The AAC (Cambustion Ltd, Cambridge, UK) is used to select aerosols within a narrow range of aerodynamic diameters and does not suffer from the issue of multiple charges that affects selection using instruments such as the CPMA and DMA. The AAC uses a centrifugal force and sheath flow between two concentric rotating cylinders to produce an aerosol classified by aerodynamic diameter. The detailed information regarding the principle of AAC can be found in Tavakoli et al. (2014).

The EXSCALABAR instrument used in this work was developed by the Met Office (Exeter, UK), which can be operated in both the laboratory and from the UK atmospheric research aircraft (FAAM BAe-146). The operating principle of EXSCALABAR has been described in detail in previous papers (Davies et al., 2018; Cotterell et al., 2020, 2019a, b). In brief, the instrument uses cavity ring-down spectroscopy (CRDS) to measure the dry aerosol extinction at wavelengths of 405 and 658 nm and photoacoustic spectroscopy (PAS) to measure the dry aerosol absorption coefficient at wavelengths of 405, 514 and 658 nm. For deployment during SASSO, all cells shared common sample conditioning; the aerosol sample relative humidity was reduced to <10 % by passing through a Nafion dryer (Perma Pure LLC), and ozone and NOx were removed by an activated charcoal “honeycomb” scrubber (custom-built in-house). The sample passed through an impactor (Brechtel Manufacturing Inc., custom-built for an 8 L min-1 volumetric flow rate) with aerodynamic cut-off (D50) of 1.3 µm before being drawn through a series of flow splitters that evenly distributed the aerosol-laden air samples to the various optical spectrometers, each operating at 1 L min-1. For both the 405 and 658 nm wavelengths, the PAS cells were mounted downstream of the CRDS cells (CRDS-PAS); i.e. the sample passed through the CRDS cell before entering the PAS cell of the same wavelength. All other cells operated in a parallel flow configuration. During the wood combustion experiments, a condensation particle counter (CPC) (Model 3776, TSI, USA) was put in series after the 405 nm CRDS-PAS to measure the number concentration of particles contributing to the extinction and absorption signal. For an additional 405 nm CRDS spectrometer, the aerosol sample passed through a HEPA filter prior to sampling to provide a continuous measurement of the baseline and highlight the presence of any gaseous absorbers affecting measurements at that wavelength. CRDS and PAS measurements rely on characterisation of the aerosol-free background. Before and after each wood combustion experiment (run), sample flow was routed through a HEPA filter immediately after EXSCALABAR's common sample inlet section to provide baseline measurements of empty-cavity ring-down time and background photoacoustic response for all CRDS and PAS spectrometers, respectively. During the Manchester chamber experiments, the CPC was fitted in parallel with the sample lines, and automated baseline measurements were made every 10 min. PAS cells with improved sensitivity – as described by Cotterell et al. (2019b) – were gradually implemented during 2019; only the 405 nm dry absorption measurement used the new cells during the Wildfire lab wood combustion experiments, but all cells were upgraded by the start of the Manchester aerosol chamber experiments. The PAS cells were calibrated using ozone before and after each set of experiments as well as at least every working week during the experimental periods, as described by Davies et al. (2018) and discussed further by Cotterell et al. (2019a).

