2.1 Experimental conditions

Experiments were conducted in the ILMARI laboratory of the University of Eastern Finland (http://www.uef.fi/ilmari, last access: 20 May 2020) with the experimental setup shown in Fig. 1. The two combustion appliances used were modern stoves with improved air intakes. First, a heat-storing masonry heater (Hiisi 4, Tulikivi Ltd., Finland) representing the typical modern logwood combustion technology utilized in northern Europe, was fired with spruce logwood. The combustion procedure in the masonry heater consisted of three 2.5 kg batches (35 min combustion time) of spruce logwood representing kiln-dried fuel (5 % moisture content, S-5 %), after which there was a 25 min char burning period prior to two 45 min batches with moist (22 % moisture content, S-22 %) spruce logwood. Second, a non-heat-retaining chimney stove (Aduro 9.3, Denmark) representing central European modern logwood stoves was fired with both beech and spruce logs. The chimney stove experiments consisted of five 2 kg batches (combustion time 40–55 min) of beech logwood (17 % H₂O, B-17 %), followed by two 50 min batches of S-22 %. Beech was used in these experiments because it is the most common firewood used in central Europe, while spruce is used both in northern and central Europe. For ignition, 150 g of dry kindling was placed on the top of the first batch in the cold furnace. Each batch was divided into three parts: ignition, flaming, and burnout phase. The ignition phase was determined to last from the beginning of the batch to the moment of batch maximum flue gas CO₂ concentration, and the burnout phase began from the moment when the CO concentration started to elevate and remained at a high level until the end of the batch (Fig. 2). The modified combustion efficiency (MCE) as a function of time was calculated from primary flue-gas CO₂ and CO concentrations as $\mathrm{\Delta }{\mathrm{CO}}_{\mathrm{2}}/(\mathrm{\Delta }{\mathrm{CO}}_{\mathrm{2}}+\mathrm{\Delta }\mathrm{CO})$.

The exhaust was sampled from the stack with a 10 µm precut cyclone. The sample was diluted with a combination of a porous tube diluter and an ejector diluter (Dekati FPS ejector, Finland) and had a dilution ratio (DR) of 40–150 (Table 1) when fed to the PEAR OFR. In the chimney stove experiments, an additional ejector diluter with a DR of 8 (Dekati DI-1000, Finland) was placed before the secondary online aerosol instruments. Compared to the no-ageing experiments, a higher DR was selected for the ageing experiments in response to the expected increase in particulate matter in the PEAR OFR during photochemical ageing. Measured concentrations were corrected for the dilution and normalized to stoichiometric dry flue-gas by multiplication with the stoichiometric correction factor (SCF) of Eq. (1) based on the background CO₂ concentration (CO_2,bg), the CO₂ concentration in the diluted sample downstream of the PEAR OFR (CO_2,sec), and the fact that the dry flue gas of wood combustion with no excess air contains 20.2 % CO₂.

In addition, the diluted concentrations were normalized to 13 % flue-gas oxygen content.

2.2 Use of the PEAR OFR

The PEAR OFR (Ihalainen et al., 2019) was used to continuously age the sample stream. In the setup, the extent of photon flux and consequential photochemical ageing were controlled by adjusting the voltage of the 254 nm UV lamps. The total flow through the PEAR OFR was 60 L min⁻¹, and the residence time within the PEAR was 139 s. In addition to the 55 L min⁻¹ sample flow, humified and purified air was introduced to the PEAR OFR to obtain a relative humidity of 45±5 %. In the PEAR OFR, the photolysis of the ozone forms hydroxyl radicals via Reactions (R1) and (R2). Thus, the extent of photochemical ageing in the PEAR OFR depends on the photon flux inside the reactor and the introduction of OH precursors, namely the externally fed O₃ and H₂O vapours.

In the ageing experiments, ozone and butanol-d9 mixed with the main sample flow before the reactor inlet (Fig. 1) resulted in concentrations of 2–11 ppm of O₃ (Table 1) and 80–200 ppb of butanol-d9 in the PEAR OFR before initiation of photochemical ageing. The OH exposure (OH_exp) was determined continuously based on the butanol-d9 concentrations downstream of the PEAR according to the method presented by Barmet et al. (2012). The OH_exp for clean air in the PEAR prior to the exhaust input ranged from 8.2×10¹⁰ to 1.6×10¹² molec cm⁻³ s, which is equivalent to 1–18 d in an ambient atmosphere (eqv. d) with an estimated 24 h average global OH concentration of 1×10⁶ molec cm⁻³ (Prinn et al., 2005). The alternate reaction pathways in the PEAR OFR, namely exposure to photolysis (flux_254 nm,exp), excited (O(¹D)_exp), atomic oxygen (O(³P)_exp), and ozone (O_3,exp), were assessed using the OFR exposure estimation equations of Peng et al. (2016) for OFRs with 254 nm lamps. Both the OH_exp and the alternate reaction pathways were affected by the external OH reactivity (OHR_ext) (Li et al., 2015) of the gases in the sample fed to the PEAR OFR. The OHR_ext was calculated with Eq. (2) from the concentrations (c_i) of the primary gaseous compounds (Table S1) and the butanol-d9 in the PEAR OFR, with their OH reaction rate constants k_i.

Due to the differences in the emission concentrations during logwood combustion, the OHR_ext and consequently OH_exp vary also within a batch (Fig. S1). Average OHR_ext was in the range 130–1300 s⁻¹ with the highest contributions from CO, NO, and unsaturated hydrocarbons (Fig. S2).

