Technical Note: Pyrolysis principles explain time-resolved organic aerosol release from biomass burning

. Emission of organic aerosol (OA) from wood combustion is not well constrained; understanding the governing factors of OA emissions would aid in explaining the reported variability. Pyrolysis of the wood during combustion is the process that produces and releases OA precursors. We performed controlled pyrolysis experiments at representative combustion conditions. The conditions changed were the temperature, wood length, wood moisture content, and wood type. The mass loss of the wood, the particle concentrations, and light gas concentrations were measured continuously. The experiments were 5 repeatable as shown by a single experiment, performed nine times, in which the real-time particle concentration varied by a maximum of 20 % . Higher temperatures increased the mass loss rate and the released concentration of gases and particles. Large wood size had a lower yield of particles than the small size because of higher mass transfer resistance. Reactions outside the wood became important between 500 and 600 ◦ C. Elevated moisture content reduced product formation because heat received was shared between pyrolysis reactions and moisture evaporation. The thermophysical properties, especially the 10 thermal diffusivity, of wood controlled the difference in the mass loss rate and emission among seven wood types. This work demonstrates that OA emission from wood pyrolysis is a deterministic process that depends on transport phenomena.

Emission of OA from biomass is observed to be variable and chaotic, making the determination of definite emission factors 20 difficult (Nielsen et al., 2017;Shrivastava et al., 2006;Jolleys et al., 2014). To explain measurement variability, the emission behavior has been related to combustion parameters. The most used parameters are heat flux (Haslett et al., 2018;Eriksson et al., 2014), wood type (Weimer et al., 2008;McDonald et al., 2000) whether softwood or hardwood (Schmidl et al., 2008; In thermally thick pyrolysis, heat transfer in and mass transfer out of the wood matrix are important limiting factors. This type of pyrolysis is less often explored because of its limited relevance to industrial processes, although decomposition rates of thermally thick wood have been examined because they affect fire safety (Tran and White, 1992;Lee et al., 1977). Samples 60 investigated are commonly 1-3 cm, with cylindrical and spherical shapes (Remacha et al., 2018;Bennadji et al., 2013;Di Blasi et al., 2001b;Ding et al., 2018b;Gauthier et al., 2013). Differences in time-dependent mass loss and in overall product yield occur over a range of wood sizes, shapes, and temperature or heat flux to which they are exposed. Throughout these works, larger sizes and non-rounded shapes have not been investigated to ascertain whether known pyrolysis principles are sufficient to explain behavior in real-world applications. Furthermore, time-dependent release occurs because of heat and mass transfer Figure 1 is a schematic of the experiment. The pyrolysis reactor was described in detail in previous work (Fawaz et al., 2020(Fawaz et al., ) et al., 2010. Chemically, the difference between softwood and hardwood is less clear (Di Blasi et al., 2001a). The main structural groups of wood are cell wall material, extractives, and ash. Hemicellulose, cellulose and lignin are the components 125 of cell wall material, volatile sugar and acids are the major components of extractives, and inorganic oxides make up ash (Reed, 2002).
Two size dimensions of wood were used; 14 cm x 3.8 cm x 2.9 cm are called in this work "large", and 2.9 cm x 2.9 cm x 2.9 cm are called "small". The wood was cut longitudinally with the the grain as shown in Fig S2. The large wood size was the same order of magnitude as those used in wood cookstoves, and the small size was used to evaluate the effect of the length of 130 the wood on emission. Both wood sizes are considered thermally thick, a condition in which a thermal gradient forms between the surface and the center when the wood is exposed to heat from its surroundings (Pyle and Zaror, 1984).
Two wood moisture contents (MC) were used for pineR: dry samples at 8±1% and wet wood at 25±2%. Dry wood samples used were received as kiln dried wood and the MC of the samples was between 7-9%. The wood MC was increased by soaking pineR wood samples for 24 hours based on the method of Lee and Diehl (1981). The MC of the soaked wood was measured 135 in four internal positions to confirm the permeation of water and a homogeneous distribution of water throughout the wood.