The refractory black carbon (rBC) mass concentration was measured by a SP2 (Droplet Measurement Technologies, Colorado, USA). The SP2 uses the laser-induced incandescence to measure the rBC mass and optical size of individual BC particles with an intra-cavity Nd:YAG laser operating at 1064 nm. The particle size can be determined by detecting the laser signal scattered by particles, with the scattering intensity maximum related to the optical particle diameter through a calibration using polystyrene latex spheres. The optical size of BC-containing particles is determined by matching the measured scattering signal with calculations from light scattering calculations assuming a core–shell structure (core–shell Mie theory) (Moteki and Kondo, 2007). Because the scattering signal of the absorbing particle will be distorted during its transit through the laser beam due to the mass loss by laser heating, the leading edge scattering signal before the onset of volatilisation is extrapolated to reconstruct the scattering signal of the absorbing particles (Gao et al., 2007; Liu et al., 2014). This calibration was performed at the start of the measurement campaign. For those particles which contain absorbing materials such as the refractory BC, they will absorb 1064 nm light and then heat up and emit visible thermal radiation (incandesce). This incandescence signal is directly proportional to the mass of rBC as determined by a calibration (Liu et al., 2010) with generated BC aerosols of known or independently measured mass. In the atmosphere, besides BC, other materials (e.g. some metals and minerals) can incandesce at 1064 nm as well. However, as boiling-point temperatures of these materials are rather different to that of black carbon, it is easy to distinguish them in measurements made using the SP2 equipped with an additional narrowband incandescence detector, such as the one used here (Liu et al., 2018). Recently, Sedlacek et al. (2018) demonstrated that charring of light-absorbing organic particles at 1064 nm can produce the refractory black carbon and then overestimate the rBC mass concentration; however this does not affect pure BC particles, and we saw no evidence for an incandescence signal associated with the pure organic particles measured in this study. We investigated the effectiveness of this calibration from different types of BC particles in this study and report the outcomes of these investigations in Sect. 3.3.

Non-refractory aerosol chemical compositions including sulfate, nitrate, ammonium, chloride, and organics were measured by a HR-AMS in real time. The HR-AMS was operated in fast mode to capture the fast transition of the combustion phase during the wood combustion experiment (Kimmel et al., 2011). The instrument operation and data analysis of HR-AMS have been described in detail elsewhere (Alfarra et al., 2006; Allan et al., 2003, 2004; DeCarlo et al., 2006). The ionisation efficiency of the AMS was calibrated using monodisperse ammonium nitrate particles according to the method described by Jayne et al. (2000) at various times during the experimental periods.

The CPMA (Cambustion Ltd., Cambridge, UK) uses opposing electrical and centrifugal fields to classify particles according to their mass-to-charge ratio. The ability to vary the electrical field and rotation speed enables particle selection based on their mass. The principles and operation of the CPMA have been described elsewhere (Olfert and Collings, 2005; Olfert et al., 2006). Combining the CPMA with a CPC (Model 3776, TSI, USA) and scanning across the mass range of interest provides the bulk aerosol mass size distribution. As the CPMA uses an electrical classification method to select particles, it also suffers from an issue of multiple charging similar to a DMA, and it has problems associated with a fraction of the uncharged particles that are also transmitted, particularly at the lower rotation speeds. In this study, an electrical ioniser (MSP Corp., USA) was used for wood combustion experiments, and a 90Sr radioactive ioniser was used for chamber experiments to neutralise particles before they were sampled by the CPMA.

Aerosol size distributions in the diameter range from 14.9 to 673.2 nm were measured by a commercial SMPS (TSI, USA, employing a model 3082 classifier unit, 3081 DMA and 3786 CPC) with sheath and sample flow rates of 3 L min-1 and 0.3 L min-1, respectively. The DMA was operated in scanning mode, with a scan time of 60 s and a retrace time of 4 s. The particle size distribution was corrected for multiple charge effects and diffusion loss using the standard inversion algorithm in the SMPS software (AIM version 10). Before the experiment, the SMPS was calibrated using NIST-certified polystyrene latex spheres (PSLs, Thermo Fisher Inc.).

To investigate morphological properties, BC particles were collected on carbon-coated copper grids using an electrostatic sampler (ESPnano, DASH Inc., USA) (Miller et al., 2010) for transmission electron microscopy (TEM) analysis. The grids were analysed using a FEI Titan operated at 300 kV. Several areas of the grid were imaged to provide representative images for each sample since both the grid film and particles contained carbon.