Particulate wall losses inside the PEAR OFR were minimized by conductive stainless-steel walls, laminar flow, and a relatively small surface-to-volume ratio (2.28 m² : 139 L) (Ihalainen et al., 2019). The particulate losses were estimated to be approximately 6 % based on the loss of elemental carbon (EC) in the PEAR OFR, determined from the difference in the thermal–optical EC concentrations measured upstream and downstream of the reactor. The fate of the gaseous organic compounds capable of irreversible condensation under the present experimental conditions, that is, the low-volatility organic compounds (LVOCs), was estimated based on Palm et al. (2016) by considering three possible depletion pathways: condensation onto particles, reactions with OH radicals, and condensation onto walls. To estimate the fraction of LVOCs condensing onto particle phase, condensation sinks (Fig. S3) were calculated according to Lehtinen et al. (2003) using an average of the particle size distributions up- and downstream of the PEAR OFR. See Sect. S4.2 in the Supplement for further information on LVOC fate estimation.

2.3 Offline filter sampling and analysis

PM₁ filter samples were collected on Teflon (PTFE, Pall Corporation, P/N R2PJO47) and quartz fibre (QF; Pall Corporation, Tissuquartz) filters simultaneously from primary and secondary exhaust at a 10 L min⁻¹ flow rate, following the methodology presented by Sippula et al. (2009). A pre-impactor (Dekati PM-10 impactor) was used to separate the particles with aerodynamic diameters of less than 1 µm (PM₁). For the masonry heater, samples were collected separately from the S-5 % combustion (100 min collection time) and S-22 % combustion (90 min collection time). From the chimney stove, two sample pairs were collected from the combustion of beech: full first two batches (85 min collection time) and fourth and fifth batches excluding the last 15 min (85 min collection time). The third chimney stove collection consisted of the combustion of two full batches of moist spruce (100 min collection time). The Teflon filters were weighted before and after sample collection to determine the total PM₁ mass of the emission. The amount of organic (OC) and elemental carbon (EC) in PM₁ was determined from the QF filters by using a thermal–optical carbon analyser (Sunset Laboratory Inc.) following the protocol NIOSH5040 (NIOSH, 1999). In situ derivatization thermal desorption–gas chromatography–time-of-flight mass spectrometry (IDTD-GC-ToF-MS) (Orasche et al., 2011), was applied for targeted analysis of organic compounds in the particulate phase from the QF filters. Nonpolar and polar compounds were identified and quantified using mixtures of isotopically labelled internal standards and calibration standards; see the Supplement, Sect. S5.1, for further information on the IDTD-GC-ToF-MS analysis procedure. The majority of the targeted compounds can be classified as either semi- or intermediately volatile, meaning that under the present experimental conditions they can be partitioned into both gaseous and particulate phases.

2.4 Online gas-phase measurements

A Fourier transform infrared spectrometer (FTIR DX4000, Gasmet Technologies Inc.) was implemented on the stack to measure the amount of NO_x, CO₂, CO, and 27 OGCs (Table S1) in the fresh exhaust. The measured primary non-methane OGCs were grouped into four subgroups: alkanes, oxygenated compounds, and unsaturated and aromatic hydrocarbons. The concentrations of CO₂, O₃, NO_x, and SO₂ were monitored at the outlet of the PEAR OFR with a trace-level single-gas analysers (ABB CO₂ analyser, Siemens), and OGCs were measured by PTR-ToF-MS (PTR-TOF 8000, Ionicon Analytik, Innsbruck, Austria), with H₃O⁺ as the reagent ion and an electric field to gas number density ratio (E ∕ N) of 130. Mass calibration was done with H₃O⁺ (m∕z 21.02) and 1,3-diiodobenzene (m∕z 203.94), which was added as a calibrant for higher m∕z. The processing of the PTR-ToF-MS data was done in a manner similar to that of earlier work (Hartikainen et al., 2018). Reaction rates by Cappellin et al. (2012) were used when available; for other compounds, the reaction rate with H₃O⁺ was assumed to be 2×10⁹ cm³ s⁻¹ (Table S2).

2.5 Online particulate-phase measurements

Particle concentrations and mobility size distributions were monitored with scanning mobility particle sizers before (SMPS 3082, TSI, size range 14.6–661.2 nm) and after the PEAR OFR (SMPS 3080, TSI, size range 15.1–661.2 nm). An electrical low-pressure impactor (ELPI; 10 L min⁻¹, Dekati) measured the particle aerodynamic size distribution and the concentration of primary particles in the size range 18.6–5950 nm. The total particulate mass after PEAR OFR was measured with a tapered element oscillating microbalance monitor (TEOM, Model 1405, Thermo Scientific).

