Thermochemical recovery of waste wood in a domestic wood stove: influence of the geometry of wood batch in the fireplace on pollutant emissions

Abstract In the current energy context, pallet boards represent a wood of opportunity strongly used on an individual scale as a means of heating at a low cost. However, French stoves are certified to burn only hardwood species. The main objective of this work is to study the combustion behavior of pallet boards in a commercial stove designed to burn hardwood. The novelty of this work lies in the study of the influence of the wood surface exposed to fire and the identification of the conditions that promote the production of ultrafine particles. The results of the combustion tests show that the surface exposed to the fire influences the combustion. Of the six wood dispositions tested, two stand out with gaseous and particulate emissions at similar levels of a conventional hornbeam log and a densified log. Low temperature conditions and high levels of unburned gaseous products were identified as promoters of ultrafine particles. Overall conclusion of the study is that it is possible to use pallets in a non-designed stove, provided that the user carefully manages the combustion. This opens the way to the clean and rational use of a new type of fuel in a low carbon circular economy.


Introduction
According to the latest synthesis report published by the Intergovernmental Panel on Climate Change [1], since 1850, human activities have caused unprecedented global warming of the Earth.European environmental initiatives have prompted France to draw up a plan to develop renewable energies, among which biomass energy is the leading contributor [2].This energy is developed on a multi-scale as it is considered a clean and decarbonized energy production source.However, to develop green energy while respecting forestry resources, the wood-energy sector needs to use, as much as possible, a type of fuel other than fresh wood.The order of 29 July 2014, setting the criteria for removing waste status for wood packaging shreds, acts as the necessary condition for the use of end-of-waste wood (EOWW) for energy production [3].EOWW concerns clean waste wood consisting of shredded packaging such as pallets, crates, or any other type of wood packaging that can replace fresh wood for energy production.Old pallets make up the largest portion of wood waste.
Currently, according to the statistical office of the European Union, in France, only 37% of wood packaging is recovered for energy purposes, which leaves a significant amount still unused [4].Even if the EOWW status is mainly dedicated to industrial sites, many individuals use this type of wood in their domestic appliances [5].The main advantage is economic because the pallet is an opportunity wood that can be obtained for free by users.Moreover, the homogeneity of the pallet is an advantage for combustion because of its low moisture content and high calorific value [6,7].The thermal degradation of wood waste has been studied at several scales and in different forms: on a reduced scale with thermogravimetry [8][9][10][11], with a fixed-bed reactor [12,13], by a gasification process at pilot scale [14,15], and in densified form as briquettes or pellets [16,17], or as chips [18].Fachinger et al. [19] worked with different biomass fuels at the scale of the wood stove and observed a high burning rate with a related high emission factor for EURO pallet fuel.They linked this behavior to the small log size of the batch but no further explanation was given concerning the combustion behavior of pallets.Thus, to our knowledge, the literature does not provide complete and detailed studies of the combustion behaviors of pallet plank at the scale of the domestic wood stove.
Due to their composition, the thermal degradation of softwoods is different from that of hardwoods, especially because of the participation of acids from the resin in the formation of aromatic polycyclic products [20][21][22][23].Fuel quality is an important factor in the combustion process and pollutant emissions, and softwood is known to be a major emitter of Particle Matter < 10 mm (PM 10 ), < 2.5 mm (PM 2.5 ), and < 1 mm (PM 1 ) [24].In parallel, the European standard EN 13229 (June 2002) [25, p. 229] on requirements and test methods for open fireplaces and inserts for solid fuels accepts only hardwood species (beech, hornbeam, and birch) as representative wood logs.In addition, the French certification NF 'Biocombustibles solides' (NF 444) offering the consumer a guarantee of a good quality fuel excludes coniferous species (softwood) from the scope of certification.Only hardwood species such as hornbeam, beech, oak, and maple are allowed in the scope.As a result, certified French stoves (like the one used in this study) are designed to burn only hardwood.However, the pallet board is a fuel that is distinguished both by its plate shape and by its composition in terms of wood species (mostly resinous).Previous studies have shown that the use of this fuel in an appliance certified to French standards (not designed for softwood combustion, and representative of the market) can quickly become a generator of pollutants [16,26].It is therefore necessary to study the optimization of the combustion in real conditions.
The pallet board has a shape that differs from the classic log in that it has a large surface area in relation to the thickness of the wood.Kuznetsov et al. [27] experimentally studied the influence of the shape and size of wood particles on their ignition during heating at high temperature in an oxidizing environment.They concluded that the shape of the particles has a significant effect on the characteristics and conditions of ignition.Furthermore, the results show that plate-shaped particles ignite faster (than rectangular and cubic ones) because they heat up faster thanks to their large heat exchange surface.Several authors highlighted the influence of the specific surface of the wood on the combustion reaction, using different fuels and combustion devices [12,[28][29][30][31][32][33][34].P erez et al. [12] showed that the specific surface of a wood particle has an impact on the combustion rate in a fixed-bed reactor.According to these authors, the surface rules the heat transfer inside the particle and thus its combustion rate.These observations were confirmed by Yang et al. [33], who studied the combustion of pinewood of several particle sizes in a packed bed.Porteiro et al. [30] studied the combustion of wood pellets in an experimental fixed-bed reactor.According to these authors, heat diffusivity takes place in large wood particles reducing the heat transfer between particles in the fuel bed.Caposciutti et al. [32] studied three sizes of poplar woodchips at pilot scale in a 140 kW boiler.Results showed a positive correlation between the size of a wood particle and the mass fraction of pollutants in the volatile fraction released.They also highlighted a difference in wood consumption: large pieces remained partially unburned at the end of the fuel-bed while small pieces were fully burned.
Based on our current understanding, no previous research has been conducted on how the surface of pallet wood affects combustion behavior at the scale of a wood stove.The aim of this study is to enhance the combustion process of pallets in a household appliance that was not originally intended to burn softwoods.The study aims to replicate real-life usage conditions to minimize the emission of atmospheric pollutants.To achieve this goal, the investigation focuses on examining how the surface of the pallet wood influences the release of pollutants.For the individual, modifying the surface of the board is a simple lever to properly burn the pallets in a French stove, thus reducing the environmental impact of home heating.

