to a diffusion coefficient of approximately 2 × 10 −15 cm 2 of 1,4-naphthoquinone in the SOA particle, which is within the range expected for a semi-solid particle (Shiraiwa et al., 2011).As such, the conditions to assume equilibrium partitioning in the bulk phase appear to be met.
We obtained an average partitioning coefficient for a single experiment by averaging the values taken over a 10 min period prior to the lamps being turned off: (7.0 ± 2.5) × 10 −4 m 3 µg −1 ; the uncertainty is based on the standard deviation of all experiments, for which the results are summarised in Table 4.As the PTR-MS was calibrated assuming full recovery of 1,4-naphthoquinone from the XAD cartridge, [OH] a /10 7 this may be a lower limit.This is a larger value than those estimated by the EPI Suite software, likely due to the potential for inaccurate predictions from structural activity relationships; the McKay model predicts a value of 1.24 × 10 −5 m 3 µg −1 , while the octanol-air model predicts a value of 1.56×10 −4 m 3 µg −1 (US EPA, 2013).However, this value is in good agreement with the previously mentioned value of 4 × 10 −4 m 3 µg −1 reported for 1,4-naphthoquinone organic matter partitioning in diesel exhaust particles (Jakober et al., 2007).

Air-particle partitioning of 1,2-naphthoquinone
A corollary to the assumptions we made to calculate the 1,4naphthoquinone partitioning coefficient is that we cannot use a similar technique to calculate a partitioning coefficient for 1,2-naphthoquinone.This is unfortunate, as we have found, like earlier studies (Lee and Lane, 2009;Eiguren-Fernandez et al., 2008a), that the 1,2 isomer of naphthoquinone partitions to the particle phase to a greater extent than the 1,4 isomer.As discussed by Lee and Lane (2009), the ratio of 1,2-to 1,4-naphthoquinone tends to be higher in the particle phase than in the gas phase.The ratio of the 1,2 to 1,4 isomers we have measured in the gas phase and in the particle phase is 0.015 ± 0.010 (as mentioned previously) and 0.58±0.27,respectively, indicating that 1,2-naphthoquinone is much more abundant in the particulate phase.This is despite previous reports that the vapour pressure of the 1,2 isomer is higher than that of the 1,4 isomer (Jakober et al., 2007;Lee and Lane, 2009), although based on the Modified Grain method structural activity relationship from the EPI Suite software package, where the calculated vapour pressures at 25 • C are 1.08 × 10 −4 torr and 1.69 × 10 −4 torr for 1,2 and 1,4 naphthoquinone, respectively (US EPA, 2013), partitioning trends are consistent with vapour pressure.The ratios themselves are lower than the range of values summarised in Lee and Lane (2009), which suggests that the yield of 1,2-naphthoquinone relative to 1,4-naphthoquinone in our system is lower than in other studies.As those values are taken from engine samples, urban PM, and high-NO x chamber studies, this may be attributed to the low-NO x conditions used in this study.Regardless, our findings agree that the vapour pressure of 1,2-naphthoquinone is in fact lower than that of 1,4-naphthoquinone based on the partitioning behaviour between the gas and particle phases, and thus 1,2naphthoquinone is potentially of more interest in the particle phase than 1,4-naphthoquinone.
It is somewhat unsatisfactory to use the filter and XAD cartridge data to estimate the partitioning coefficient for 1,2-naphthoquinone, as during the filter collection period, the system deviates from equilibrium.The calculated partitioning coefficient increases as the chamber is pumped, presumably due to dilution rates of the particles occurring on a faster time scale than the diffusion of 1,4-naphthoquinone out of the particle that is required to re-establish equilibrium.In absence of appropriate measurements, the ratio of the 1,2 to 1,4 isomers can be used to obtain an estimate of the partitioning coefficient by combining partitioning coefficient expressions for the two species to give K p1,2 = K p1,4 (r p /r g ), where r p and r g are the ratio of the 1,2 to the 1,4 isomer of naphthoquinone in the particulate and gaseous phases, respectively.Using data from our chamber, the value for K p is estimated to be (3.8 ± 2.7) × 10 −2 m 3 µg −1 ; this calculation was performed using only experimental data that included simultaneous gas-and particle-phase measurements.Using ratios obtained in other studies, we obtained values of (5.3 ± 4.6) × 10 −3 m 3 µg −1 (Lee and Lane, 2009) and (3.6±2.2)×10−3 m 3 µg −1 (Eiguren-Fernandez et al., 2008a).
We note that the estimate based on our data is approximately an order of magnitude higher than when we use the values from other studies.As the Lee and Lane (2009) values are obtained from a larger chamber that does not require dilution during particle sampling, and as the Eiguren-Fernandez et al. (2008a) values are taken from ambient particles, these values are more likely to be taken under equilibrium conditions.Given the close agreement between these estimated coefficients, we consider the calculations using the Lee and Lane (2009) and the Eiguren-Fernandez et al. (2008a) values to be a more reliable estimate of the gas-particle partitioning coefficient for 1,2-naphthoquinone.

