The curious case of the second/end peak in the heat release rate of wood: A cone calorimeter investigation

The reasons behind the occurrence of a second/end peak heat release rate (PHRR) during wood combustion under radiative heating were determined. Effects of the type of rear material, wood thickness, char progression, and its microstructure, as well as moisture content/transport in spruce wood, were studied. Rear materials used were insulating Kaowool, conducting steel, and the same wood but physically separated from test specimen by aluminium foil. The intensity of the second/end PHRR with Kaowool was almost 50% more than that of the sample with steel. Thus, the second/end peak is governed by the boundary condition defined by the rear material, which determines the heat losses at the rear side of the specimen and consequently the temperature of the specimen. Higher specimen temperature enhances the pyrolysis rate, thereby causing the second/end PHRR. The appearance times and values of the second/end PHRR for 30, 20, and 10 mm wood were 1740 s/78 kWm−2, 685 s/134 kWm−2, and 450 s/160 kWm−2, respectively. Char progressed to the rear of the samples even with a thin (8 mm) conductive steel substrate. Cracks in char grew almost three times wider during the second/end PHRR compared to the sample with no second/end peak. Char cracking had no significance on the time of occurrence of the second/end PHRR but affected the overall heat release. High moisture content reduced the charring rate and delayed the time of occurrence of the second/end PHRR as more water was needed to undergo a phase change, requiring a higher amount of energy.