Fuel and characterization
To simulate the combustion of pallet wood, fir boards (softwood) of dimensions (L Â W Â D) 2.5 Â 14.4 Â 33.4 cm and 2.5 Â 9.7 Â 33.4 cm were used.This type of wood was chosen as it is the major component of a pallet.Characterization tests were carried out to determine the proximate characteristics of our sample of wood.The moisture was determined in an oven at 105 C following the EN 14774 standard.Before ash and higher calorific value (HCV) characterization tests, the biomass sample was dried and milled (mean diameter < 4 mm) using a Retsch SM300 knife shredder.Ash content was measured in a muffle furnace at 550 C by the application of the EN 14775 standard.The BS EN 14918 standard was used to determine the HCV using an IKA C200 calorimeter.Results show a low ash content in the fuel of 0.19 wt.% (on a dry basis) and an HCV (18.1 kJ/g on a dry basis) in the range of what is presented in the literature [8,9,28,[35][36][37][38][39][40][41].The moisture content was 10.07 wt.% (on a raw basis).

Experimental platform and acquisition equipment
Combustion tests were performed in a domestic wood stove 'LORFLAM XP-68 IN' with a nominal output of 8 to 14 kW and certified by the French label 'Flamme verte 7 Ã ' which ensures the respect of pollutant emission thresholds.The device is sealed and ducted, made of cast iron with a natural draw (no control or reading flow meter is installed on the air inlet of the unit).Resinous fuels are not recommended by the manufacturer for this device.Three air stages are available in the firebox.The first is situated under the grate and the second one takes place on the side wall of the firebox.They are both adjustable with a lever.The air supply is sourced from the ambient air and is preheated by the hot walls of the wood stove before entering the firebox.While the first two air supplies are used for combustion, the third one, located at the window, is used to keep this last clean.A post-combustion system called 'C-2 box' is situated above the firebox to increase the gas and smoke path.The wood stove is extended by a chimney covered with insulation represented on the diagram of the experimental platform in Figure 1.Sampling probes are positioned at the center of the cross section of the chimney stack and all sampling lines are heated to 130 C and this temperature is maintained throughout the experiment.There is no dilution system installed in the chimney stack and smoke ends its course by passing through an exit situated on the rooftop of the laboratory.The depression inside the chimney is acquired over time.The whole combustion platform is placed on a weighing scale in order to follow the mass loss.There are three thermocouples (K type, ± 1.5 C) situated at the beginning of the chimney stack, at the front of the firebox, and at the rear of the firebox.
Gaseous emissions were measured from the beginning of the test (ignition) to the end (extinction).Acquisition of CO 2 , CO, O 2 , SO 2 , NO and NO 2 is carried out with ROSEMOUNT NGA 200 analysers.Uncertainty is considered to be 1% of the full scale of the gas concerned (CO 2 : 0.2%; CO: 0.01%; O 2 : 0.25%; SO 2 : 10 ppm; NO: 10 ppm; NO 2 : 10 ppm).Detailed data for SO 2 , NO and NO 2 emissions are not shown here.In fact, SO 2 and NO 2 concentrations measured here are below the uncertainty of the analyser.With respect to NO emission, in the literature, it is known that NO x emissions from wood combustion can be calculated as Fuel À NO x [17,42].Total hydrocarbon emissions (THC) are measured with a JUM 109 L analyser.Data from this analyser are expressed in CH 4 equivalent.Particle emissions are measured by means of an electrical low-pressure impactor (ELPI) from Dekati preceded by a fine particle sampler (FPS) diluter.The FPS system dilutes smoke with a dried and heated air supply to preserve particles from coagulation and condensation effects.So, particle emissions and particle size distribution (PSD) are representative of the smoke leaving the firebox.Mass particle emissions are collected following the EN 16510 standard using borosilicate glass fiber filters (Whatman EPM 2000).The sampling rate is maintained constant at 0.51 Nm 3 /h.At the end of the particle collection, filters are dried in an oven at 110 C for 24 h before weighing.

