Residential wood combustion is a notable source of atmospheric aerosols, particularly in regions with moderate to cold climate, as it is a common heating method. For example, during the winter, residential wood combustion was found to contribute between 17 and 49 % to submicron organic aerosol mass at various rural and urban sites throughout central Europe (Lanz et al., 2010); between 9 and 64 % to total particulate carbon at six rural sites in Portugal, France, Germany, Austria and Hungary (Gelencsér et al., 2007); an average of ∼ 41 % to organic carbon in Fresno (USA) (Gorin et al., 2006); an average of 79 % to organic carbon in Prague (Czech Republic) (Saarikoski et al., 2008); 13–15 % to primary submicron aerosol and up to 66 % to total submicron aerosol in Paris (France) (Crippa et al., 2013; Petit et al., 2014); 31–83 % to PM1 in northern Sweden (Krecl et al., 2008); and 30–60 % to organic carbon in various European alpine locations (Herich et al., 2014). Globally, it is estimated that approximately 3 billion people burn biomass or coal for residential heating and cooking needs (IPCC, 2007).

Although wood combustion is known to contribute to the global aerosol burden, large uncertainties in aerosol composition and quantification remain. Aerosol characterization is important for ambient source apportionment and for understanding impacts on health (Naeher et al., 2007; Mauderly and Chow, 2008; Bølling et al., 2009), air quality (Finlayson-Pitts and Pitts, 2000) and climate (IPCC, 2013). Direct particulate emissions from wood combustion are a complex mixture of organic compounds, inorganic compounds, elemental and/or black carbon, and metals (Fine et al., 2001, 2002a, 2004a; Hedberg et al., 2002; Johansson et al., 2004; Schmidl et al., 2008), making characterization difficult. Previous studies have shown that organics often constitute more than 50 % of particulate emissions (Schauer et al., 2001; Fine et al., 2002b; Hedberg et al., 2002; Schmidl et al., 2008; Elsasser et al., 2013), making this an important class of compounds to characterize. However, characterization is complicated by burn variability, both during a single fire and between individual fires (Weimer et al., 2008; Heringa et al., 2012; Elsasser et al., 2013; Eriksson et al., 2014). Burn parameters, such as wood type, wood loading and appliance type, can also influence emission composition greatly (McDonald et al., 2000; Fine et al., 2002b; Jordan and Seen, 2005; Weimer et al., 2008; Pettersson et al., 2011; Orasche et al., 2012, 2013; Elsasser et al., 2013; Eriksson et al., 2014).

Gases emitted together with primary organic aerosol (POA) can undergo oxidation in the atmosphere to form lower-volatility compounds, which condense to form secondary organic aerosols (SOA) (Kroll and Seinfeld, 2008). However, most previous laboratory studies of residential wood burning have only characterized primary emissions after undergoing varying degrees of cooling and dilution (McDonald et al., 2000; Fine et al., 2001, 2002a, b, 2004a, b; Schauer et al., 2001; Hedberg et al., 2002; Johansson et al., 2004; Schmidl et al., 2008; Weimer et al., 2008; Orasche et al., 2012, 2013; Elsasser et al., 2013; Eriksson et al., 2014). The few laboratory studies investigating SOA formation from residential wood burning have shown that it can be substantial; aged OA was 1.8–5.3 times that of POA for log wood burners during stable burning conditions (Grieshop et al., 2009a, b; Hennigan et al., 2010; Heringa et al., 2011). Also, phenol, syringol and guaiacol, which have been identified in primary biomass burning emissions, form SOA in laboratory studies (Yee et al., 2013). There is evidence of SOA formation from residential wood burning in ambient measurements as well (Iinuma et al., 2010; Mohr et al., 2013). For example, methyl-nitrocatechols, reaction products from the oxidation of m-cresol, present in primary wood smoke, have been measured in significant quantities in rural Germany (Iinuma et al., 2010). The extent to which wood combustion SOA is included in ambient measurements is unclear and more work is needed to characterize SOA quantities and composition. For example, 59 % of the carbon in semi- and low-volatility oxygenated organic aerosol (largely secondary) measured during winter in Barcelona consists of non-fossil carbon, suggesting this is from sources such as biomass burning and cooking (Mohr et al., 2012). Improved understanding of wood burning SOA is also needed to improve modeling results (Simpson et al., 2007).