We explored the impacts of the partitioning of coating materials on the structure of BC. For this BC restructuring experiment, a thermal denuder was utilised to remove the coating materials on coated BC particles. The home-built TD consisted of a stainless steel tube in a temperature-controlled furnace (Voliotis et al., 2021). The TD had a length of 0.97 m and an internal diameter (i.d.) of 0.15 m. Aerosols entered and exited the TD unit via a cylindrical 0.12 m long and 0.037 m i.d. stainless steel compartment. The temperature in the heating section (0.51 m × 0.15 m i.d.) was controlled by four PID controllers (Watlow EZ-ZONE) with additional temperature sensors on the outside of the tube. It is necessary to ensure flow through the TD whenever it is heated, even when bypassed, to allow for measurements of the unheated sample. A vacuum line maintained 2.0–2.5 L min-1 through whichever of the bypass or TD line was not in use. The residence time of the air sample in the heating section was 216–270 s. The TD was given 30 min for its temperature to stabilise before sampling. The temperature of the TD was calibrated by measuring the temperature at the axial centre of the denuder's heating zone. In this study, the purpose of the TD is to remove as much of the organic coatings from the combustion-generated particles as possible, rather than probing the volatility properties of the coating. Therefore, all heating zones of the TD were set to their maximum temperature of 180 ∘C. This upper temperature is lower than that achieved by other commercial TD units and minimises the risks of charring.

The multi-tube Nafion humidifier in the Manchester home-made Hygroscopicity Tandem Differential Mobility Analyzer (HTDMA) system was used for the BC restructuring experiment. The principle and configuration of the HTDMA system including the Nafion humidifier flow system can be found in Good et al. (2010). The RH in the Nafion humidifier was controlled by adjusting the relative mixing ratio of dry air from a compressed air source with humidified air generated by bubbling compressed air through a water-filled glass bulb.

The Manchester Aerosol Chamber comprises an 18 m3 collapsible Fluorinated ethylene propylene Teflon bag (3 m (H) × 3 m (L) × 2 m (W)). It was used by Alfarra et al. (2012) to investigate the effect of photochemical ageing and initial precursor concentration on the composition and hygroscopic properties of secondary organic aerosol. The chamber was run as a batch reactor in which the composition of the gaseous precursors, oxidising environment, primary emissions or seed particles, relative humidity and temperature were controlled. Air was supplied to the chamber by a blower at a flow of 3 m3 min-1. The air was dried and filtered for gaseous impurities and particles using a series of Purafil (Purafil Inc., USA), charcoal and HEPA filters (Donaldson Filtration, USA) prior to humidification with ultrapure deionised water. Halogen bulbs and two 6 kW Xenon arc lamps were mounted on the inside of the enclosure housing the bag, which was coated with a reflective space blanket (mylar) to maximise the irradiance in the bag and to ensure even illumination. The Xenon arc lamps were mounted on two opposite sides of the enclosure at different heights. The combination of illumination was tuned and evaluated to mimic the atmospheric actinic spectrum over the wavelength range 290–800 nm and had a maximum total actinic flux of 1.4 × 1014 (photon s-1 m-2 nm-1) over the region 460–600 nm. The calculated j(O1D) value during the reported experiments was 1.23 × 10-5 s-1 (290–340 nm), and j(NO2) was 1.5 × 10-3 s-1 (290–422 nm). The relatively large volume of the chamber allows the dilute sample to be held for several hours without significant aerosol removal from wall losses, allowing for the study of particles introduced directly or formed within the chamber over a period of several hours. The mean number and mass wall loss rates of particles inside the chamber were estimated as 9.17 ± 1.3 and 8.16 ± 1.5 × 10-5 s-1, respectively (Shao et al., 2021). Further details on the chamber are given by Alfarra et al. (2012).

The engine used in this study was a Volkswagen 1.9 L SDI light-duty diesel engine (EURO 4 car equivalent), mounted on a test rig (CM12; Armfield Ltd., Hampshire, UK) and coupled to an eddy current dynamometer. This setup was used for investigating the light absorption properties of black carbon aerosol by Liu et al. (2017) previously. The engine throttle and the load on the dynamometer were controlled with dedicated software. The engine exhaust was passed through an oxidising catalytic converter (a retrofitted diesel oxidation catalyst consisting of a mixture of platinum and rhodium) followed by a standard Volkswagen silencer then a computer operated pneumatic valve connected to the chamber, with 2 in. i.d. stainless steel tubing between each component. The whole length of the exhaust line was 4 m. The fuel used was standard UK low-sulfur diesel, obtained from a local fuel station.