Test set-up
To determine the impact of the wood layout by increasing the area exposed to the combustion, six different board layouts were tested.Table 1 presents photographs and surface, weight, and density data of the six dispositions tested, with a constant geometry from D1 to D3 and a constant mass from D3 to D6.The reactive surface (the wood exposed to the hot thermal environment of the firebox) of the wood load increases from D1 to D6.For repeatability, at least three nominal phases were conducted for each disposition.A combustion test is carried out by the ignition of 2.3 kg of cut wood (the same wood used in all nominal batches) with the reverse ignition technique and maximum air supply.After ignition, a first batch (2.5 kg) called the 'nominal setting' is conducted to bring the combustion device to stable conditions.For all tests, the D5 layout was used to realize this first batch because of its great stability results.Following these two sequences, three consecutive nominal batches are conducted with a nominal air supply.The nominal batch is considered to have ended when the outlet carbon dioxide concentration falls below 4%.At this moment, embers are poked before adding a new batch of fresh wood to the firebox.At the end of these three batches, the extinction phase begins, in which the smouldering conditions are allowed to continue until the carbon dioxide concentration in the stack falls below 2%.

Data processing
Acquisition of data starts at the ignition of the starter batch.Except for total suspended particles (TSP), data are collected continuously from ignition to extinction of the fuel.All the gaseous emission results are expressed in the form of an emission factor (EF) in milligrams per kilogram of dry wood (mg/kg dry wood ).In accordance with standard NF EN 14785, all pollutant emissions were corrected to 13% of O 2 before any analysis or treatment.The particle emissions are expressed as an EF in the form of numbers of particle per kilogram of dry wood (part/kg dry wood ).Mass particle emission results are expressed as milligrams per kilogram of dry wood (mg/kg dry wood ).A variability coefficient is calculated (standard deviation divided by the average) for each result.Theoretical power output for a given period is calculated thanks to the mass loss and the high heating value (HHV) of the fuel.In order to compare multiple parameters to determine whether they are correlated, several correlation tables are calculated.Each correlation table includes data from all combustion tests representing the results of 18 nominal batches.Tables take the form of a matrix presenting several coefficients between À1 and þ1.The more a coefficient tends to À1, the more the two parameters studied are inversely correlated.In the same way, if the result tends to þ1, the two parameters are positively correlated.No correlation is observed in the case of a coefficient close to 0.
Calculations to analyze the production of ultrafine particles of each wood layout were also conducted.It provides information on the behaviors of the combustion tests and has the purpose of confirming or invalidating several observations by taking into account all emission values.Three results are investigated: which particulate aerodynamic diameter (Dae) is the most emitted, how much of the combustion time it represents, and during what part of the combustion ultrafine particles (20, 40, and 70 nm) are the most emitted.Then, the time during which a given diameter was most emitted (DME) can be summed and reported tfor the total number of values.In addition, ultrafine particles (20, 40, and 70 nm) are targeted to evaluate their contribution to particle emission.Equations ( 1) and ( 2) show the calculations from the DME table: For the sake of clarity, only the results for nominal batches will be discussed in the following.The values presented correspond to an average of three nominal batches.

Results
Table 2 shows the average gaseous and particulate emission results for the nominal batches for all combustion tests as well as the results of a previous study [26] conducted using the same stove.In general, layouts D1, D2, D3, and D6 showed the highest EFs for unburned products and particles.Conversely, for all parameters except TSP, the EF results are lower for D4 and D5.
The CO 2 and O 2 EFs of the present study (6.6-8.7% and 11.4-13.3%,respectively; see Table 2) are consistent with prior investigations [43,44] where CO 2 EF ranges from 6.0 to 11% and O 2 EF ranges from 9.6 to 14.5% were measured (detailed data available in the Supplementary data).The same is found for CO and CH 4 , even though various articles in the literature report wood combustion by varying settings and multiple stove generations.Particle masses measured in the present study (163-409 mg/kg dry wood or 13-34 mg/Nm 3 ) are close to the modern stove emissions reported by Brandelet et al. [43] (6-44 mg/Nm 3 ; detailed data available in the Supplementary data).However, particle EF values are much lower than those found in the literature ; this may result from differences in sampling methods.