The composition of submicron particulate matter after the PEAR OFR was measured by soot particle aerosol mass spectrometer (SP-HR-ToF-AMS, Aerodyne Research Inc). The SP-HR-ToF-AMS dual vaporizer mode was used, with the combination of the thermal vaporizer (600 ^∘C) and the continuous wave laser vaporizer (1064 nm) enabling the study of both nonrefractory (NR-PM) and refractory light-absorbing submicron aerosol particles (R-PM, e.g. refractory black carbon, rBC) (Onasch et al., 2012). Standard mass-based calibrations were performed for the ionization efficiency in the SP-HR-ToF-AMS using ammonium nitrate and Regal Black (Regal 400R Pigment Black, Cabot Corp.) particles (Jayne et al., 2000; Onasch et al., 2012). SP-HR-ToF-AMS was operated in V-mode from 12 to 555 m∕z, and the two vaporizer configurations were alternated every 120 s, including the particle time-of-flight (PTOF) mode (duration 20 s). The SP-HR-ToF-AMS data were analysed using the standard analysis tools SQUIRREL v1.62A and PIKA v1.22D adapted in Igor Pro 8 (Wavemetrics). The data were corrected by time-dependent background gas-phase CO₂ subtraction using the online HEPA filter measurement technique. The interactions of inorganic salts with pre-deposited carbon on the tungsten vaporizer were corrected according to Pieber et al. (2016). The CO₂-AMS interference calibration value was 0.3 % of the NO₃ concentration determined by NO₃NH₄ calibration and corrected via the fragmentation table according to Pieber et al. (2016). Before correction, the relative effect of the interference on O : C ratio was typically under 1 % and 5 % at maximum. The applied relative ionization efficiency (RIE) was 1.4, which agrees with the OA mass concentrations of biomass-burning emissions within the stated ±38 % uncertainty in AMS (Bahreini et al., 2009; Xu et al., 2018). The elemental analysis of the OA was conducted using the improved-ambient method (Canagaratna et al., 2015), and the average carbon oxidation state (OS_C) of the OA was estimated as OS${}_{\mathrm{C}}\approx \mathrm{2}\times$ O : C − H : C (Kroll et al., 2011).

The SP-HR-ToF-AMS OA mass spectra was further examined by positive matrix factorization (PMF) using the method described in Lanz et al. (2007) and Ulbrich et al. (2009). Similar methods have been used previously for example for the assessment of RWC generated POA and SOA in a chamber (Bruns et al., 2015a; Tiitta et al., 2016) and in an OFR (Bruns et al., 2015a) or for time-resolved analysis of RWC OA emission constituents (Elsasser et al., 2013; Czech et al., 2016). The PMF Evaluation Tool v.3.05 was applied, and the standard data pretreatment process was completed based on Ulbrich et al. (2009), including the application of minimum error criteria and down-weighting weak variables as well as m∕z 44 (${\mathrm{CO}}_{\mathrm{2}}^{+}$) and water-related peaks. The final four-factor PMF solution covered 98 % of the OA spectra (2.2 % residual). Additional factors did not increase the realistic physical meaning of the solution, while fewer factors were insufficient for a meaningful presentation of the data. The factor identification was confirmed by comparing the time series and mass spectra of each factor with external tracers (nitrate, sulfate, ammonium, chloride, PAH, ${\mathrm{CO}}_{\mathrm{2}}^{+}$, C₂H₃O⁺, ${\mathrm{C}}_{\mathrm{4}}{\mathrm{H}}_{\mathrm{9}}^{+}$, and ${\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{4}}{\mathrm{O}}_{\mathrm{2}}^{+}$). Furthermore, the factors were compared to logwood combustion mass spectra measured by Tiitta et al. (2016) from the ageing of spruce logwood exhaust in a smog chamber. The agreement of spectra was denoted with both a coefficient of determination (R²) and the angle between two mass spectra vectors (Kostenidou et al., 2009), where an angle of less than 15^∘ indicates a good agreement between two mass spectra.

The polycyclic aromatic hydrocarbons and other PAC in the exhaust were analysed by using the PAC molecular ions as a proxy, following the P-MIP methodology presented by Herring et al. (2015). The base molecular ions [M]⁺, their fragments ([M–H]⁺ and [M–2H]⁺), and isotopes (²H, ¹³C, ¹³C₂, ¹⁵N, ¹⁷O, and ¹⁸O) were isolated and quantified using the SP-HR-ToF-AMS high-resolution analysis software tool (version 1.22D). The ions targeted with the high-resolution (HR)-PAH analysis included those previously connected with PACs (Herring et al., 2015) and compounds typically released from RWC (Avagyan et al., 2016; Bertrand et al., 2018; Bruns et al., 2015b; Czech et al., 2018; Miersch et al., 2019). The 61 PACs considered (Table S3) were separated into five subgroups: unsubstituted PAHs (UnSubPAHs), oxygenated PACs (OPAHs), methylated PACs (MPAH), nitrogen-substituted PACs (NPAHs), and amino-PACs (APAHs).

3.1 Combustion conditions

The average modified combustion efficiency was greater than 0.97 for the masonry heater and greater than 0.95 for the chimney stove, with lower MCE occurring mainly during the char burnout periods (Fig. S4). These values are typical for modern batchwise-operated logwood appliances (Bhattu et al., 2019; Czech et al., 2018; Heringa et al., 2011). In addition to the variation within a combustion batch, there were also differences between the individual batches (Fig. 2). Notably, the combustion conditions during the first batch were distinct from later batches, because of the ignition in a cold firebox. This caused a longer ignition period (determined by the rising CO₂ concentrations), which lasted for 24±5 % (masonry heater) and 35±4 % (chimney stove) of the total duration of the batch, whereas in the later batches the firebox temperatures are higher, and the ignition shortened to 9±3 % (masonry heater) or 8±4 % (chimney stove) of the total combustion time. Furthermore, the flaming phases of the first batches were shorter than those of the following batches of dry spruce or beech fuels. The emission profiles were affected by these batchwise differences, with ignition being the period for enhanced organic emissions, whereas the flaming phase was characterized by an increase in particulate emissions consisting mainly of black carbon, as expected based on previous work (Kortelainen et al., 2018). The char burnout phases with these fuels were characterized by high CO concentrations, whereas in moist spruce combustion elevated CO concentrations were measured throughout the batch, thus making the burnout phases less distinguishable.