| INTRODUCTION
Timber/wood is one of the world's oldest construction and building materials. Timber/wood-based construction always had a strong economic and social value, and today it is used globally. 1 As climate change becomes increasingly critical, it has been pivotal to consider and produce sustainable building materials and structures. Therefore, the environmental significance of wood as a natural resource with renewable and sustainable properties has become more important than ever. 1 As timber/wood receives more attention as a construction material, it is important to understand its combustion properties as well as enhance the understanding of its behaviour when exposed to fire.
Fire safety is of great importance in the building industry leading to high demands for design requirements and limitations in construction codes. To fully understand the fire and burning behaviour of wood, extensive modelling and tests have been conducted making its fire-performance characteristics (thermal degradation, ignition, smoke and heat release, charring rate, and flame spread) well documented. 2,3 However, there is a dearth of literature on the phenomena that affect the various stages of heat release rate (HRR), especially the second or end peak, as most commonly observed in a cone calorimeter. As HRR is one of the most important variables in the evaluation of material fire hazards, [4][5][6] it is of great importance to study it and gain a deeper understanding of its complexity.
The cone calorimeter is one of the most widely used benchscale instruments to elucidate the reaction-to-fire properties of materials. The apparatus exposes a sample to a specific heat flux, and subsequently, it can measure and analyse the combustion gases and smoke. HRR is proportional to oxygen consumed during combustion by a material and is described as the rate of fuel volatiles generated per unit area of fuel surface multiplied by the effective heat of combustion. 7 The cone calorimeter calculates HRR and measures mass loss, time to ignition, and smoke production of a specific fuel. The produced data can be used in fire modelling, prediction of fire behaviour, pass and fail tests of new products and materials, and fire performance ranking. 8,9 It indicates the amount of energy released per unit time, which influences the combustion properties such as hot gas temperatures, plume flows and rate of descent of hot gas layers. 10 Fire development is often characterised in terms of HRR versus time 11 making HRR one of the most important variables when it comes to evaluating material fire hazards. 4,5 Over time, the HRR curves of materials with thick charring, such as wood, tend to peak at the start where no charring has yet occurred as well as at the end of the combustion cycle, see Figure 1A. By understanding in detail why and when the second/ end peak occurs in HRR, it is possible to improve the prediction of the burning behaviour of wooden structures.
Wood is a thick charring material. The char forms as a result of the contributions mainly by lignin and cellulose 12 , and according to Lowden et al. 13 the first peak in HRR is a response to ignition of the surface. The generated heat from combustion will maintain the pyrolysis of the material and release more volatile gases. 14 The decrease in HRR is caused by the formation of a char layer that has an insulating effect slowing down the pyrolysis process as well as hindering the transport of heat and volatile gases through the material. 15,16 The second/end peak is believed to be a result of material burn through and char cracking, which enable the escape of additional volatile gases. 17 F I G U R E 1 (A) A typical heat release rate curve of wood (adopted from Schartel & Hull 8 ); (B) Wood samples exposed to different heat fluxes ranging from 20 to 65 kWm À2 (Forest Products Laboratory 2 ); (C) Heat release rate of thermally thick noncharring samples; and (D) Heat release rate of thermally thick charring samples (adopted from Schartel & Hull 8 ) The final decrease in HRR occurs when no more volatiles are released and the flaming combustion ends, resulting in a steady baseline. Elsewhere, Tran et al. 18 suggests that the second/end peak is a result of a raise in the bulk temperature of the remaining material when the pyrolysis zone reaches the rear side of the sample. The author argues that if the material is sufficiently thick, the second peak will not occur.
The author's experimental observation shows that a wood sample that was 18 mm (analysed with a cone calorimeter) had a second peak while a sample with 64 mm thickness (analysed with the Ohio State University heat release rate apparatus), that was only analysed until the char depth reached 36 mm, did not experience a second peak.
Studies by forest products laboratory 2 suggest that the second/end peak is an effect of the pyrolysis zone reaching the sample's rear side, which increases both temperature and HRR. Forest products laboratory 2 studied HRR when a sample is exposed to different heat flux intensities. Results shown in Figure 1B indicate that a higher heat flux gives quicker and higher HRR peaks, meaning that burn through and flameout take place earlier when exposed to high temperatures and heat flux. The second/end HRR peak is, according to Gan et al. 19 caused by cracks forming in the char layer, creating an easy pathway for the combustible gases inside the wood, and as the cracks grow larger, the peak heat release rate (PHRR) is reached. The authors stated that a densified wood will have a lower and delayed PHRR compared to normal wood because the char structure possessed thinner and shallower cracks. Studies by Das et al. 20,21 corroborate the findings of Gan et al. 19 where the authors stated that the char structure is important for the extent of HRR and that a dense and rigid char will hinder the mass and energy transport more than a cracked char structure. However, Bartlett et al. 22 also studied the effect of wood density on charring rate. Herein, it was discovered that higher density decreases charring rate. The authors argue that higher density generates char at a slower pace because more energy is required to pyrolyse a greater material mass. In addition, an increase in moisture content reduces the charring rate of wood, as shown in the work of Mikkola. 23 The author found out that charring rate is proportional to 1/(1 + 2.5 Â moisture content). Schartel and Hull 8 presented HRR curves of samples of thermally thick charring and noncharring materials ( Figure 1B,C). The thermally thick noncharring sample showed a second/end peak in HRR similar to the HRR curve of wood ( Figure 1A). The authors believed the second/end peak to be an effect of the pyrolysis zone reaching the rear side of the sample, where the insulating material beneath the sample prevents downward heat transfer and instead causes heat build-up, which increases the HRR.
On the other hand, the HRR curve of the thermally thick charring material did not experience a second/end peak and as the layer of char thickened, the HRR decreased. The HRR of woody materials is also a factor of the moisture content. The mass flow of water through the material cools the pyrolysis zone and slows down the charring rate. 22 A high moisture content prolongs the time to ignition and slows the rate of pyrolysis. Therefore, a previous study has found that wood with, for example, 12% moisture content ignited after almost double the time as that of dry wood. 22 In a numerical study, Hostikka and Matala, 24 reproduced the second/end PHRR based on boundary condition effects only, and no other effects, such as char cracking, for example, were implemented in the model.
From the aforementioned studies, it can be inferred that the layer of char may or may not have a significant influence on the second/ end peak of the HRR curve. The thermally thick noncharring material shows that char cracking is not needed for the sample to experience an end peak, and the peak will only appear when the pyrolysis zone reaches the rear side of the sample. However, the thermally thick charring material indicates that a layer of char prevents an end peak from occurring. This is contrary to the notion that sample burn through is the reason for a second HRR peak. From the literature review, it is clear that researchers are equivocal about the reason for the end or second PHRR of wooden materials. This warrants a systematic study that will shed some light on the dominant mechanism responsible for the occurrence of the second/end PHRR peak, as commonly observed in cone calorimeter experiments.
The factors that have a significant influence on the nature of HRR of woody materials, especially towards the end, are the type of rear material/substrate, thickness of the wood, char progression during combustion, char microstructure and moisture transport. 25 The present investigation is aimed at systematically elucidating the occurrence of the second/end peak in the HRR of wood during cone calorimeter analysis. Three types of rear materials were chosen: Kaowool, which is an insulator; steel plates, which are conductors or heat sinks; and aluminium foil wrapped around the same type of wood, where no further combustion will occur and which will possess the same thermal properties as the tested wood. The tested wood had three thicknesses: 10, 20, and 30 mm. Additionally, the effects of char progression, char microstructure, moisture content, and its transport on HRR were also determined. This study is novel in the sense that evidence-backed phenomena that govern the occurrence and extent of the second/end peak of wood HRR is presented. The information existing in this investigation can be beneficially used by academia and industry, alike, to comprehend and presage the burning behaviour of wood, at a fundamental level, and bestow essential fire safety in buildings.