PAHs are formed during incomplete combustion, emitted during wood burning (McDonald et al., 2000; Schauer et al., 2001; Hedberg et al., 2002; Fine et al., 2004b; Jordan and Seen, 2005; Fitzpatrick et al., 2007; Bari et al., 2011; Pettersson et al., 2011; Eriksson et al., 2014), and of particular interest due to their deleterious effects on health (Kim et al., 2013). Functionalized PAHs, such as oxygenated or nitrogenated compounds, can be emitted directly during burning or can form in the atmosphere from the oxidation of parent PAHs, and these heteroatom derivatives often have higher toxicity than their PAH analogues (Yu, 2002; Fu et al., 2012). Recently, increased PAH emissions were observed under highly loaded burner conditions (Elsasser et al., 2013; Orasche et al., 2013; Eriksson et al., 2014). However, no work has been published on the effect of burner fuel load on SOA.

Here, we present results from the first smog chamber investigation of the impact of wood load on the quantities and composition of primary and secondary residential wood combustion products. Online measurements were made with a high-resolution time-of-flight aerosol mass spectrometer (AMS) and compared to results from offline filter analysis, with a particular focus on PAH emissions and their evolution with aging.

Six wood combustion experiments using beech wood with a moisture content of 10 ± 2 % (mean ±2 sample standard deviations, 2 SD) were conducted in a 27 m3 smog chamber (Paulsen et al., 2005) to investigate the effects of wood loading on primary emissions and secondary products. Three of the experiments (1–3, Table 1) were conducted with a starting wood load of 2.21 ± 0.09 kg (three logs without bark, four pieces of kindling and three fire starters comprised of pine wood shavings, paraffin and natural resin) and three experiments (4–6, Table 1) were conducted with a starting wood load of 7.4 ± 0.2 kg (nine logs without bark, eight pieces of kindling and four fire starters) in the ∼ 0.037 m3 burner combustion chamber of a modern log wood burner (Avant, Attika) (Fig. S1 in the Supplement). Hereafter, these two cases are referred to as “average load” and “high load”, respectively. Prior to each burn, ash and residue from the previous burn were removed. Each burn began with ignition of the fire starters using a hand-held lighter and immediately closing the burner door. The damper was left completely open to allow maximum air flow into the combustion chamber until the logs caught fire (∼ 10–15 min), after which the air flow was reduced.

Before each experiment, the chamber was cleaned by injecting O3 for 6 to 8 h and irradiating with a set of 80 UV lights (100 W, Cleo Performance, Philips) (Platt et al., 2013) for at least 10 h while flushing with pure air (120 L min-1, 737-250 series, AADCO Instruments Inc.). The chamber was then flushed with pure air in the dark for at least 20 h. After cleaning, the chamber was partially filled with humidified pure air. Wood was combusted as described above and emissions were sampled from the chimney, diluted using two ejector dilutors in parallel (DI-1000, Dekati Ltd.) and injected into the chamber. Lines from the chimney to the smog chamber, the ejection diluters and the dilution air (equal mixture of air purified from 737-250 series, AADCO Instruments Inc. and 250 series, AADCO Instruments Inc.) were heated to 473 K to reduce line losses of semi-volatile compounds due to condensation of the hot emissions. The total dilution ratios of the raw emissions after the dilutors ranged from 13.6 to 15. Emissions underwent another dilution of roughly a factor of 5–20, depending on the experiment, when injected into the chamber. The average temperature and relative humidity in the chamber after emission injection was 294.0  ± 0.5 K and 60 ± 5 %, respectively.