Gas emissions
Analyses of carbon dioxide and carbon monoxide emissions allow the evaluation of combustion efficiency by determining the level of volatiles converted during the thermochemical reaction.Results show higher CO 2 concentration in the wood smoke for the last three arrangements, indicating a higher combustion rate.The lowest CO EF is observed for D4.It is interesting to note that D5 and D6 present high levels of CO 2 and CO.CO emissions of fir plank used in a previous study [26] are on the same order of magnitude as those resulting from the D2, D3 and D6 layout combustion tests (Table 2).D4 produces less CO than the combustion of densified log, and CO on the same order of magnitude as that of the hornbeam log (22,079 ; 33,357 ; and 21,091 mg/kg drywood , respectively).It can be noted that for all arrangements a peak of CO is observed as long as the temperature of the firebox has not reached a higher value (T $ 550-570 C).The beginning of the smouldering conditions was also clearly visible with the increase in CO emission followed by a stabilization (typical run available in the Supplementary data).Similar trends are observed in the literature [28].
Hydrocarbon emissions are important factors for wood combustion, especially for conifer as it contains terpenoids (molecules coming from resins) which can decompose into several molecules comprising the polycyclic aromatic hydrocarbon (PAH) retene [20,21,35,45].For example, terpenes are released by pyrolysis of resins and can be converted to retene leading to soot formation by 'H abstraction-C 2 H 2 addition' (HACA) or 'cyclopentadienyl' (CPDyl) [21,35,36,45].In our test, the resin was observed coming out of the interstices of fresh wood planks and bubbled when subjected to the hot thermal conditions of the fireplace.The lowest THC emissions (597 mg/kg dry wood ) are observed for D4 while the highest emissions (3900 mg/kg dry wood ) are observed for D6.As with CO, a peak of hydrocarbons occurred just after reloading with a fresh batch of wood.This is due to the lower temperature which is not sufficient for correct oxidation of the first volatiles [28,34,40,[46][47][48].There is another THC emission peak at the end of the batch when the smouldering conditions begin but, unlike CO, the peak does no't reach a plateau but falls after reaching a maximum.Hydrocarbon emissions from burning fir boards in the previous study [26] are much higher than the present tests, which found an average of 10,686 mg/kg dry wood .Results for D1, D2, D3, and D6 are on the same order of magnitude as those observed for hornbeam and densified log, while the lowest THC emissions are found for D4 and D5.

Number, mass and size distribution
Measured PM and TSP EF are given in Table 2. Figure 2 shows the mean number of particulate matter (PM) emissions for PM 2.5 , PM 1-2.5 , PM 0.1-1 , and PM 0.1 .Regardless of their diameter, PM emissions tend to decrease as the surface area of wood exposed to fire increases, up to 2975 cm 2 (Figure D5).Above this wood surface area, PM emissions increase again.D5 shows the lowest results for all particle diameters and D1 the highest.As for gaseous pollutants (in particular CO and Hydrocarbons (HCs)), D4 and D5 show low particle EF.The evolution of these particulate emissions, regarding their disposition in the shape of an inverted parabola (Figure 2), seems to show that there is an optimal surface for this combustion device.Interestingly, unlike the number of particulate emissions, the mass of particulate emissions seems to increase with the increase in surface area of the wood arrangement (Table 2).
Similarly to CO and HCs, the emission in terms of number of particles is higher at the ignition of the batch of fresh wood as long as the firebox environment has not reached a sufficient temperature.During this ignition phase, several types of particles can be emitted simultaneously such as partially decomposed pieces of wood, unburned products, soot agglomerates, and fly ash [47].Smouldering combustion (at the end of the nominal batch) showed, for all arrangements, the same trend, with PM of 20, 40, 70, and 120 nm diameters being the most emitted.The literature presents comparable findings and suggests that these particles are fly ash from the remaining glowing char [47,49].Figure 2 compares the emission of PM 2.5 in the present study with those in a previous study [26].D1, D2, and D6 show similar emissions levels (1.31 Â 10 14 to 2.19 Â 10 14 p/kg dry wood ) to those of fir plank in the previous study using the same stove (1.76 Â 10 14 p/kg dry wood ).D4 and D5 (5.29 Â 10 13 to 2.72 Â 10 13 p/kg dry wood ) have lower EFs that fall within the range of hornbeam wood and densified log (which are 3.58 Â 10 13 and 4.60 Â 10 13 p/kg dry wood , respectively). Figure 3 shows the PSDs of D3 (which is representative of the PSD of D1, D2, and D6) and D4 (which is representative of the PSD of D5).Conversely to D4 and D5, all arrangements show a bimodal PSD centered on ultrafine and fine particles.Arrangements D4 and D5 stand out from the others in having a unimodal PSD centered only on fine particles at 100 nm.This clearly highlights that ultrafine particle emissions are drastically decreased under favorable combustion conditions (i.e. higher temperature) as obtained for D4 and D5.In addition, this shift to a larger diameter is accompanied by a decrease in particle emission for all diameters.The D4 PSD is very similar to those of the combustion of 11 types of pellets (a mix of hardwood and softwood) tested by Sippula et al. [50] where the authors obtained a unimodal distribution with geometric mean aerodynamic particle sizes between 126 and 188 nm.

Production of ultrafine particles
The following analysis is intended to specify the cleavage between D4 and other layouts concerning ultrafine particle production.It allows us to determine whether there are some diameters that are most emitted (DME) compared with others and for how long they are emitted.Table 3 shows the results for all the wood arrangements.Except for D4, the particle DME is 20 nm for 55% to 89% of the combustion time.A difference in ultrafine particle production is also visible with only 25% of the burn time for the D4 test where ultrafine particles are prevalent in terms of   numbers.It seems that low firebox temperature and high concentration of unburned products promote the production of ultrafine particles.Conversely, the combustion of the D4 layout seems to be different, with a higher diameter of particle emitted (DME of 120 nm during 74% of the burn time).