Experimental Design
Two groups of experiments were performed. The first group varied pyrolysis conditions of reactor temperature, wood moisture content and wood size for two wood types: birch and pineR. PineR is a low-density softwood and birch is a high-density hardwood. These two woods were chosen to evaluate whether the response to changing input conditions depends on wood 140 type. The second group of experiments explored how emissions from pyrolysis change based on the type of the wood. Seven wood types, including Birch and PineR, were used in constant temperature experiments at 400, 500, and 600 • C. This range was selected because at temperatures less than 400 • C pyrolysis was extremely slow and did not occur at certain conditions, and at temperatures higher than 600 • C secondary reactions consumed OA precursors before formation. In combustion conditions, temperatures higher than 600 • C are associated with flaming combustion. The conditions of all the experiments used in the 145 analysis are shown in Table S1.

Reactor Operation
The experiments were performed under isothermal conditions; the wall temperature of the reactor was maintained at a set point temperature of either 400, 500, or 600 • C. To begin an experiment, the reactor was raised to the set point temperature, after which data collection was initiated and nitrogen gas purged the reactor for 2 minutes at 20 LPM. The wood sample was inserted 150 after the purge period, and this event marked the experiment start time.
After the wood sample was inserted, the reactor was kept partially covered with firebricks to prevent heat loss to the ambient atmosphere. The experiment ended when the residual mass of the solid stopped changing within the uncertainty limit of the mass balance and when the hydrocarbon gas concentration returned to background values measured before the experiment.
Each experiment was replicated to determine the uncertainty.

Analysis
A LabVIEW program collected the thermocouple, mass balance, gas analyzer, and flowmeter data. TSI softwares were used to collect data from the EEPS and DustTrak, and the data streams were synchronized with the signals collected in LabVIEW.
The background concentration of the gases in ambient air was measured before and after the experiment and subtracted from the gas signal during the experiment. We assumed a linear drift in the signal between initial and final measurements. The 160 real-time concentrations were smoothed using a MATLAB moving mean function with an averaging period of 3 secs, unless otherwise stated. Product collection was calculated as the total mass of all measured products divided by the initial mass of the wood. The products measured include the final char mass and the total mass of PyOM, CO, CO 2 , and HC. The yield or mass fraction of each product is the sum of its real-time measured mass divided by the initial mass of the wood. In the elevated MC experiments, calculated mass fractions are provided on an as-received basis and wet basis, where the as-received mass of the 165 wood was the mass before water absorption and the wet mass was after absorption, respectively.
The EEPS measured the number concentration of the particles from 6 -560 nm. A density of 1. Pyrolysis experiments at 400 • C released particles with diameters larger than the upper limit of the EEPS measurement range, causing an underestimation in the mass concentration. When the particle sizes exceeded the size range measured by the EEPS, the mass concentration measured by the DustTrak was used for the total mass of the particles. In section S3, we compare the 175 PyOM concentrations from the EEPS and DustTrak measurements at 400 and 500 • C to evaluate the differences between the two instrument outputs and the adequacy of using the DustTrak data at 400 • C.
The measured particle concentration was multiplied by the dilution ratio of the secondary dilution stage during the experiment. Thus, the PyOM mass concentrations reported here are those that would be measured in the primary dilution stage, so they are directly comparable to the gas concentrations.  Table 1 shows the yields of all measured products for the two groups of the pyrolysis experiments. The yields of PyOM show that 10-30% of the initial wood mass was emitted as particles, depending on the conditions of the experiment. One test (large birch at 500 • C) was repeated nine times, and the average coefficient of variation of the real-time mass concentration of PyOM was 15±3% (Fig S21), indicating that the tests were repeatable. Collected product percentages, excluding water, of all 185 experiments were between 77% and 90% of the wood mass. The lower end of the mass closure was observed at 400 • C because of the loss of dark heavy sticky material onto the reactor. The assumption of the PyOM density did not have a quantified effect on the PyOM yield and mass closure. The real-time modified combustion efficiency of birch experiments at the three temperatures is shown in Figure S24.