Emission injection into the chamber began at least 15 min after ignition to ensure that the emissions were not influenced by fire starters. Injection continued until either flames were no longer visible or the desired mass concentration was reached in the chamber, which ranged from 41 to 82 min for experiments 1–5. Experiment 6 was performed in a similar manner, except at the end of the flaming phase, the injection into the chamber was stopped and two batches of two additional logs each were added to the burner. The second and third injection into the chamber began after the new logs caught fire (approximately 2–4 min after addition) for a total injection period of 113 min.

A burn proceeds through phases: a starting phase, at least one stable flaming phase and a smoldering phase, all of which can have different chemical profiles. Emissions are typically more variable during the start and smoldering phase than the flaming stage. To isolate the effect of fuel loading, only emissions from the flaming phase were sampled. In this study, the average modified combustion efficiency (MCE), defined as the ratio of CO2(g) to the sum of CO2(g) and CO(g) (Ward and Radke, 1993), of the injected emissions was greater than 0.90 for each experiment (Table 1), indicating injected emissions were dominated by the flaming phase (Lee et al., 2010).

After injection, the gaseous and particulate emissions in the chamber were characterized using a variety of online and offline techniques described below. Following approximately 2 h of primary emissions measurements, the toxicological responses of human epithelia cells to the particle emissions were investigated (presented in a future publication). During the cell exposures, particles were sampled at ∼ 80 L min-1 from the chamber. This necessitated refilling the partially empty chamber during the exposure with humidified pure air.

After exposing the cells to primary emissions and the concurrent refilling of the chamber, a 9-times deuterated butanol sample (butanol-D9, 98 %, Cambridge Isotope Laboratories) was injected into the chamber. This compound was monitored ([M-H2O+H]+ fragment, m/z 66) using a proton transfer reaction time-of-flight mass spectrometer (PTR-MS, [H3O]+ reagent ion, Ionicon Analytik GmbH) to calculate OH exposures in the chamber, using a rate constant of 3.4 × 10-12 cm3 molec-1 s-1 (Barmet et al., 2012). The PTR-MS was not used during experiments 3 and 6.

In addition to the butanol-D9 injection, a continuous injection of nitrous acid (HONO), which dissociates upon irradiation (λ < 400 nm) to form hydroxyl radical (OH), began 30–35 min prior to the start of photooxidation. The HONO was generated and injected by bubbling pure air (2.5 L min-1) through a mixture of H2SO4 (95–97 %, Merck) in water (10 mM, 18.0 MΩ cm, Milli-Q, Reference A+) and NaNO2 in water (1 mM experiments 1, 2, 4, 5; 3 mM experiments 3, 6; ≥ 99.0 %, Fluka) and passing through a particle filter into the chamber (Taira and Kanda, 1990; Pfaffenberger et al., 2013). Finally, the emissions were aged with UV light for 4.5–16 h, corresponding to maximum OH exposures of ∼ (2–5) × 107 molec cm-3 h in experiments 1, 2, 4 and 5. A significant increase in OA upon photooxidation was observed during all experiments, and after the maximum OA concentration was reached, a second cell exposure and chamber refill was performed.

Primary and secondary particle- and gas-phase products were characterized using a variety of online and offline techniques. In the gas phase, CO2, CO and CH4 (Picarro Inc.); O3 (S300 ozone analyzer, Environics); total hydrocarbons (THC, model VE 7 THC analyzer with flame ionization detector, J.U.M.); and NO, NO2 and NOx (trace level 42C, Thermo Environmental Instruments with a photocatalytic converter and 9841A NOx analyzer, Monitor Labs) were measured.

Non-refractory particle chemical composition and size were measured using a high-resolution (HR) time-of-flight AMS (Aerodyne Research Inc.) (DeCarlo et al., 2006) operated in V mode with a 2.5 µm inlet lens (Williams et al., 2013). Two Aethalometers each measured black carbon mass concentrations at seven wavelengths (Magee Scientific Aethalometer model A33, Aerosol d.o.o.). A thermal desorber comprised of a 50 cm heating section held at 423 K followed by a 50 cm denuder section was located directly upstream of one Aethalometer to volatilize and remove organic species (Burtscher et al., 2001). A condensation particle counter (CPC, 3025A TSI) measured total particle number concentrations and a scanning mobility particle sizer (SMPS, CPC 3022, TSI, and custom built DMA with a length of 0.44 m) measured particle size distributions. Particles were dried (Nafion, Perma Pure LLC) upstream of the AMS, Aethalometers, SMPS and CPC.