For each run, the engine was started and warmed up at 2000 rpm with 30 % load. Once the engine reached a steady temperature (after ∼ 10 min), a controlled amount of the exhaust was injected into the Manchester Aerosol Chamber by switching the valve for a set time (this condition is denoted as “hot engine”). For the cold-start runs (in which no throttle or load was applied, denoted as “cold idle”), the exhaust sample was injected within a minute of the engine starting.

A miniature inverted soot generator (MISG; Argonaut Scientific) operated with propane was used to generate black carbon aerosol. The design, operation and performance of the MISG have been previously reported (Kazemimanesh et al., 2019; Moallemi et al., 2019). The MISG was operated under two conditions: with the flow rate of propane and air at 0.0625 and 10 L min-1 respectively (Condition 1) and 0.0625 and 7.5 L min-1 respectively (Condition 2). Under these two conditions, the MISG produces black carbon with average elemental carbon / total carbon of 95 %, determined via thermal-optical analysis.

Our wood combustion experiments were designed to expand on the studies reported by Haslett et al. (2018) to produce repeatable combustion events with discernible transitions between the three burning phases of pyrolysis, flaming and smouldering combustion. Haslett et al. (2018) used the FM Global Fire Propagation Apparatus (FPA) (FTT, East Grinstead, UK) as their controlled ignition source, while we used the iCone Calorimeter (Fire Testing Technology, FTT, East Grinstead, UK), located in the wildFIRE laboratory at the University of Exeter (Fig. 1a). The main difference between the FPA and iCone approaches is that the there is a forced flow of air from beneath the sample in the FPA, whereas in the iCone the sample sits in ambient air conditions. Otherwise, both approaches rely on oxygen consumption calorimetry to measure the heat release rate from a burning object.

The aim of the wood combustion experiments was to produce different phases of combustion that were as well separated as possible, rather than under natural burning conditions. This allowed the aerosol emissions from different burn phases, pyrolysis, flaming and smouldering, to be analysed separately. Three core wood types were selected for all the experiments which we hypothesised would produce different burn conditions, i.e. different durations of pyrolysis, flaming and smouldering. These wood types were Sequioadendron giganteum (Giant Redwood), Pinus sylvestris (Scots pine) and Populus nigra (poplar), with bulk densities of 0.45, 0.51 and 0.4 × 103 kg m-3 respectively. The Giant Redwood and Scots pine are gymnosperm “softwoods”, with the former considered to have low density and the latter a high density, while the poplar is a low-density angiosperm “hardwood” species. In addition, we also sampled combustion emissions from Thuja plicata (Western red cedar) with a bulk density of 0.38 × 103 kg m-3. Both Scots pine and Western red cedar are resin-rich, while the Giant Redwood contains less resin, and the poplar contains barely any resinous compounds. Hence, all should be capable of producing different volumes and types of aerosol particles. There are many factors that can influence composition and properties of generated aerosols during combustion, such as the wood resin content and moisture content, which would result in a highly extensive variable set. In the work presented here, we only focus on the influence of water content of wood samples on the physical and chemical properties of the particles from wood combustion. We assessed wood of two different levels of fuel moisture: fully oven-dried samples and moist samples with ∼ 25 % moisture content.

During the experiment, a wood sample (L×W×H (mm): 90×90×30) was first placed into a custom-made stainless steel basket and then exposed to a radiant heat flux of 40 kW m-2 using the iCone Calorimeter. This radiant heating thermally decomposes the wood sample and generates volatile gases (pyrolysate) as the wood begins to pyrolyse. A continuously operated spark igniter that was positioned in the released stream of pyrolysate acted as a source of ignition. Immediately following the initial placement of the wood sample under the radiant heat source, pyrolysate was continuously released, and the pyrolysate concentration immediately above the wood sample and in the vicinity of the spark igniter increased also. The sample ignited once this pyrolysate reached sufficient concentrations and was well mixed with the surrounding air, at which point the igniter was switched off and removed from the air flow. Measurements were taken as soon as a sample was exposed to the radiant heat flux. Each sample was allowed to flame for 5 min and the aerosols captured from this phase, and then the flames were snuffed out manually leaving the fuel smouldering so that separate measurements of the smouldering phase could be taken alone for a further 5 min.