Figure 3Relative shares of the OGC groups in the exhaust downstream of the PEAR OFR as measured by PTR-ToF-MS, averaged over the total experiment time covering 2–5 batches.

Download

Figure 4Average concentrations of organic gaseous compounds in the dry primary exhaust at the different combustion phases (I: ignition, F: flaming, and B: char burnout phase) for each batch measured by FTIR from the stack. Error bars denote standard deviations between experiments (n=5).

Download

3.2 Primary emissions

3.2.1 Gaseous organic emissions

The primary organic gaseous emissions in the undiluted flue gas were measured continuously by an FTIR multicomponent analyser. Additionally, the PTR-ToF-MS measured diluted, unaged emissions during the no-ageing experiments. These datasets complemented each other because, while FTIR was calibrated for 27 typical combustion-derived compounds (Table S1), a more detailed insight into the composition of the gaseous organic phase was acquired via the PTR-ToF-MS, with which 126 different molecular formulas for OGC in the primary emissions were identified in the m∕z range of 40–180 (Table S2). However, PTR-ToF-MS is unable to detect compounds with proton affinities lower than that of water (691 kJ mol⁻¹), such as alkanes. Furthermore, fragmentation in the PTR-ToF-MS limits the identification of compounds with similar mass-to-charge ratios as common fragment ions, including unsaturated aliphatic compounds, such as propene (${\mathrm{C}}_{\mathrm{3}}{\mathrm{H}}_{\mathrm{6}}^{+}$) at m∕z 41.04 or butene (${\mathrm{C}}_{\mathrm{4}}{\mathrm{H}}_{\mathrm{8}}^{+}$) at m∕z 57.08. However, these compounds were detected by the FTIR.

Table 2Primary PM₁ concentrations in dry, 13 % O₂ flue gas conditions. The number concentration and geometric mean mobility diameter (GMD) were derived from a scanning mobility particle sizer and PM₁, organic carbon, and elemental carbon from filter samples.

Download Print Version | Download XLSX

The major OGC groups measured by PTR-ToF-MS were carbonyls, aromatic hydrocarbons (ArHCs), furans, and phenols (Fig. 3; Table S2). In addition, unsaturated aliphatic compounds constituted a substantial fraction of the total non-methane OGC (NM-OGC) measured using the FTIR (Fig. 4). The NM-OGC emissions based on FTIR were 42.6±9.2 and 101±24.6 mg C m⁻³ for dry and moist spruce combustion in the masonry heater, respectively, and 89.4±11.3 and 148.3±23.0 mg C m⁻³ for beech and moist spruce in the chimney stove, respectively. Thus, the lowest NM-OGC concentrations were measured from the dry spruce combustion. Moist spruce combustion produced a factor of 1.7–3.6 higher emission than dry wood, which is well in line with studies by e.g. McDonald et al. (2000), where moist fuel produced 2–4 times more NM-OGC than dry fuel combustion. The difference between concentrations was highest for oxygenated compounds (factor of 4.2±1.7) and for unsaturated compounds (2.7±0.6), but statistically significant for all subgroups (paired t-test p values ≤ 0.02 for all groups for the consecutive dry/moist experiments; Table S4). Differences in NM-OGC emissions were also observed between the masonry heater and the chimney stove. Emissions from moist spruce combusted in the chimney stove were higher by a factor of 1.5 (p value of 0.05) compared to the masonry heater, with a statistically significant increase in unsaturated and aromatic hydrocarbons (factors of 1.6 and 1.7, respectively, p values < 0.01) based on FTIR measurements.

The NM-OGC emissions of the first batches exhibited a distinct time-dependent behaviour in comparison to the following batches: the emitted concentrations always increased both after ignition and at the end of flaming phase (Fig. 2), whereas in the following batches there was a sharp emission peak at ignition, after which the concentrations declined as soon as the flaming phase began. These findings agree with those by Kortelainen et al. (2018) and are influenced by the fact that cold ignition is performed from the top of a fuel batch, while the following batches are ignited from the bottom by the glowing embers. NM-OGC emissions were lowest during the burnout period when most of the fuel was already consumed (Fig. 2). Thus, NM-OGC emissions had a reverse time profile to the CO emissions, which peak during the char burnout phase. Furthermore, the contribution of the different organic compound groups to the total NM-OGC concentration differed in relation to time (Fig. 4). For example, the importance of aromatic species increased for dry fuels (S-5 % and B-17 %) for the flaming and burnout phases, while other species were pronouncedly emitted at ignition. In addition, the composition of aromatic species measured by PTR-ToF-MS in the unaged emission depended on the phase (Fig. S5). In the masonry heater, ArHCs had the highest contribution during the ignition phase, but their relative share decreased during flaming phase, while the share of furanoic and phenolic compounds increased. The relative importance of furanoic and phenolic compounds in the fresh exhaust of the flaming phase from a masonry heater has been also previously established (Czech et al., 2016). Overall, the share of ArHC in the fresh exhaust is higher and less phase dependent for beech combustion in a chimney stove than for other experiments. The share of N-containing aromatic compounds, namely nitrophenol and nitrocresol, also increased after ignition. These findings are important also when considering the potential of SOA formation, as aromatic compounds have been observed to be the major SOA precursors in RWC emissions, and removal of ArHCs from the flue gas either by improving combustion conditions or using e.g. catalytic cleaning has been noted to decrease also the resulting SOA formation (Bruns et al., 2016; Hartikainen et al., 2018; Pieber et al., 2018).