Losses in the thermal desorber were determined by nebulizing NaCl (≥ 99.5 %, Fluka) in water and passing through the thermal desorber. Size distributions were measured using an SMPS before and after the thermal desorber. It was determined that 24 % of the mass is lost in the thermal desorber at a mobility diameter of 100 nm and 9 % of the mass is lost at a mobility diameter of 200 nm. The average mobility diameter of the particles after injection ranged from 50 to 120 nm and losses were accounted for using the data collected at a mobility diameter of 100 nm.

Additionally, particles were collected on filters (47 mm Tissuquartz, Pall Corporation, 26 L min-1 for 30–32 min) after injection of the primary emissions and after observation of the maximum OA signal during aging. Prior to deposition on the filter, particles were sampled through a charcoal denuder to remove organic gases. Filters were immediately stored at 253 K until analysis. Filters were analyzed by in situ-derivatization thermal desorption gas chromatography time-of-flight mass spectrometry (IDTD-GC-TOF-MS) as described by Orasche et al. (2011). Briefly, a small portion of each filter was spiked with internal standards (isotopically labeled compounds; Orasche et al., 2011) and N-methyl-N-(trimethylsilyl)-trifluoroacetamide (MSTFA) was added before and subsequently during a thermal extraction period by introduction into a carrier gas stream. Thermally desorbed and derivatized molecules were transferred onto the cold GC column and the GC-MS run was started with pure carrier gas. Filter analysis from experiment 2 is not available.

For all experiments, primary particulate emissions were a mixture of organic and inorganic compounds and black carbon (Table 1). Emission factors of various species in the chamber are given in Table 1 and corresponding concentrations in the chamber are presented in Table S1 in the Supplement. Organic contribution to the total particulate matter (PM) mass was substantial in all experiments, ranging from 22 to 87 % (Table 1), in agreement with the wide range reported in previous studies (Schauer et al., 2001; Hedberg et al., 2002; Fine et al., 2004b; Schmidl et al., 2008; Elsasser et al., 2013). Total PM concentrations were determined by summing BC and all species measured by AMS. Non-carbon elemental species were not measured during this study, and although they are known to be emitted from wood combustion, they are not expected to greatly influence the results, as previous studies have typically found contributions of less than 5 % to the fine PM mass (Schauer et al., 2001; Hedberg et al., 2002; Fine et al., 2004b).

The total PM emission factor was similar for each experiment, except experiment 1, for which it was over 3 times greater than the other burns due to increased organic emissions (Table 1). There were no obvious reasons why the total PM emission factor in experiment 1 was considerably higher than the other experiments. Although the MCEs were similar for all burns (Table 1), this illustrates the real-world variability of wood combustion emissions and the large range of possible emission factors from the same burner and operator. Variability is also observed in previous comparisons of high- and average-load burns. Orasche et al. (2013) found total PM emission factors were a factor of about 1.4 higher for high-load flaming conditions than for average loads, whereas Elsasser et al. (2013) found 4-times-higher PM emissions generated under high-load conditions compared to average-load conditions. However, it is noted the Elsasser et al. (2013) result was determined from a single measurement.