| Materials
The material used for the research was untreated spruce wood (softwood) with a moisture content of 11.2%-20.8% procured from Beijer Byggmaterial, Luleå, Sweden. Three rear materials were used during the experiments: steel, Kaowool, and aluminium foil wrapped around spruce wood. A humidity sensor, Tinytag Plus 2 (TGP-4505), having an external temperature and humidity probe, was procured from Gemini data loggers, Intab, in the United Kingdom to measure the relative humidity. The sensor measures temperatures between À25 and +85 C and relative humidity from 0% to 100%. In addition, a type K thermocouple with a diameter of 0.25 mm from Intab was used to measure the temperature at the rear side of the sample.

| Cone calorimeter
The cone calorimeter experiments were performed according to the standards stipulated in ISO 5660 using equipment from Fire Testing Technology (FTT) at Luleå University of Technology. The coil temperature was set to 722 C, which equals a heat flux of 35 kWm À2 and the exhaust duct flow rate was 24 ls À1 . A smaller intensity of heat flux, that is, 35 kWm À2 was more suitable for this study as the results were generated more moderately and therefore it was easier to analyse the burning process. The computer connected to the cone calorimeter was adjusted to record HRR every 5 s. This was calculated from the data collected through the consumption of oxygen. The computer programme Easyview was used to visualise the temperature profile during the test. The programme was set to collect new values each second, providing a temperature curve of the fire scenario.

| Scanning electron microscopy (SEM)
SEM images of pristine spruce wood and char samples were taken before and after the second/end HRR peak was obtained. Burnt samples from the cone calorimeter were immersed in liquid nitrogen and cut into two by a hand saw to observe the char depth. The char samples for the microscopy were taken from sections of the top surface.
A JEOL JCM-6000 tabletop SEM was used. A low-vacuum mode was applied with an acceleration voltage of 15 kV and a working distance of 6 mm. The images were generated from the detected signals of backscattered electrons.