Power of the reaction
Figure 4 shows a comparison of THC emissions between two nominal batches (D4 and D6).The instantaneous power released by combustion is associated with emissions.The time distribution shows that the power released is not constant over time.With respect to the D6 layout, the fast volatilization phase at the first third of the combustion time generates a power that largely surpasses the nominal power of the device (8-14 kW) and leads to high THC emissions.The two highest peaks of THC emissions for layout D6 (A) are associated with a power release of 22.6 and 30.2 kW which is double the preconized values (given by the constructor) for this stove.In addition, the power decreases rapidly (as does temperature) during the following two thirds of the combustion time.These two facts show that the power released by the reaction has not been evenly distributed all along the nominal batches.In contrast to this wood layout, the D4 (Figure 4B) combustion test presents much more stable volatiles releasing with a power situated in the nominal power of the stove device.

Category of combustion behaviors
Results of the combustion tests show that the surface exposed to the fire influences the combustion in the firebox.Distinct combustion behaviors can be classified into three groups: These categories will be used in the following analyses and discussion.

Correlations between combustion parameters
The correlation analysis is used to check several observations based on emissions data and the quality of the acquisition system.Parameters used to realize the correlation matrix were: back firebox temperature (T bf ), total hydrocarbons (THC), CO 2 , CO, and O 2 emissions.Table 4 shows the results of the correlation matrix.
The strongest coefficient is attributable to T bf /CO 2 with þ0.93 to þ0.97.This is not surprising as these two factors are indicators of complete combustion conditions and thus, this indicates that T bf values can be considered a good indicator for combustion efficiency.Strong coefficients of T bf and O 2 confirm this conclusion.There is a negative coefficient for T bf /CO, of about À0.69 to À0.88 (categories I and II).A more complete combustion will produce CO 2 instead of THC and CO by correctly oxidizing volatiles.Conversely, when incomplete combustion conditions occur, volatiles cannot be destroyed and less CO 2 is produced which benefits THC and CO creation.Sippula et al. [50] present, for a top-feed pellet stove, a link between fire  chamber outlet temperature, oxygen, and CO, since when the burning rate is low, the production of CO is high.The strong positive correlation for THC/CO confirms that THC is an indicator of incomplete combustion conditions.In the case of D6, the strong volatilization followed by the decrease in the firebox temperature leads to the emission of large quantities of HCs due to poor combustion conditions.These events following each other make the coefficients THC/T bf , THC/CO 2 , and THC/O 2 tend to zero, whereas they would be positive during the runaway and negative during the last 2/3 of the nominal batch.

Correlations of gas/particles
When comparing gaseous and particulate emissions as a function of the combustion time, recurring behaviors can be observed: An increase in T bf and CO 2 with a decrease in CO and THC that led to a decrease in PM of 20, 40, and 70 nm jointly with an increase in PM of higher diameters.An increase in CO and THC with a decrease in T bf and CO 2 that led to an increase in PM of 20, 40, 70, and 120 nm.
To discuss the potential correlation between combustion parameters and the emission of particles, correlation matrices were created.Two parameters were analyzed: T bf and THC (judged relevant for their influence on particle emissions).

Hydrocarbons and particles
Table 5 shows the results of the correlation matrix that allow us to analyze the relationship between THCs and particle emission (for each ELPI impaction step).
Layouts D1, D2 and D3.Results for D1 show a low positive coefficient for ultrafine particles ($0.09 to þ0.30).Particle diameters having the strongest coefficients are 120 and 200 nm with þ0.59 and þ0.50.The largest diameters have only slightly higher values than ultrafine diameters.Results for the D2 arrangement show the same tendency as for D1 but with slightly higher values for all diameters.The arrangement D3 presents higher coefficient values for ultrafine particles (coming from the premature extinction of the wood load generating high THCs and 20 nm particle diameter emissions) but it is again the 120 and 200 nm particle diameters that have the highest coefficients.
In general, a positive correlation coefficient for all particle diameters can be observed, with maximum coefficient values for diameters around 120 and 200 nm.
Layouts D4 and D5.There are two distinct behaviors for this category.With respect to D4, a positive correlation between THC and particle size (only for the ultrafine diameter) can be observed: coefficients for 20, 40, and 70 nm of around þ0.50.For the other diameters, there is either no correlation or a slightly negative coefficient.For D5, there are low coefficients for ultrafine particles, explained by the long heterogeneous combustion phase.In general, the absence of correlation for the major diameters confirms that the more favorable combustion (i.e.adequate air supply and higher temperature) conditions seem to lead to the emission of particles of a different type than ultrafine ones.
Layout D6.A positive correlation coefficient for all particle diameters can be observed.The highest values are attributed to ultrafine particles at 20, 40, 70 nm and fine particles at 120 nm with coefficients around þ0.60.

Temperature and particles
As for hydrocarbon products, the higher temperature seems to be linked to the emission of the highest particle diameter.Table 6 shows the results of the correlation matrix for temperature and the several impaction stages of the ELPI.

Layouts D1, D2 and D3.
There are two distinct phases for D1: positive correlation for ultrafine particles and negative correlation for the highest diameters.The comparison of emission versus time reveals that ultrafine emissions follow the increase in temperature.D2 and D3 show negative values for nearly all diameters.
Layouts D4 and D5.With respect to D4 (which has the lowest results fot THC but not for PM number), the tendency seems to be the opposite of that in category I. Temperature appears to produce a decrease in ultrafine particles, especially for 20 and 40 nm.The biggest particles show a positive correlation coefficient centered around 200 to 760 nm.This trend is confirmed in the D5 arrangement Layout D6.This combustion test shows a scheme where no correlation is visible for the finest particles.With respect to the highest particle diameters, a positive coefficient value for 320 nm is noted.