The OM : BC ratio varied between burns, ranging between 0.33 and 9.4, with no significant dependence on wood loading (Table 1). These values are within the large range of measurements reported in the literature, although generally lower than ambient measurements (Gianini et al., 2012) and some direct emission studies (e.g., McDonald et al., 2000; Fine et al., 2001, 2002a, 2004a, b; Schauer et al., 2001; Schmidl et al., 2008). Similar values to those found in this study were reported in measurements from previous chamber experiments (Grieshop et al., 2009b; Heringa et al., 2011) and other direct emission studies for flaming conditions with modern small-scale wood burning appliances (Lamberg et al., 2011; Eriksson et al., 2014). Variability in literature OM : BC values arises not only from burn variability but also from measurement/analysis methods. The burner used here is relatively new and expected to burn more efficiently (i.e., lower OM : BC) than burners with older technologies. Also, potential difficulties in separating primary and secondary signal in ambient source apportionment studies can result in incorrectly apportioned primary and secondary signals (Lanz et al., 2010). As OM : BC is typically higher for aged aerosol compared to primary, this results in overestimated OM : BC ratios. In addition, previous direct emission studies often used lower dilution ratios than used in this study and collected material on quartz fiber filters without the use of an upstream denuder (e.g., McDonald et al., 2000; Fine et al., 2001, 2002a, 2004a, b; Schauer et al., 2001), both of which can result in substantial positive artifacts, as discussed previously (Subramanian et al., 2004). This would result in overestimated OM : BC ratios.

AMS and offline filter measurements provide data on the effect of fuel loading on particulate composition. Figure 1 shows the average AMS mass spectrum of the primary emissions from each experiment. The mass spectral signal is separated into different chemical classes (i.e., organic, PAH, nitrate, sulfate, ammonium and chloride) as described in the data analysis section. There were significantly higher PAH contributions to the total organic signal for the high-load experiments compared to the average-load experiments (Fig. 1, Table 1). For the high-load experiments, PAHs contributed 14–17 % to the total organic signal, compared to only 3.4–4.7 % for the average-load experiments (Table 1).

Higher PAH fractions of the total organic mass with higher wood loading are in agreement with Elsasser et al. (2013), who observed an increased signal at higher m/z values in AMS mass spectra under high-load conditions, compared to average-load conditions, and attributed this to PAHs. Eriksson et al. (2014) measured PAH contributions to the total OA of 6–14 % under hot, air-starved conditions (generated using higher wood loadings) compared to only 2–4 % under nominal burn conditions during the flaming stage using AMS. Also, previous GC-MS analysis of primary filter samples shows a contribution of PAHs to total PM mass of about 13 % during high-load burning, compared to less than 1 % during average-load burning of beech wood (Orasche et al., 2013).

As filter-based methods can be subject to considerable artifacts and the analysis method used here for PAH quantification from HR-AMS data is novel, results from filter and AMS PAH analysis were compared. The total primary PAH emission factors measured from the filters were 0.88 mg kg-1 (experiment 3) and 4.2 mg kg-1 (experiment 1) for average-load experiments and ranged from 2.2 to 4.8 mg kg-1 for high-load experiments (Table 3). The primary PAH emission factors measured by AMS ranged from 2.8 to 27 mg kg-1 for average-load experiments and 5.4 to 13 mg kg-1 for high-load experiments. AMS PAH emission factors were higher than those determined from the filter analysis, but within a factor of 2–3 for all experiments, except experiment 1, which differed by a factor of 6. Due to uncertainties associated with both analysis methods, agreement within a factor of 2–3 is satisfactory.

As a further comparison between the filter and AMS PAH analysis for primary aerosol, compounds with molecular weights corresponding to base ions measured by the AMS above the background level (i.e., 202, 226, 230, 252 and 276 including all associated ions) were compared (Fig. S4 in the Supplement). With the uncertainties associated with both methods, good agreement was observed for molecular weights of 230, 252 and 276, with AMS : filter ratios ranging from ∼ 0.9 to 2 for the primary emissions. However, for MW 202 and 226, the filter measurements were a factor of ∼ 6–15 lower than the AMS measurements. The discrepancies between the filter and AMS measurements were highest for the highest volatility compounds compared, suggesting negative artifacts during the filter sampling, attributed to the use of the denuder.