3.3 OFR conditions

During photochemical ageing, OHR_ext is an important parameter affecting OH-radical consumption and reaction pathways of organic species. For the masonry heater experiments, the average OHR_ext was 130–150 s⁻¹ for dry spruce samples and 160–220 s⁻¹ for moist spruce. In the chimney stove, the average OHR_ext was 280–300 s⁻¹ for beech and 320–370 s⁻¹ for moist spruce samples, except for the low-DR experiment (Exp. 5, DR of 30) where the average OHR_ext reached 820 and 1300 s⁻¹ for the beech and moist spruce samples, respectively. The OFR conditions were divided into good, risky, and bad based on the ratio of the photon flux exposure to OH_exp by the definitions of Peng and Jimenez (2017) (Fig. S7). Here, the ageing conditions were defined as mainly “risky” (4×10⁵ cm s⁻¹ < flux_254 nm,exp ∕ OH_exp < 10⁷ cm s⁻¹) during all the experiments. The OHR_ext varied considerably during combustion cycles, but the limit for “bad” conditions was exceeded only during ignition phases of chimney stove experiments when the OHR_ext peaks above 1000 s⁻¹ for brief periods accounting for up to 4 % of total experiment time. In terms of OA emission, the bad conditions accounted for 3 %–7 % of chimney stove OA emissions, excluding the S-22 % combustion for the low-DR experiment where 30 % of OA was emitted under bad conditions.

Evaluation of the OHR_ext was limited to the externally added butanol-d9 and the compounds measured from the primary exhaust with FTIR, with NO, CO, and unsaturated hydrocarbons as the main OHR_ext producers (Fig. S2). In other words, the products of later-generation oxidation or from fragmentation from the particulate phase were not considered. The average NO_x concentrations input to the PEAR OFR ranged from 150 to 420 ppb, except for the low-DR experiment with concentrations of 1140 and 751 ppb for beech and spruce combustion, respectively. NO is rapidly oxidized to NO₂ with the addition of O₃ and then partially converted to particulate nitrate. The subsequent low-NO conditions in the PEAR decreased the reactions of organic peroxy radicals (RO₂) with NO. Based on the RO₂ chemistry model by Peng et al. (2019) for 254 nm OFRs, the high photon fluxes and high concentrations of both OH and HO₂ in the PEAR OFR result in lower RO₂ lifetime and consequently lower RO₂ isomerization rate than in atmosphere, while the importance of RO₂+OH reaction is enhanced due to lower-than-ambient HO₂-to-OH-ratio (Fig. S8) (Peng et al., 2019).

The average batchwise OH exposure in the PEAR OFR during photochemical ageing experiments ranged from 0.5 to 7 eqv. d depending on the applied photon flux and the OHR_ext of the sample. During the experiments, the photochemical exposure varied in line with the varying OHR_ext during batchwise combustion, with the highest exposure occurring during the flaming phase and lower exposure during the ignition period (Fig. S1). The extent of alternative non-OH reaction pathways during photochemical ageing in the PEAR OFR were compared to the tropospheric conditions (Sect. S4.1), and the exposures to O(¹D) and O(³P) in the PEAR OFR were estimated to be similar to those in ambient conditions (Fig. S9), excluding the ignition period during the low-DR experiment, where the importance of O(³P)_exp briefly exceeded ambient conditions. The exposure to ozone in relation to OH radicals was estimated to be lower in the PEAR OFR than in the troposphere (${\mathrm{O}}_{\mathrm{3},\exp }/{\mathrm{OH}}_{\exp }$ < 10⁵). However, our estimations were based on the initial O₃ concentrations, whereas O₃ is expected to form in the PEAR OFR during photochemical ageing as a product of the OH + OGC reactions (Carter, 1994). This may have led to the ${\mathrm{O}}_{\mathrm{3},\exp }/{\mathrm{OH}}_{\exp }$ being slightly underestimated.

Several of the emitted OGCs were susceptible to photolysis at 254 nm, most importantly the ArHCs, which are among the main SOA precursors from RWC. Photolysis as a degradation pathway can exceed OH reactions also in the atmosphere for compounds with high photolysis rates (Hodzic et al., 2015). Here, the importance of photolysis was notable, e.g. for benzene, of which more than 40 % may have degraded via photolysis (Fig. S10). However, the importance of photolysis during photochemical ageing is inversely proportional to the ratio of the OH reaction rate to the photoabsorption cross section (σ_abs), and it can be considered a minor pathway for other main OGCs (Table S7), including other aromatic species such as toluene (< 10 % of total degradation during these experiments; Fig. S10).

The detailed results of the LVOC fate calculations are presented in Sect. S4.2. Briefly, the majority of LVOCs were estimated to condense into the particulate phase in all experimental conditions (Fig. S11). Similar to earlier OFR studies (Bruns et al., 2015a; Ihalainen et al., 2019; Simonen et al., 2017), the downstream particle number concentration and condensation sinks were influenced by new particle formation (Fig. S12) with the extent of nucleation depending strongly on the ageing conditions and the concentrations of condensable vapours. However, the distinct formation of ultrafine particles does not necessarily represent the atmospheric fate of condensable vapours because of the faster-than-ambient oxidation, which results in higher saturation ratios and faster gas-to-particle conversion in the PEAR OFR. Gas-to-particle conversion is also affected by the dilution ratio, which was in the PEAR OFR lower than what typically exists in atmosphere. Therefore, it has to be noted that the relative fraction of the measured semivolatile species in the condensed phase are likely lower at ambient dilution ratios than in the conditions of this study.