Atmospheric implications
Naphthoquinones are already known to be minor contributors to particle mass, and as such the generation of quinone species is not a concern to SOA mass loadings.However, due to high SOA mass yields, naphthalene itself has a potential to contribute considerably to particulate mass in areas with strong PAH emissions.For example, it has been estimated that naphthalene, 1-and 2-methylnaphthalene, acenaphthalene and acenaphthylene from mobile sources could contribute 37-162 kg m −3 d −1 of SOA in Houston, Texas, compared to a total SOA estimate of 268 kg m −3 d −1 from mobile sources, with a substantial fraction (36-48 %) of the PAH SOA mass coming from naphthalene oxidation (Shakya and Griffin, 2010).For diesel exhaust, it was estimated that of SOA production from oxidation of light aromatics, PAHs, and long-chain alkanes, 58 % of the aerosol produced arises from PAH precursor species (Chan et al., 2009).With a significant amount of naphthoquinone production from naphthalene oxidation and sufficient organic aerosol mass loading available for gas-particle partitioning, naphthoquinones may have a significant impact in terms of the redox behaviour and toxicity of ambient particles.
There are few studies that have attempted to model the atmospheric burden of 1,4-naphthoquinone.One study modelled the atmospheric burden resulting from oxidation of naphthalene and 1-naphthol and the photolysis of 1nitronaphthalene in Southern California, and much of central Los Angeles was predicted to have 1,4-naphthoquinone concentrations in excess of 0.5 ng m −3 and up to 2.5 ng m −3 based on modelled naphthalene emissions for a week in July 1998 (Lu et al., 2005).These values are similar in magnitude to those that were subsequently measured in Southern California (Eiguren-Fernandez et al., 2008a).We have used the modelled as well as the observed quinone concentrations to make an estimate as to how abundant quinones should be in the particle phase and what their contribution to redox cycling activity would be over a wide range of particle mass loadings.
To predict particulate quinone mass loadings under different mass loadings of organic aerosol, the K p equation was rearranged to yield c p = (K p /M p )c T /(1 + K p /M p ), where c T is the total quinone mass loading (c p + c g ) and K p /M p is the ratio of the particle-phase mass loading and the gas-phase mass loading of the quinone.Using a quinone concentration of 1 ng m −3 , the particulate mass loading was calculated for 25 • C based on the experimentally determined gas-particle partitioning coefficient of (7.0 ± 2.5) × 10 −4 m 3 µg −1 for 1,4-naphthoquinone.For 1,2naphthoquinone, the partitioning coefficient was estimated as 4.4 × 10 −3 m 3 µg −1 based on the average of the two estimates discussed previously, with the uncertainty in the range of (1.4-9.9)×10 −4 m 3 µg −1 based on the upper and lower standard deviations of those estimates.The results of the calculations for a range of organic particle mass loadings approximately 0.1-100 µg m −3 are shown in Fig. 10.DTT activity was calculated based on an average DTT decay rate of 2.1×10 −3 pmol min −1 pg −1 and 5.7×10 −3 pmol min −1 pg −1 for 1,4-and 1,2-naphthoquinone, respectively.
Unsurprisingly, the particulate mass loading of each quinone rises substantially as more organic particles are present for partitioning.As a result, the DTT activity per unit volume of air, which is an important metric for exposure to toxic particles, rises with particle mass loading.The DTT activity normalised to particle mass, however, remains relatively unchanged and actually falls somewhat at high particle mass loadings as the additional mass partitioning to particles begins to diminish.Expected massnormalised DTT decay rates are anticipated to be on the order of 0.02 pmol min −1 µg −1 for 1,2-naphthoquinone and 0.0015 pmol min −1 µg −1 for 1,4-naphthoquinone.As PM in Los Angeles has been measured to decay dithiothreitol at rates greater than 20 pmol min −1 µg −1 (Ntziachristos et al., 2007), both quinones seem to be unlikely contributors to redox activity of ambient particles in that locale.This is a simplistic analysis of the partitioning behaviour, assuming equilibrium conditions with bulk partitioning.Measured ambient particulate loadings of 1,4naphthoquinone (Cho et al., 2004;Eiguren-Fernandez et al., 2008a;Valavanidis et al., 2006)  100 pg m −3 , which is higher than the range we have estimated based on equilibrium partitioning, perhaps due to non-equilibrium partitioning or partitioning to non-organic aerosol species such as black carbon.We have also not explored how ambient temperature might change the outcome of such an analysis.Naphthoquinone species may become more important under conditions that favour partitioning, namely low temperatures, provided significant naphthalene emissions exist and photoreactivity is sufficiently high.This treatment is beyond the scope of this project but may be important for determining how atmospheric processes may contribute to the toxicity of particles.Furthermore, there is currently little information on the production rates of the more redox-active and less volatile 1,2-naphthoquinone and 5-hydroxy-1,4-naphthoquinone products for the purpose of atmospheric modelling, and there remains, at minimum, about 70 % of the redox activity with unknown partitioning behaviour that we could not account for based on the three naphthoquinone species measured.The missing activity could arise from highly functionalised products, which are likely to have low vapour pressures and could contribute to redox cycling under ambient particle loading; structural activity relationships predict the vapour pressure of 5-hydroxy-1,4-naphthoquinone to be about an order of magnitude lower than that of 1,4naphthoquinone (US EPA, 2013).In terms of importance, oxidation of phenanthrene to produce phenanthrenequinone (e.g., Wang et al., 2007;Lee and Lane, 2010), which is of much lower volatility than either 1,2-or 1,4-naphthoquinone and is more highly redox-active, may be a greater source of particle-phase redox-active organic species and has been shown to be an atmospherically relevant species (Eiguren-Fernandez et al., 2008b).Certainly, previous apportioning of DTT activity to chemical species has shown phenan-threnequinone to be a much more substantial contributor to redox activity than 1,2-and 1,4-naphthoquinone (Chung et al., 2006;Charrier and Anastasio, 2012).The influence of naphthalene oxidation on redox activity in relation to other redox active species such as 9,10-phenanthrenequinone, transition metals, black carbon, and humic-like substances -particularly in warmer climates when partitioning favours the gas phase for small naphthoquinones -may ultimately depend on the volatility of the unknown redox-active species generated through its photooxidation.
by Copernicus Publications on behalf of the European Geosciences Union.