Discussion
Category I: D1, D2, and D3 Among all of the results, D1, D2, and D3 have the lowest average stack temperatures, at 297, 302, and 294 C, respectively (Table 2).This is a direct consequence of the wood surface area being exposed to the combustion.For these three arrangements, the surface area of the wood exposed is high enough to create volatile-releasing conditions but not enough to maintain high temperatures.Therefore, some of these volatiles will not be degraded and will become unburned products.P erez et al. observe, by varying the operational parameters of a gasifier, that the increase in particle size has a significant effect on the thermochemical process.Indeed, results show that the larger the size of the wood particle, the lower the flame front velocity.Similarly, the maximum temperatures reached decrease with increasing particle size.This is linked to the effective surface area of the particle available for the reaction.Since the surface area is reduced, the mass exchange and heat transfer in the thermochemical process will also be low [12].This is the reason why it is possible to observe high HC and CO EFs, which are indicators of incomplete combustion conditions.This is consistent with the negative correlation coefficient for T bf /THC and THC/CO 2 (Table 4), indicating that THC emissions are primarily from lower volatile oxidation.The combustion efficiency also impacts particle production.The first category of combustion tests has the highest particulate emissions and is characterized by a high emission of ultrafine particles.This can be seen in the PSD where a bimodal mode was present.This represents up to 89% of the combustion time where 20, 40, and 70 nm are the most emitted particle diameters.Concerning PM 120 and 200 nm, the correlation results indicate that they appear to be closely related to THC (Table 5).This is consistent with the correlation coefficients of temperature and PM.Negative coefficient values seem to indicate that the increase in temperature inside the firebox brings the energy required to destroy the particles.Thus, these seem to be unburned products like soot agglomerate or hydrocarbon condensates.Other studies point out that particles from incomplete combustion contain high levels of elemental carbons and organic carbons [38,41,45,[51][52][53][54][55].Our results are consistent with those of Vincente et al. [28] who measured PM 10 emission for several wood batch sizes in a Portuguese stove.They showed that the highest emissions rates are measured for the lowest wood loads.According to those authors, condensation and agglomeration processes occur in poor combustion conditions (low temperature and bad mixing) which generate an increase in the size of particles emitted.This is also consistent with Torvela et al. [52], who obtained higher PM 1 emissions for poorer combustion conditions (low load and fuel-rich).It is further supported by Ozgen et al. [40] who studied the relation between particle emissions and combustion products using multiple regression analysis.According to their results, particle diameters of 39-488 nm come from elemental soot with the addition of organic gaseous carbon condensation on particles when the temperature is lower.
The D3 arrangement was chosen to represent a reconstituted log by overlaying four planks of wood.This configuration blocked the thermal conduction of heat inside the arrangement, leading to a premature end of the flame period with consequences such as the release of a great deal of unburned products like CO, HCs, and particles.Poking the burned wood revealed that the center of the arrangement remained unburned and intact.This thermal gradient phenomenon is known in the literature for other fuels [29,30,32].The difference seen with other arrangements resides in their capacity to bend which allows the exposure of fresh intact wood to the heat of the combustion chamber.Thus, the wood can gradually degas volatiles during the load.This highlights the fact that, with a stove certified for hardwood combustion, pallet plank combustion is singular and cannot be considered entirely the same as the combustion of a traditional wood log.