Discrepancies between the filter and AMS measurements increased with aging (Fig. S4 in the Supplement), particularly for the higher volatility compounds (i.e., MW 202 and 226). The AMS : filter ratios for MW 230, 252 and 276 were 0.9–4, whereas AMS : filter ratios increased drastically for MW 202 and 262 to 13–43. The increased AMS : filter ratios during aging could be due to the formation of oxygenated PAHs, which are more likely to fragment in the AMS than unfunctionalized compounds, and could contribute to the base ion and/or associated ions. Thus, it is not clear whether the AMS signals during aging correspond solely to the compounds identified in the filter samples. During aging, gas-phase compounds are oxidized and partition to the particles and the AMS : filter ratios for the lower volatility compounds, which are already predominately in the particle phase, would not change as much during aging as higher volatility compounds, as observed.

The primary organic emission factors for the average-load experiments were a factor of 0.8–18 times higher than for the high-loading experiments. Although there were higher PAH fractions of the organic mass in the high load cases, there were generally higher primary organic emission factors for the average-load burns compared to the higher load burns, resulting in comparable PAH emission factors. The PAH emission factors in this study are considerably lower than the 196 mg kg-1 found by Orasche et al. (2013) under stable burning conditions of beech wood in an overloaded burner; however, total PM mass determined by Orasche et al. (2013) by weighing dried filter samples was also considerably higher. The higher emission factors observed by Orasche et al. (2013) may be due to differences in burner technologies or sampling methods. The emission dilution ratio affects the partitioning of semi-volatile species, with higher dilution ratios shifting the distribution to the gas phase and thus decreasing particle emission factors (Lipsky and Robinson, 2005). The dilution ratios were only a factor of 4 in Orasche et al. (2013), compared to ∼ 70–300 in this study. Also, their lack of a denuder upstream of the filter may have resulted in positive artifacts (Subramanian et al., 2004; Orasche et al., 2013).

Due to the semi-volatile nature of many of the PAHs, quantification is also subject to temperature (Boman et al., 2005; Hytonen et al., 2009). While 74 % of pyrene is in the particle phase at 294 K, 99.8 % of pyrene is expected to be in the particulate phase at 263 K (vapor pressure of 7.6 × 10-7 Pa at 263 K, activity coefficient in the organic mass of 1 and a total organic aerosol mass concentration of 30 µg m-3) (Pankow, 1994; Donahue et al., 2006). These experiments were conducted at 294 K; however, most residential wood combustion occurs when outdoor temperatures are well below this. Thus, if these experiments had been conducted at lower temperatures, particulate PAH emission factors would be higher.

The ion C2H4O2+ is used as a biomass burning tracer in ambient AMS data. Interestingly, the fraction of C2H4O2+ relative to the total primary OA (fC2H4O2+) was higher in the average-load experiments, ranging from 0.050 to 0.094, compared to the high load experiments, ranging from 0.012 to 0.029. As summarized by Fitzpatrick et al. (2007), lower burn temperatures favor the formation of products from the pyrolysis of lignin and carbohydrates, including levoglucosan. As the burn temperature increases, the relative contribution of these compounds decreases and products such as polycyclic oxygenates and PAHs dominate. The higher fC2H4O2+ and lower fPAHs in the average-load experiments suggests that the variability in fC2H4O2+ was due to lower burn temperatures in the average-load experiments, likely due to different fuel / air ratios.

The chemical differences in organic emissions from high- and average-load cases lead to differences in the evolution of these emissions with aging, which has not previously been investigated. Figure 2 shows the evolution of the components measured by AMS, corrected for wall losses and dilution, with photochemical aging. In all experiments, a significant increase in organic mass was observed with photochemistry. At the peak during aging, or at the end of the experiment for experiment 6, OA increased by a factor of 2.3–3.4 for high loads, whereas by only a factor of 1.4–1.8 for average loads. There was also an increase in PAH concentration with initial aging followed by a decrease with extended aging and eventual stabilization. As shown in Fig. S5 in the Supplement, PAHs above m/z 200 were stable or decreased with the initiation of photochemistry, whereas PAH signals at lower molecular weights increased. As discussed previously, the PAH signals in the lower m/z range are likely from PAHs that have undergone fragmentation in the AMS vaporization–ionization process. Functionalized PAHs are more likely to undergo fragmentation than non-functionalized PAHs (McLafferty and Turecek, 1993). Increases in these lower m/z PAH signals with aging are likely due to the formation of functionalized PAHs with sufficiently low volatility to partition to the particles.