Throughout the experiment, the exhaust emissions from combustion were collected in a hood and entered into an exhaust duct. The concentration of oxygen, carbon dioxide and carbon monoxide in the exhaust gas was measured by gas analysers integrated into the iCone system and used to calculate the heat release rate from the burning fuel. We installed an outshoot on this duct to allow part of the exhaust to be carried to a separate sampling system that allowed for the measurement of the aerosol properties in the exhaust. Thus monitoring could occur in real time by the instruments (described above and shown in Fig. 1), after the exhaust was diluted with the compressed air through a set of Dekati DI-1000 ejector diluters.

For the purposes of SP2 incandescence signal calibration and morphology investigation of BC particles, three types of the BC particles other than those emitted from wood combustion were investigated through chamber experiments (Fig. 1b). For the BC particles generated by the MISG, BC particles were introduced into the chamber through a Dekati DI-1000 ejector diluter. For the BC particles from the Aquadag standards, the Aquadag standard was first generated by an aerosol atomiser (Topas ATM 226) and then passed through a diffusion dryer (using silica desiccant) prior to entering the chamber. For the BC particles emitted from the engine, a hot engine running condition (2000 r.p.m., 30 % load and 10 min warm-up) and a cold idle condition were investigated. After the BC mass concentration had accumulated to adequate levels in the chamber, a catalytic stripper (operated at 350 ∘C, Model CS10, Catalytic Instruments, Germany) was used to remove the coatings on the BC particles before they were sampled by the instruments.

For the purpose of testing models of BC and SOA mixing during the SASSO project, two procedures for generating SOA were used: one that yielded non-absorbing SOA and one that produced brown carbon. Experiments were conducted in the photochemical aerosol reaction chamber at the University of Manchester. During the experiment, NO2 (10 % v/v, with a balance gas of high-purity N2 (BOC, UK)) was injected directly into the bag from a custom-made gas cylinder via stainless steel tubing, and its concentration was measured using a chemiluminescence gas analyser (Model 42i, Thermo Scientific, MA, USA). After the NOx concentration in the bag reached the desired value, the precursor volatile organic compound (VOC) was injected into a heated glass bulb (80 ∘C) and flushed into the bag with high-purity nitrogen. Subsequently, chamber lights were turned on to trigger SOA formation. The EXSCALABAR system was used to determine the light absorption properties of the formed SOA particles in real time throughout the experiment.

In this study, 50 ppb NOx and 250 ppb α-pinene (Sigma-Aldrich) was used to form non-absorbing SOA; this route to forming non-absorbing SOA is well characterised and has been reported previously (Nakayama et al., 2010). Here, we only focus on the methodology used to produce the brown carbon SOA. Previous work has reported “brown” SOA formation through oxidation of aromatic precursors under high-NOx conditions (Laskin et al., 2015). In this study, we injected ∼ 400 ppb cresol (Sigma-Aldrich) and ∼ 15 ppb NO2 (∼ 60 ppb NOx in the chamber) into the chamber and removed the UV filter from one of the Xenon arc lamps to increase photochemistry and accelerate and enhance the “brown” SOA formation.