Category II: D4 and D5
The increase in CO 2 concentration in the wood smoke indicates that the combustion rate was higher, and this is directly connected to the surface area.When the surface area of wood exposed to the combustion increases, volatiles released by the wood batch increase too.As richness (fuel/oxidant ratio) is higher, the heat released by the reaction increases, allowing more volatiles to be oxidized and giving rise to a higher combustion rate.This is supported by the work of Porteiro et al. who studied the parameters influencing the propagation of the ignition front in a fixed-bed combustor.According to their results, there is a significant difference in reaction rate between large and small particles.Small particles are more affected by surface reactions like thermal radiation because of their high surface/volume ratio, leading to higher volatilization rates than larger particles [30].In D4 and D5, the increase in combustion rate can be observed by the higher stack temperature (Table 2).It is also visible from the power released by the reaction.With respect to D4 and D5, the release of volatiles was more stable over time, generating power in the range of the stove's rated power (Figure 4).Therefore, with better combustion quality, unburned products have drastically decreased as suggested by the low level of hydrocarbon emissions (Table 2).Bhattu et al. [46] concluded, by studying several stove technologies and combustion conditions, that for all devices, an increase in the temperature of the combustion chamber leads to a decrease in CO, CH 4 , and non-methane volatile organic compounds.The lowest value of T bf /THC (r ¼ À0.38) (Table 4) means that all unburned products are not necessarily coming from the decrease in the temperature but are due to other factors such as mixing, residence time, or richness.The efficiency of the combustion also has an effect on the correlation between hydrocarbon and particulate emissions.First, positive coefficients for finest diameters (r around þ0.50 for D4) (Table 5) prove that, in our cases, hydrocarbon products can participate actively in ultrafine particle formation.Ozgen et al. [37] showed in their combustion tests that gaseous organic carbon explained 71% of the nanoparticle variance (range 7-39 nm).This is in accordance with pictures of soot agglomerates (primary particle size of 30-50 nm) made by Torvela et al. [52], which show among three types of soot, the amorphous type with a coarse surface where semi-volatile hydrocarbons seem to have been absorbed.Tissari et al. [53] describe the range of 6-30 nm as the nucleation mode.They observed this type of particle in conditions of incomplete combustion.In such an environment, the high concentration of volatile organic compounds favors homogeneous condensation into liquid hydrocarbon droplets that constitute the analyzed nuclei.It can be assumed that under favorable and stable combustion conditions, condensation mechanisms are avoided, relegating hydrocarbons to the formation of ultrafine particles which are significantly reduced.In fact, unlike all other wood layouts, D4 and D5 show a unimodal PSD centered on 100 nm fine particles.The decrease in ultrafine particles is more visible with layout D4 for which during 25% of the combustion time ultrafine are the most PM emitted, versus 64-89% for the other arrangements (Table 3).This is visible in the correlation matrix between temperature and particle emissions (Table 6) where the results show that only as the temperature decreases does the ultrafine increase (the value is stronger for D5 due to the heterogeneous reaction).
The second fact is that the larger diameter particles are emitted throughout the burn time and appear to be related to the burn rate.It is well known that favorable combustion conditions promote inorganic particle emissions, in contrast to incomplete combustion conditions, which generate unburned products [38,41,[52][53][54].These results highlight the difference between two types of particles: ultrafine assimilated to unburned products (soot, hydrocarbon droplets) and larger particles assimilated to inorganic products (minerals or fly ash).According to Stanmore et al. [56] temperature is a key factor for the soot oxidation mechanism.Lea-Langton et al. [36] studied the smoke formation of several burning wood components.They could observe soot formation constituted by individual spherules from 20 nm to 60 nm depending on the nature of the component burned.Lamberg et al. [41] tested several combustion condition cases including an efficient combustion obtained with a pellet boiler using pellets mainly made of pine wood.They observed that in contrast to incomplete combustion conditions, the efficient one was promoting particles whose chemical composition is mainly inorganic alkali metals (with K þ , SO 4 2À , Cl À , and Na þ as the most abundant chemical species).In the same way, Tissari et al. [53] observed, for a modern device, fine particles with low organic compounds but high inorganic species.They state that this indicates much better combustion in comparison to another device studied.In contrast, Torvela et al. [52] obtained a particle mean diameter of 25 nm for efficient combustion with a unimodal distribution centered on ultrafine particles, while intermediate and smouldering conditions produced a bimodal distribution centered on ultrafine particles and the accumulation mode.The high level of CO, combined with the low concentration of hydrocarbons in the D5 combustion test, is the consequence of a long period of heterogeneous combustion.It is known that a high level of CO is produced in smouldering combustion conditions [57].So, despite a flame phase with low emissions, this disposition of wood generates high CO levels.Thus, we could optimize global emissions for these arrangements (especially D5) by deleting the heterogeneous phase, by poking the embers and reloading fresh wood in the stove.
In comparison with a previous study [26], the results showed that D4 and D5 emit pollutants at the same level as those emitted by hornbeam and densified logs (Table 2).Thus, the main conclusion drawn from these wood layouts is that the combustion of fir boards can be adapted to this type of wood stove to achieve acceptable emission levels.

Category III: D6
The particularity of the D6 combustion test is the high temperature and CO 2 concentrations jointly with high levels of CO and hydrocarbon products.The very high surface of D6 causes runaway combustion leading to the release of high quantities of volatiles.The volatilization rate increases with the increase in the specific area available to the reaction.In this period, which happens just after the reloading of the stove, the very low oxygen concentration (< 2%) does not allow the burning of all the volatiles.This results in high levels of unburned products of which CO is the indicator.Yang et al. [33] studied the effect of the fuel size on pinewood combustion in a fixed-bed reactor.One of their main conclusions is that smaller particles lead to richer fuel combustion as the burning rate increases with decreasing particle size (and thus increasing reactive surface area).This leads to high CO and CH 4 concentrations in the flue gases.Oxygen is essential to obtain correct combustion insofar as it allows the cracking of large molecules into smaller ones and also promotes the decomposition of aromatics [39,58].Nystr€ om et al. [59] observed, by burning large batches of finely cut logs, short moments of low O 2 concentration (< 2%) and indicate that partial gasification may have occurred during this phase.The role of the high surface area is highlighted by the power released by the combustion.Figure 4 shows that the power produced by the reaction is outside of the range of the nominal power accepted by the stove (the air supply is not mechanized but natural).These runaway conditions provide a lot of unburned products in the wood smoke like heavy tar, lowmolecular-weight gases, or PAHs, resulting in high hydrocarbon emissions [37,41,51,60].For example, Hoerning et al. [61] observed, with an experimental set-up using a fixed grate combustor, that fuel-rich conditions promote high emissions of benzene, naphthalene, acenaphthylene, and anthracene.
The correlation matrix in Table 6 shows that high particle diameters were emitted in conjunction with high temperature.At that moment, particles of several types were likely emitted, such as unburned wood, unburned products, soot agglomerates, and inorganic products [42,59].The hydrocarbon-particulate correlation matrix also shows that particulate emissions are associated with a high concentration of hydrocarbons.In addition to the increase in PAH formation under fuel-rich conditions, the participation of THC in particle formation might be by insemination on the elementary nucleus (ultrafine up to 120 nm) and condensation (since there are high THC concentrations) [61].These results can be linked to the higher value of TSP for the D6 layout.Fitzpatrick et al. [42] measured species with higher molecular weights (adsorbed on soot) and pieces of unburned fuel while they burned pine wood in a more intense combustion.In the same way, Nystr€ om et al. [59] saw an increase in inorganic particle emissions when they burned, in a stove, high batches of wood logs (pine and beech), in comparison to low loads.After the strong volatilization of the wood in the first instant, there is a rapid decrease in the firebox temperature during the last 2/3 of the combustion time where a great deal of hydrocarbons are emitted due to poor combustion conditions.This leads to THC/PM coefficients with higher values for ultrafine particles (þ0.58 < r < þ0.62) (Table 5).The combustion behavior is, then, similar to category I.