Effect of Temperature
190 Figure 2 shows the real-time mass loss rate and mass concentration of emitted gases and PyOM during pyrolysis of large pineR and birch wood at 400, 500, and 600 • C. The time series all have two peaks, although some peaks are more pronounced than others. At each fixed temperature, the mass loss rate, PyOM, and gas concentrations exhibit similar behavior with time but the relative heights of the two peaks differ among temperatures. Figure  The heat the wood receives at the surface initiates reactions and is conducted towards the center of the wood. The surface 205 layers of the wood gain temperature rapidly, reacting and losing mass to form the first peak in emissions. When these layers are depleted the declining portion of the first peak forms. Each location inside the wood receives heat conducted from the surface, and transfers heat toward the wood center. As pyrolysis progresses, the outer layers of the wood become char, which has a lower thermal conductivity than wood. This thermal resistance barrier slows the travel of the elevated temperature zone, creating the portion of the mass-loss curve that appears as a saddle at 500 and 600 • C. The rate of mass loss declines when 210 most of the material has reacted in the heated section, and when unreacted wood is not heated rapidly enough to maintain the same level of product formation. This principle explains the sharp second peak, which occurs when the thermal wave reaches the center. The earlier peak times and shorter overall pyrolysis duration of pineR compared to birch at all temperatures is due to the lower density and greater thermal diffusivity of pineR compared to birch. The higher the thermal diffusivity the faster the heat transfer rate from the surface to the center, making pyrolysis reactions in pineR faster than birch. Birch has a larger 215 overall mass loss; the higher density of birch compared to pineR means there is more mass to react and form products. Mass loss as a function of time and temperature is repeatable (Fig S21) because it is governed by the relative rates of heat transfer, temperature increase, and char formation within the wood.
At all temperatures, the real-time mass concentration of PyOM (Fig 2d-f) had similar real-time features to the mass loss rate.
The concentration of CO, CO 2 , and HC (Fig 2g-o) are also similar in behavior. For example, the slope of the first peak increases 220 with temperature for all measured products. Relative peak heights and the nature of the saddle between peaks changed with Evaporation occurs at 100 • C and pyrolysis reaction rates become significant at temperatures higher than 280 • C (Broido, 1976).
The first peak results from the balance between radiative heat flux at the surface and the formation of char and is most affected by the change in relative rates. At 400 and 500 • C, heat transfer into the wood is slow enough that evaporation at 100 • C occurs before pyrolysis reactions at the surface and the first peak of PyOM emission is reduced to a broad shoulder appeared instead.
At 600 • C, heat transfer is rapid and both pyrolysis and evaporation occur simultaneously at the surface. After the initial peak, 260 continuous internal heat transfer in the wood sustains pyrolysis and product formation and there is little difference between the second peak height in the dry and wet cases.
The sharing of energy between water evaporation and pyrolysis shifts product yields towards char (Beaumont and Schwob, 1984;Peters and Bruch, 2003;Di Blasi et al., 2000) and reduces the reaction temperature. The effect of reaction temperature on yield reduction is evident in the change in yield between 400 • C and 500 • C where more char, less gases, and less PyOM 265 were produced at the lower temperature. Further, the yield of PyOM from pyrolysis of wet pineR on an as received basis was lower than that of pineR at MC=8% at the same reactor temperature (Table 1).
Published findings show conflicting effects of moisture content on particle emission yields. Some observe that higher mois- has a higher ignition delay and a higher critical heat flux for ignition (Simms and Law, 1967), prolonging pyrolysis and CP emission that occur in the absence of oxidation processes. As shown here, the rate of condensable product formation itself may be diminished, and that could reduce the rate of oxidation reactions after ignition (Price-Allison et al., 2019).

Wood type
Seven types of wood were used in pyrolysis experiments at each reactor temperature. Thermal properties of dry wood depend 275 on density (MacLean, 1941), so we present results grouped by density, although there may be other causes of differences.