3.4 Transformation of gaseous phase

The composition of the OGCs changed throughout the studied exposure range, and the abundance of the secondary compounds increased with an increase in photochemical exposure. Namely, the OGCs measured with PTR-ToF-MS from the aged gas phase were governed by small carbonyls and fragmentation products such as acetic acid (C₂H₄O₂, m∕z 61.02). However, while the small oxidized organics became increasingly dominant in the gas phase, photochemical ageing particularly decreased the share of aromatic compounds (Fig. 3). The evolution of the total chemical composition is visualized in the van Krevelen diagram, which is often used for simplified characterization of particulate OA (Heald et al., 2010) and was here extended for the investigation of the gaseous organic phase measured by PTR-ToF-MS (Fig. 5). Photochemical ageing caused several simultaneous and subsequent functionalization reactions of the organic compounds, and the increase in the average H : C ratio together with an increasing O : C ratio led to linear slopes from +0.69 to +1.0 in the van Krevelen diagram for gaseous phase. Similar, positive slopes are derivable also from previous studies of ageing of RWC OGC in smog chambers (Bruns et al., 2017; Hartikainen et al., 2018). Most of the OGCs in primary RWC exhaust have relatively high OH reactivities, and thus their reactions are expected to take place within the first atmosphere equivalent hours of photochemical ageing (Bruns et al., 2017; Hartikainen et al., 2018). Therefore, the secondary OGC products, including the oxidized fragmentation products from the particulate phase, dominate the total OGC in the long-range transported smoke. In addition, the loss of compounds with low H : C and O : C ratios, such as aromatic species, is important for the change in the average composition. Furthermore, the oxidation of compounds undetected by the PTR-ToF-MS, such as alkanes, may introduce secondary products with higher proton affinities and their subsequent appearance in the spectra. Conversely, ageing may also lead to the growth of compounds outside the observed mass range (m∕z 40–180). Compared to previously measured changes in RWC exhaust in a chamber (Hartikainen et al., 2018), ageing in the PEAR OFR led to a slightly higher increase in the H : C ratio. This difference implies an increased fragmentation which is likely the result of faster OH exposure and differences in the RO₂ chemistry (Peng et al., 2019) in the PEAR OFR compared to ageing in a smog chamber.

Figure 5Van Krevelen diagram of the OGCs measured by PTR-ToF-MS. Numbers indicate batches combusted in the masonry heater (1–3: dry spruce; 4–5: moist spruce); letters refer to chimney stove batches (A–E: beech; F–G: moist spruce). Ozonolysis experiments (in black) were not considered in the slope calculations. ^* Behaviour of RWC emission aged in a chamber (Hartikainen et al., 2018).

Download

Aromatic compounds consisting of ArHCs, phenols, furans, N-containing aromatic compounds (N-aromatics), and other oxygenated aromatics were important constituents of the primary organic gas phase and formed 37 %–39 % of the fresh emission from the masonry heater and 33 %–34 % from the chimney stove, as measured by the PTR-ToF-MS (Fig. 3). Similar shares have been measured from fresh RWC exhaust by Bruns et al. (2017) (13 %–33 %) and Hartikainen et al. (2018) (33 %–36 %). However, after 2 eqv. d of exposure, their share decreased to less than 20 % of the identified OGCs, which agrees with previously reported conversions of aromatics during photochemical exposure (Bruns et al., 2017; Hartikainen et al., 2018). Overall, the photochemical reactions of aromatics are an important source of SOA because they form products that efficiently partition into the particulate phase. However, there are large differences between the conversion efficiencies of aromatic compound groups. While ArHCs comprise approximately half of the aromatic compounds in fresh gaseous exhaust, their share grows to over 70 % in 5–7 eqv. d, while the share of oxygenated aromatics decreases with ageing in line with their higher OH reactivity. Similar aromatic behaviour in RWC exhaust was observed earlier in a chamber with spruce exhaust (Hartikainen et al., 2018), where the molar share of ArHCs in the total aromatic content increased from 45 % and 32 % to 63 % and 54 % during ageing of 0.6 and 0.8 eqv. d, respectively, while the share of furanoic and phenolic species decreased. Furthermore, N-aromatics were not detectable here with ages exceeding 1 eqv. d, although they have been observed to form with shorter exposures (Hartikainen et al., 2018). The N-aromatics produced by the first stages of ageing may have partitioned to the particulate phase but are also degraded by subsequent reactions with OH. Simultaneously, the share of aliphatic nitrogen compounds (CHN and CHNO) to the total concentration was not substantially affected by photochemical exposure.

Carbonyls were divided into primary and secondary subgroups based on their behaviour during ageing. The primary carbonyl group, consisting mainly of acetaldehyde and to a smaller extent of compounds such as acrolein and butadiene, was prevalent in the fresh exhaust, but their share of identified compounds decreased from 13 %–27 % in unaged exhaust to 3 %–12 % in the highest exposures. This is the result of the degradation of these compounds and the introduction of high amounts of carbonyls in the secondary carbonyl group. The ratio between the two carbonyl groups increased linearly with age. The secondary carbonyl group was dominated by acetic acid, which was the most prevalent compound after extensive ageing in all experiments and covered over 30 % of the total measured OGC concentration from the highly aged S-5 % and B-17 % exhaust. The mainly small acidic compounds in the secondary carbonyl group were formed from the photochemical reactions of OGCs and from particulate OA, which is a consistent source of oxygenated OGCs such as acetic acid, formic acid, acetaldehyde, and acetone (Malecha and Nizkorodov, 2016). Of these, acetaldehyde is classified as a primary carbonyl because it reacts with OH 2 orders of magnitude faster than the other considered small carbonyls (Atkinson et al., 2006), and thus its concentration remained at a similar level throughout the ageing process.