Conclusion
The interest of this study lies in the recovery of waste wood, sometimes called 'opportunity wood', at an individual scale in domestic wood stoves.Several wood layouts were tested in a domestic wood stove to study the influence of the surface on pollutant emissions.Three combustion behaviors could be observed.
First: Efficient combustion conditions (D4 and D5) thanks to an optimal surface area, characterized by a stable release of power throughout combustion.CO and HC EFs were obtained corresponding to the emissions of the reference fuel of the stove.It should be noted that these two wood layouts generate fewer ultrafine particles, showing a reduction of 72% compared to the highest emissions.
Second: Poor combustion conditions (D1, D2 and D3) due to insufficient exposed surface area (in relation to the stove's capacity), resulting in a reduced combustion rate and temperature, and high emissions of unburned products (CO and HC) and particulates.Among these particles, emissions of ultrafine particles are the highest (especially for 20 nm).
Third: Poor combustion conditions (D6) caused by a toolarge surface area exposed to the firebox (for the capacity of the stove), with high average temperature and significantly higher EFs for all pollutants.The immediate power of 22-30 kW released by the reaction during the first third of the combustion time greatly surpassed the nominal power of the stove (8-14 kW).
This work provides a better understanding of pallet burning at the individual stove scale, with the main findings being that coniferous and waste wood can be burned in a stove designed for hardwood only, and managing the surface area is an efficient way for the user to optimize combustion.However, the user must be careful with the layout, as overlapping plank configurations can block heat conduction, and too much wood can impair the stove's ability to supply oxygen.
At the academic level, the contributions are as follows: The surface area of wood offered to the fire strongly impacts the quality of combustion and thus the production of gases and particles.Ultrafine particles, i.e. smaller than 70 nm (determined by ELPI impaction plateaus), increase significantly in concentration in cases of more incomplete combustion.On the other hand, ultrafine particles are greatly reduced during more complete combustion, i.e. with a higher temperature (from 550 C in the case of this device) and a better mix between pyrolysis gases and oxygen.Heterogeneous combustion phases always show the same diameters of emitted particles.The correlation of gaseous hydrocarbons to particles is relative to the quality of combustion.The same occurs with the temperature, which can be correlated or not depending on the quality and the phase of combustion.
However, there is a need for further study of softwood combustion in other appliances designed for hardwoods.It would be interesting to examine how parameters such as the use of mechanical ventilation or the use of weathered pallet waste (the influence of leaching of wood subjected to rain has been demonstrated [62]) may influence home heating using pallet wood.A more detailed analysis of the nature of the emitted particles through photography or chemical analysis would also be interesting.Although the results of this work were obtained under laboratory conditions supported by analyzers, simple observations by the user can avoid poor combustion conditions.This opens the way for the clean and rational use of a new type of fuel in stoves not designed for such use.

Figure 2 .
Figure 2. Mean number particle emissions of the six combustion tests performed, and results from a previous study.PM : particle matter.

Figure 3 .
Figure 3. Size distribution of the particles emitted by the combustion of D4 (left) and D3 (right).D average : average particle diameter.

Figure 4 .
Figure 4. Emission of the THC and power release for D6 layout, nominal batch no. 3 (A) and for D4 layout, nominal batch no. 1 (B).

Table 1 .
Characteristics of the six different disposition tested in the domestic wood stove.

Table 3 .
Influence of wood arrangement on particulate diameter most emitted (DME).

Table 4 .
Correlation matrix of T bf , THC, CO, and O 2 for six combustion tests.

Table 5 .
Correlation matrix between total hydrocarbons and particle emissions for each impaction stage of the ELPI impactor.

Table 6 .
Correlation matrix between back firebox temperature and particle emission for each impaction stage of ELPI.