Fig. 1 .
Fig. 1.A general redox cycle for 1,2-naphthoquinone.A generic reducing agent (Red) transfers a single election to 1,2naphthoquinone to generate the semiquinone radical.The quinone is regenerated when the electron is transferred to oxygen, generating superoxide.

Fig. 3 .
Fig. 3. Average mass spectrum of naphthalene SOA coloured by fragment family.
Fig. 9.Representative experimental data for a naphthalene generation experiment.AMS organics (green) and m/z 158 (orange) were corrected for a collection efficiency of 0.5.The PTR-MS signal for protonated naphthoquinone at m/z 159 (blue) was normalized to the reagent ion count.The gas-particle portioning coefficient K p (black triangles) was calculated for each time point using AMS and PTR-MS data calibrated to filter and XAD measurements of 1,4naphthoquinone, respectively.

Fig. 10 .
Fig.10.Calculated quinone particulate mass loadings and mass-and volume-normalised DTT decay rates as a function of organic particle mass loading.Calculations are performed using a total mass concentration of 1 ng m −3 of quinone.Shaded areas represent the uncertainties as discussed in the text.

Table 2 .
Average elemental composition of particles at the end of the photooxidation portion of the experiment.

Table 3 .
Summary of DTT and quinone analysis for naphthalene oxidation experiments.

Table 4 .
Summary of AMS and PTR-MS measurements for naphthalene oxidation experiments.
Calculated from first-order decay of naphthalene (k OH = 2.2 × 10 −11 molec −1 cm −3 , Sasaki et al., 1997); relative standard deviation of fit was typically less than ±1 %. b Uncertainty based on ±10 % relative standard deviation of PTR-MS calibration for naphthalene.c Values averaged from 10 min of data prior to lamp shutoff time.d Uncertainty based on standard deviation of averaged values 10 min prior to lamp shutoff.e Uncertainty based on a ±25 % relative standard deviation of the calibration slope for calculation of 1,4-naphthoquinone from AMS m/z 158.
a f Uncertainty estimated at ±20 %.