3.5 Transformation of particulate phase

3.5.1 OA enhancement and composition

The photochemical ageing process increased the mass of particulate organic carbon measured with thermal–optical carbon analysis by factors of 1.3 and 3.9 for dry and moist spruce combustion in the masonry heater, respectively, and by factors of 2.0 and 3.0 for beech and moist spruce combustion in a chimney stove, respectively. This agrees well with the previously observed SP-HR-ToF-AMS-based OA enhancement factors (1.6–5.3) for RWC (Bertrand et al., 2017; Bruns et al., 2015b; Grieshop et al., 2009; Heringa et al., 2011; Tiitta et al., 2016) and with the thermal–optical analysis-based organic carbon enhancements previously measured for the same chimney stove (1.3–1.4 after 1.7–2.5 eqv d) (Miersch et al., 2019). The organic carbon concentrations after the PEAR OFR correlated well with the OA measured with SP-HR-ToF-AMS (R²=0.85, Fig. S6). Ageing led to a linear increase in the SP-HR-ToF-AMS-derived ratio of organic matter to organic carbon (OM : OC, Fig. S13), which rose from the initial average ratio of 1.8–2.2 to 2.7–3.0 during extended ageing. Similarly, the OS_C of the OA increased as a function of the photochemical age throughout the tested exposure ranges (Fig. S13), indicating the existence of continuing reactions of the particulate phase after the rapid consumption of the majority of the primary gaseous SOA precursors. As expected, the organic carbon mass enhancement did not increase with continuous ageing because the major SOA precursors were already consumed by relatively short OH exposures. Continuing photochemical exposure may instead reduce the amount of particulate organic carbon (Kroll et al., 2015), which acts as a source for volatile acidic compounds during photochemical ageing (Malecha and Nizkorodov, 2016; Paulot et al., 2011). See Table S8 for experiment-wise average concentrations and chemical compositions of the particulate matter measured by SP-HR-ToF-AMS.

The oxidation states measured after the PEAR OFR without oxidative ageing were highest for the chimney stove OA, with average unaged OS_C's of 0.22 and 0.41 for beech and spruce combustion, respectively, while dry and moist spruce in the masonry heater had average unaged OS_C's of −0.18 and 0.15, respectively. This indicates the existence of different combustion conditions in the studied appliances, with the masonry heater having lower emissions of highly oxygenated compounds and a higher share of unsaturated hydrocarbons compared to that of the chimney stove emissions. However, as a result of oxidative ageing, OS_C surpassed 1.5 after 5 eqv. d regardless of the type of experiment, which exceeds the values typically observed in ambient aerosol (Kroll et al., 2011; Ng et al., 2011). This is likely because highly aged aerosol in ambient air is mixed continuously with fresh and less oxidized sources. However, the chemical evolution of OA in the PEAR OFR followed a similar trend to that observed for typical ambient semivolatile oxygenated aerosol, with a van Krevelen slope of −0.5, which can be interpreted as a result of simultaneous fragmentation and acid-group addition (Ng et al., 2011). Furthermore, the changes in OS_C's were similar to those previously observed for ambient aerosol aged extensively in a potential aerosol mass (PAM; Kang et al., 2007) OFR (OS_C of up to 2; Ortega et al., 2016). The OS_C's of RWC exhaust aged in a PAM OFR have, however, been notably lower than observed here (Fig. 6; Bruns et al., 2015a; Pieber et al., 2018).

Figure 6Van Krevelen diagram of the particulate organic aerosol measured by SP-HR-ToF-AMS. Numbers indicate batches combusted in the masonry heater (1–3: dry spruce; 4–5: moist spruce), whereas letters refer to chimney stove batches (A–E: beech; F–G: moist spruce) of this study. Ozonolysis experiments are marked with black and were not considered in slope calculations. Results from ageing of RWC exhaust with a PAM OFR by Bruns et al. (2015a) and Pieber et al. (2018), as well as behaviour of RWC OA aged in a chamber (Tiitta et al., 2016) are marked with scatter.

Download

Figure 7Compounds measured by IDTD-GC-ToF-MS in different experiments (ozonolysis, low, and high photochemical exposure) with respect to their carbon number (n_C) and oxidation state (OS_C). Enhancement ratios (ER) compared to the experiments without oxidative ageing are shown in colour, and the size indicates the dilution-corrected concentrations (normalized dry, 13 % O₂ flue gas conditions) in the secondary exhaust.

Download

A higher initial oxidation state in the chimney stove exhaust led to shallower slopes (−0.17 and −0.34) in the van Krevelen diagram than those in the masonry heater exhaust (−0.46 and −0.49, Fig. 6). The slope for the beech combustion aerosol was higher than for spruce, which may result from differences in fuel composition as has been observed also previously for biomass burning OA (BBOA) (Ortega et al., 2013). Previously, the ageing of spruce combustion exhaust from a masonry heater in a chamber has produced similar but slightly steeper van Krevelen slopes of −0.64 to −0.67 (Tiitta et al., 2016). The steepness may be affected by the shorter ageing (< 1 eqv. d) and consequent lower final OS_C (maximum of +0.14) in the chamber, as the O : C ratio has been noted to level off at higher oxidation states (Ng et al., 2011). Slopes are also positively affected by fragmentation, which may be enhanced in the PEAR OFR because of the more intensive UV radiation than in smog chambers or in ambient conditions. The van Krevelen slopes agreed well with those previously measured from single precursors aged in the PAM OFR, such as −0.48 for toluene or −0.46 for xylene (Lambe et al., 2011). Similar slopes have been observed also for ageing of OA from pellet combustion (−0.47 to −0.44; Czech et al. 2017) or open biomass combustion (−0.5; Ortega et al. 2013) with a PAM OFR.