Chemical composition provides additional insight into the evolution of organic aerosol with aging. Although SOA formation was observed in all experiments, the chemical composition of the aged aerosol was different between the high-load and average-load experiments (Fig. 3). Figure 3 shows HR-AMS mass spectra of organic ions categorized by composition. As discussed previously, mass above m/z 200 is not fully included; however, the unaccounted mass is less than 4 % of the total organic mass. Mass spectra of the entire measured m/z range and all compound classes are shown in Fig. S6 in the Supplement. Experiments 1, 2, 4 and 5 were compared after an OH exposure of 1.6 × 107 molec cm-3 h. The OH exposures were not measured during experiments 3 and 6, and the mass spectra in Fig. 3 correspond to 3 h aging, which is the average time after lights-on of experiments 1, 2, 4 and 5 corresponding to an OH exposure of 1.6 × 107 molec cm-3 h.

In all mass spectra, the largest peak is m/z 44, largely from [CO2]+, formed from the decarboxylation of organic acids. For the high-load experiments, there are notable aromatic signature ions separated by 12, 14 or 16 Da, starting at [C6H5]+ (m/z 77) (e.g., [C7H7]+ (91), [C7H5O]+ (105), [C7H5O2]+ (121), [C8H5O2]+ (133)) (Fig. 3). Although unfunctionalized PAHs undergo minimal fragmentation in the AMS, functionalized aromatic compounds, including PAHs, are known to fragment when ionized by electron impact to produce other characteristic peaks in the lower m/z region as well (i.e., < m/z 77). Common fragment ions from PAHs containing electronegative substituents are at m/z 38, 39, 50, 51, 63–65 and 74–76 (McLafferty and Turecek, 1993). These peaks become more dominant with aging in the high-load mass spectra, suggesting an increase in functionalized PAHs with aging (Fig. S7 in the Supplement) (McLafferty and Turecek, 1993). There are also considerable quantities of oxygenated ions at higher m/z values in the high-load experiments (Fig. 3). For example, the fraction of total oxygenated peaks, not including [CO2]+, at or above m/z 80 is 0.45 for high-load experiment 4, compared to only 0.15–0.20 for average-load experiment 2. This fraction is also higher than the < 0.15 observed during an α-pinene photooxidation experiment (unpublished data) (Fig. S8 in the Supplement).

From the filter analysis, the largest increases with aging were observed for functionalized naphthalene compounds (i.e., 2-naphthoic acid, 1(2H)-acenaphthylenone, 1,8-naphthalic anhydride and 2-hydroxynaphthalene), particularly 2-naphthoic acid, which increased by a factor of ∼ 120 (Fig. 4b, Table 3). In addition to the large increases with aging, 2-naphthoic acid and 1,8-naphthalic anhydride contributed the most to the total PAH signal. This further suggests that the differences in the aged AMS mass spectra between the high and average loads are largely due to the presence of PAHs which have undergone oxidation in the gas phase to form compounds with sufficiently low volatility to partition to the particle phase. The half-life of naphthalene with respect to OH (average [OH] from experiments 4 and 5 = 6.3 × 106 molec cm-3) in the chamber was 71 min, and half-lives of methylated naphthalene are even shorter (Atkinson and Aschmann, 1986), which are within the time frame observed for the increase in particulate PAH signal with aging. PTR-MS data show a decrease in m/z 129, likely dominated by naphthalene, during aging, with ∼ 50 % lost within the first 65–70 min of aging and ∼ 75 % lost by the end of the experiment (Fig. S9 in the Supplement). General PAH oxidation pathways are discussed in detail elsewhere (Finlayson-Pitts and Pitts, 2000), as well as the formation pathways and identification of naphthalene oxidation products (Kautzman et al., 2010).