Figures 4a and 4c show mass loss rate and the real-time mass concentration of PyOM from woods with densities below 600
kg.m −3 , and Fig 4b and 4d shows the same quantities for woods with densities above 600 kg.m −3 . The reactor temperature 500 • C is shown because it is least sensitive to mass transfer resistance (low temperature effect) and secondary reactions (high temperature effect), as discussed previously. For each wood, regardless of density, its real-time PyOM mass concentration and 280 mass loss rate share the same features in terms of pyrolysis duration, peak shape and relative peak heights.
Low-density wood types include softwoods (pineE, pineR, and Douglas fir) and one hardwood (poplar). These wood types had similar behavior, except for Douglas fir with a 20% longer pyrolysis duration. The peak heights in the mass loss rate and real-time mass concentration are within 25% of each other, showing that for this group the pyrolysis behavior is similar.
The high-density wood types include birch, maple, and ipe, all hardwoods. The mass loss rate of the wood types in this group 285 had different pyrolysis durations and ratios between the first and second peak compared to the low-density wood, with a shorter first peak and a larger second peak at the end. The higher second peak, a repeatable feature for high-density wood, is caused by the availability of more unreacted wood mass internally than in low-density wood. Ipe emitted more particles than birch and maple and had a higher yield even though the magnitude of the mass loss rates were comparable among the three wood types. This work's thermochemical approach does not offer the reason behind the high particle emission for ipe compared to other In pyrolysis studies, wood type effect has been distinguished in thermally thin wood based on the fractions of cellulose, hemicellulose, and lignin (Grønli et al., 2002). Di Blasi et al. (2001a) found that the mass loss rate behavior for thermally thick wood can be related to the chemical makeup at low heat flux, and at higher heat fluxes the mass loss rate was explained by transport phenomena. When heat transfer controls pyrolysis and burning, the difference in the mass loss rate of the wood 295 can be explained using the thermal diffusivity of the wood (Spearpoint and Quintiere, 2001). As demonstrated in this work and modeled (Fawaz et al., 2020), the mass loss rate change as a function of wood type and heat flux can be explained by the heat transfer in the wood. Section S4.3 shows that the mass loss rate of birch and pineR can be predicted by the same global kinetics using Gpyro (Lautenberger and Fernandez-Pello, 2009). No studies have investigated the role of the chemical makeup of the wood in the pyrolysis of thermally thick wood. The difference in real-time emissions among each group of wood (low-300 and high-density wood groups) can be attributed to the differences in wood composition. well et al., 2014;Chen et al., 2007). Tabulations for use in atmospheric models have been grouped by ecosystem type (Akagi et al., 2011;Andreae, 2019), sometimes separating by flaming and smoldering (Koppmann et al., 2005). Measurements of wood burning in fireplaces and domestic stoves have provided in-depth chemical composition of gas and particulate emission 305 for different wood species (McDonald et al., 2000;Schauer et al., 2001;Vicente and Alves, 2018). However, these studies could not attribute variations of measured emission yields to wood type (Gonçalves et al., 2011;Ozgen et al., 2014). Temperatures and heating rates in these studies were not reported.

Measurements of emission factors to represent open biomass burning have often been reported for individual wood types (Stock
We have shown that differences in yields of particles and gases could be explained by reactor temperature, moisture content, and wood size, and that the difference in real-time emission among wood types can be explained by the difference in the 310 physical properties of the wood. Varying PyOM and gas emissions from different vegetation types, as widely reported in literature, are likely attributable to these factors rather than to differences among wood species.

Conclusions
The results of the developed approach reported here show that time-dependent emission of particles and gases from pyrolysis is a repeatable and deterministic process. Coupled with model findings, we found that the real-time mass loss rate can be 315 explained by heat and mass transfer processes, and in turn, real-time emission of particles and gases follows the mass loss rate.
The conditions of pyrolysis influence the product yield and the real-time concentrations. We reported here the effect of the heating conditions, wood size, moisture content, and wood type on product emission. Comparative yields of particles and gases can be explained by thermochemical principles that are well known in pyrolysis literature. Increasing temperatures increase mass loss rates and increase the concentration of emitted particles between 400 and 500 • C. At 600 • C the particle 320 concentration decreased, likely due to gas phase secondary reactions outside the wood. Increasing wood size decreased the yield of particles at the lowest temperature due to mass transfer resistance. Increasing moisture content reduced the yield and real-time concentration of particles and gases when drying reactions consumed some of the energy that the wood received for pyrolysis. The mass loss rate of different wood types showed differences between low-density (ρ < 600 kg.m −3 ) and high-density wood (ρ > 600 kg.m −3 ). 325 We have demonstrated that the first step in biomass emission-release of particles and gases from pyrolysis-is predictable and repeatable. The principles demonstrated here have not previously been exploited to guide understanding of biomassburning emissions to the atmosphere. Quantifying relevant physical factors, such as temperature, wood size, and thermal diffusivity of the wood matrix would likely reduce the unexplained variation in reported biomass emission rates. A more predictive understanding would require investigation of the remaining steps in the process, including gas-phase oxidation and 330 feedback from exothermic gas-phase reactions to the solid phase.  Table 1. Average yields of measured products for each replicated experimental condition. Wet pineR yields are expressed on a wet basis (the mass of the wood used included the moisture content) and on an as received basis (the mass of the wood before increasing its moisture). The last column is a semi-quantitative indicator of whether the production of the dark sticky material was detected during the experiment. In the table (-) indicates the lack of production, (+) indicates production with (++) signifying more production than (+).