The evolution of particulate organic aerosol was assessed also by the IDTD-GC-ToF-MS analysis of filter samples. When comparing the concentrations in the exhaust after the PEAR OFR, the concentrations of compounds with high oxidation states and low carbon number (n_C) increased during photochemical ageing (upper-right corner of Fig. 7). The locations of the measured organic compounds in the OS_C : n_C space are shown in Fig. S14, and their dilution-corrected concentrations downstream of the PEAR OFR in Table S11. In the OS_C : n_C space, the compounds which exhibited a major increase during photochemical ageing were located in or above the location of the low-volatility oxidized organic aerosol (LV-OOA) classified by Kroll et al. (2011). These compounds are products of the multistep oxidation process including both fragmentation and oxidative reactions.

OA formation is tied to the availability of organic precursors, and thus the formation of SOA was highest at the ignition phase of each batch (Fig. S15). In the masonry heater, the enhancement in particulate OA mass increased considerably with the introduction of moist logs, simultaneously with the increase in gaseous organics in the fresh emissions compared to the preceding dry spruce combustion. In contrast, the low organic emission by dry spruce combustion was reflected as a lower SOA formation. Another aspect related to the primary exhaust was the extent of OH exposure in the PEAR OFR, which was directly connected to the sample concentrations, as discussed in Sect. 3.3. Namely, the photochemical ageing was lower during periods of high emission, leading to lower oxidation states for the OA emitted during ignition.

Figure 8Average shares of the four PMF factors in the exhaust at different exposure levels.

Download

3.5.2 PMF analysis of particulate OA composition

PMF analysis applied to the exhaust produced a four-factor solution for the OA covering 98 % of the spectra. The spectra of the factors are shown in Fig. S16. Two of the factors were associated in particular with the primary OA from biomass combustion: pyrolysis-BBOA, formed especially during ignition, and flaming-BBOA, emitted pronouncedly during the flaming phase. The other two factors, semivolatile oxygenated OA (SV-OOA) and low-volatility oxygenated OA (LV-OOA), represent oxygenated organics with notably higher OS_C's than those of the primary OA factors (Table 3). Flaming-BBOA comprised 76 % and 55 % of the unaged OA from dry and moist spruce combustion in the masonry heater, respectively, but only 27 % and 23 % in the beech and moist spruce combustion OA from the chimney stove, respectively (Fig. 8), indicating a less-oxidizing higher-temperature flaming phase in the modern masonry heater. Flaming-BBOA is strongly related to the ${\mathrm{C}}_{\mathrm{4}}{\mathrm{H}}_{\mathrm{9}}^{+}$ ion (main peak of m∕z 57; R²=0.90, Fig. S17), which is used as a tracer for hydrocarbon-like compounds (Aiken et al., 2009). However, in contrast to the typical hydrocarbon-like OA factor characterized by a relatively low OS_C (−1.7 to −1.6; Kroll et al., 2011), the flaming-BBOA contains more oxygen-containing functional groups and is similar to the primary biomass burning OA factor measured in the RWC exhaust in a chamber (Tiitta et al., 2016; θ=13.6^∘; R²=0.95). Pyrolysis-BBOA, on the other hand, consisted of ions typical to the low-temperature pyrolysis products of wood combustion and its time series correlated well with the PACs (R²=0.86, Fig. S17).

Table 3Properties of the PMF factors.

Download Print Version | Download XLSX

Of the more-oxygenated factors, the LV-OOA was dominated by the CO⁺ and ${\mathrm{CO}}_{\mathrm{2}}^{+}$ ions and thus represented highly oxidized OA. The LV-OOA spectra corresponded well with the OH-induced SOA factor identified from the RWC exhaust aged in a smog chamber (Tiitta et al., 2016.; θ=8.4^∘; R²=0.98) and was comparable with that of ambient LV-OOA (Aiken et al., 2009) and the LV-OOA spectra of unaged wood combustion exhaust (Kortelainen et al., 2018). LV-OOA was also present in the unaged exhaust of this study, excluding the dry spruce combustion in the masonry heater, which produced the least-oxidized primary exhaust. The SV-OOA, on the other hand, was related to the C₂H₃O⁺ ion (R²=0.78), which is indicative of carbonyl formation in the particulate OA (Ng et al., 2010). The SV-OOA was also comparable ($\mathit{\theta }=\mathrm{11.5}{}^{\circ }$; R²=0.96) to a factor of SOA generated in a chamber in the previous work (Tiitta et al., 2016), where this factor was interpreted as ozonolysis-generated organic OA. Interestingly, SV-OOA was formed only during photochemical ageing and increased in line with higher exposures despite semivolatile compounds being products of the initial stages of OA oxidation. This further demonstrates the long-continuing functionalization of OA during ageing alongside with the fragmentation processes.

Figure 9Relationship of photochemical ageing to the batchwise average ratios of (a) total HR-PAH concentration to OA concentration and of (b–f) HR-PAH subgroups to the total PAC concentration measured by the SP-HR-ToF-AMS.

Download

Figure 10Dilution-corrected concentrations for selected compounds measured by IDTD-GC-ToF-MS at different exposure levels. Normalized to dry, 13 % O₂ flue gas conditions.

Download