The Van Krevelen diagram provides information on the bulk organic aerosol composition and compositional changes with aging (Heald et al., 2010). Figure 5a shows that the primary emissions from the average- and high-load experiments occupy different regions of the plot, further illustrating the bulk compositional differences between the different loading conditions. The primary emissions from the high-load experiments had a lower H : C (1.1 ± 0.1) compared to the average-load experiments (1.47 ± 0.09), due to the larger presence of PAHs (Fig. 5a). For example, [C16H10]+, the highest intensity primary PAH ion identified by the AMS, has an H : C of only 0.625. Primary O : C was similar for the high-load experiments (0.3 ± 0.1) compared to the average load (0.4 ± 0.1). With aging, all experiments showed an initial sharper decrease in H : C, followed by a more gradual decrease. Throughout aging, O : C steadily increased.

The evolution of the fraction of [CO2]+ (fCO2+) relative to the fraction of [C2H3O]+ (fC2H3O+) to the total organic signal also provides insight into changes in the chemical composition of the organic material with aging (Ng et al., 2011). In the AMS, [CO2]+ is formed from the decarboxylation of organic acids during vaporization, whereas [C2H3O]+ largely comes from fragmentation of carbonyl-containing compounds, as well as alcohols and peroxides. Ng et al. (2011) found that the majority of atmospherically relevant SOA measurements fall within a certain space on the f44 vs. f43 plot. Wood combustion emissions typically lie near the left-hand side of this space, as observed here (Fig. 5b) (Heringa et al., 2011; Heringa et al., 2012).

Within the first 20 min of aging, in which H : C decreased considerably, there was also an increase in fC2H3O+ for all experiments (Fig. 5b), suggesting an increased fraction of carbonyl, alcohol- and/or peroxide-containing compounds. For the high-load experiments, the majority of compounds measured from the filters that increase with aging contain at least one carbonyl group. During aging, fCO2+ continued to increase, whereas fC2H3O+ tended to decrease, supporting the more dominant formation of acids with increased aging.

The increase in functionalized naphthalene species measured on the filters with aging suggests notable concentrations of naphthalene were in the gas phase prior to aging during the high-load experiments. Also, phenols, syringol and guaiacol are considerable fractions of wood burning emissions. For oxidation of naphthalene, phenols, syringol and guaiacol, Chhabra et al. (2011) observed near-vertical movement in f44 vs f43 space, similar to all experiments here. However, in the Van Krevelen diagram, Chhabra et al. (2011) observed movement with a slope close to -1 during oxidation of phenols, syringol and guaiacol, whereas a smaller slope was observed during naphthalene oxidation. The slope after initial oxidation of the high-load experiments was -0.19 ± 0.05, smaller than for the average-load experiments at -0.4 ± 0.2. The smaller slope in the high-load experiments may be due to the higher fraction of PAHs, particularly oxidized naphthalene species, compared to the average-load experiments. Although there are clear differences between the high- and average-load experiments in the Van Krevelen diagram, it is likely difficult to use these differences to determine wood burning or PAH contribution in ambient measurements, even when dominated by wood burning, due to variability during burns and varying burner parameters and technologies.

High wood loads result in an increased fraction of PAHs to the total organic aerosol compared to average wood loads. With aging, AMS and filter data indicate that gas-phase compounds, including PAHs, undergo oxidation to form lower volatility products which partition to the particulate phase. As functionalized PAHs often have more deleterious effects on health than their parent analogues (Yu, 2002; Fu et al., 2012), these findings have an impact on toxicological implications. Even when gas-phase PAH emissions are relatively low, their contribution to SOA can be considerable. For example, using yields from the oxidation of PAHs in the laboratory, Chan et al. (2009) estimated that gas-phase PAHs, despite being only half the concentration of light aromatics, produce 4 times more SOA during the first 12 h of oxidation of emissions from the burning of pine wood under normal conditions. The toxicological effects on human health of the primary emissions from average and higher load burning and the changes in chemical composition with aging will be detailed in a future publication.