For the purpose of testing the influence of organic coatings and relative humidity on the structure of the BC particles, a series of experiments were conducted in this study. First, bare BC particles with minimal accompanying VOCs were injected into the chamber by using an oxidising catalytic converter, heated ejector dilutors, a catalytic stripper, Purafil, and activated charcoal denuders between the exhaust line of the engine and the chamber inlet. After the BC reached the desired concentration, BC injection was stopped. The particle size distribution of the dried bare BC particles was scanned by the AAC. In addition, the particle size distribution of the dried bare BC particles following “humidity cycling”, i.e. exposure to 90 % RH for around 10 s and then dehydration to 10 % RH, was measured by AAC for comparison. Afterwards, we injected 50 ppb NOx and 250 ppb α-pinene into the chamber. Photochemical reaction was initiated by turning the chamber lights on, and organic materials condensed onto the BC particles. After the condensed organics equilibrate with the surrounding VOCs and the particles stabilised at a certain size (295 nm in aerodynamic size), the aerodynamic particle size distribution of the dried organic-coated BC particles, and that of the dried organic-coated BC particles experienced the humidity cycling, was measured by AAC. Hereafter, a thermal denuder operated at 180 ∘C was added, and the organic coatings of the coated BC particles, that had either experienced the humidity cycling process or not, were removed by the TD, and then the size distribution of the BC core was measured by the AAC and SMPS.

The most common way to calibrate the incandescence channel of SP2 is using monodisperse Aquadag standards (Acheson Inc. USA), which are then corrected for ambient rBC by applying a constant factor of 0.75 based on a previous laboratory comparison (Laborde et al., 2012b). As the incandescence signal is not independent for the different types of BC particles, using a constant factor of 0.75 may represent an uncertainty in the retrieved BC mass concentration. Thus, quantifying the correction factor for many types of BC particles found commonly in the atmosphere may enable better constraints to be placed on this factor depending on the BC source being sampled, thus improving the accuracy of future SP2 measurements of rBC mass concentrations. The broad range of BC aerosols generated in this study from different combustion sources served as an ideal platform to assess the variation in this correction factor to SP2-derived rBC mass.

In this study, the incandescence signal of the SP2 was measured for BC particles from catalytically stripped diesel engine exhaust emissions, an inverted flame burner and controlled flaming wood combustion, respectively, and compared with that measured from an Aquadag standard. The uncertainties here refer to precision of the fitted parameters reported by the Igor Pro fitting algorithm, based on analysis of residual data. As shown in Fig. 4, for the BC particles emitted from the diesel engine under hot engine and cold idle conditions (Fig. 4a), the slopes of the incandescence signal with BC mass are 0.821 ± 0.002 and 0.794 ± 0.005 times of that measured from the Aquadag standard, respectively. Note that while some deviation from a perfect linear response is noted, this is small compared to the variation in slopes, so it represents a minor source of uncertainty in comparison. These correction factors are 9.4 % and 5.6 % different, with the common value of 0.75 (with the uncertainty less than 5 %) recommended by Laborde et al. (2012b) when deriving the mass concentration of BC emitted from diesel engines. For the BC particles generated from the flame burner (Fig. 4b), the correction factor is 0.879 ± 0.003. Meanwhile, for the BC particles emitted from the flaming phase during the combustion of Scots pine, poplar, Giant Redwood or Western red cedar, the correction factors are 0.913 ± 0.009, 0.906 ± 0.014, 0.889 ± 0.027 or 0.843 ± 0.028, respectively. We stress that, for the SP2 calibrations here from wood combustion emissions, the BC particles were not treated with a catalytic stripper before sampling by the SP2. While coating materials may char under 1064 nm to produce refractory black carbon and therefore cause overestimates in the incandescence signal (Sedlacek et al., 2018), as shown in Fig. 2c, the BC particles generated at the beginning of the flaming phase contained almost no organic species, with rOA values less than 0.05. Even if this organic carbon (OC) were to be converted to elemental carbon (EC) with 100 % efficiency (which we consider to be highly unlikely), this would represent a very small error.

The differences in the correction factors derived in this study with the default value of 0.75 are 9.4 % (5.6 %), 17.2 % and 12.4 %–21.7 % for the BC particles emitted from an engine with hot engine (or cold idle) condition, a flame burner and wood combustion, respectively. We recommend that future studies utilising the SP2 for rBC mass concentration measurements use SP2 calibrations with the same type of BC as that to be studied.