| Experimental design
The experiments were designed to determine the effect of the type of rear materials, thickness of tested wood, char progression and its microstructure, as well as moisture content and transport on the HRR. Table 1 gives a holistic view of the entire experimental plan. The samples used for SEM analysis were prepared by removing three samples from the cone calorimeter before and after the second PHRR.
To immediately stop the burning process and avoid damaging the char surface, liquid nitrogen was poured on top of the samples. Furthermore, in order to analyse the effect of moisture on HRR, samples of 10, 20, and 30 mm with moisture ratios ranging from 11.2% to 20.8% were dried at 100 C for 1 week to attain a moisture content of 5%-8%. These samples were tested with Kaowool as the rear material. To determine the effect of moisture transport during combustion, Kaowool (8 mm) was wrapped around the base of the cable attached to the humidity sensor, leaving the sensor exposed to avoid damaging the cable by radiation from the cone. The wood sample was then placed in the lid of the sample holder, and the probe was inserted at the backside into the centre of the sample through a hole specially created on the side of the lid of the sample holder, see Figure 2D. The  wood sample, protective insulation, and humidity sensor were assembled as shown in Figure 2E. The images and other details of the cone calorimeter used, thermocouple placement, humidity sensor, and its placement are given in the Data S1. Temperature was recorded for all samples with replication, and Figure 3B presents one curve from each set of tests that is closest to the mean value. Figure 3B indicates that the 10 mm sample had a faster increase in temperature compared to the 20 and 30 mm samples. Flameout occurred at the steep decrease in temperature seen in Figure 3B, and at this time the samples were also removed from the cone calorimeter. Figure 3C shows the different stages of the HRR curve for a 10-mm sample. Figure 3C(a) presents the initial PHRR, where the entire surface is aflame. This occurred right after ignition of the sample. In Figure 3C(b), char has begun to form, creating a protective layer over the virgin wood underneath. Thus, the burning process slowed down and HRR decreased. The second PHRR is seen in Figure 3C(c), where the flames became more intense and appeared in and around the cracks of the char. Figure 3C(d) shows the sample after the second PHRR; the burning was slow, but the sample began to glow and expand, and ash was also produced on the surface. Similar to the temperature curves of Kaowool, Figure 4B shows the temperature curves of one sample closest to the mean value for each set of tests. Figure 4C

| Comparisons of the effect of the rear materials
In this section, the HRR and temperatures of 10 and 30 mm thick samples, having various rear materials, are compared to give a deeper insight into the effect of these substrates. Figure 6A

| Char progression
To analyse the char progression during the second/end PHRR, three samples of 10 and 20 mm were tested with Kaowool in the cone calorimeter. The samples were removed prior to, during, and after the second/end peak. 30 mm samples were not analysed as the second/end PHRR on these samples were less defined and, hence, difficult to "catch." Figure 7A shows the progression of the char zone during the second/end peak for 10-mm wood samples. The char zone before the peak at around 280 s can be seen in Figure 7A(a). The char layer had reached the middle of the sample, and beneath the layer of char was fresh virgin wood. Figure 7A(b) shows the progression of the char zone at the second/end PHRR, roughly around 420 s. It was observed that the wood had almost burned through. Figure 7A(c) shows the wood sample after the second/end peak at around 520 s. No fresh wood remained, and the sample was completely burnt with a brittle structure and ash showing on the sample surface. Over time, as the creation of char proceeded further into the wood sample, cracks appeared early and grew wider and deeper as the char layer became thicker. Moreover, the sample deformed gradually as the char layer enlarged, this is seen in Figure 7A(c), where the sample has been bent upwards.
The progression of the char zone in 20 mm samples is shown in Figure 7B, which has a similar trend to that of the 10-mm samples.
Prior to the second/end PHRR, around 580 s, the char layer approached the middle of the sample, see Figure 7B(a). The char progression at the peak around 760 s can be seen in Figure 7B Similarly to the aforementioned, 10 mm samples were tested with an 8 mm steel plate and removed at multiple timesteps throughout the second/end PHRR. The second/end peak occurred around 430-460 s; hence, the tests were stopped at 440, 460, 490, and 780 s, as seen in Figure 7C(a-d). There was an insignificant difference between Figure 7C(a,c), which represents a timespan over the second/end peak. However, it was observed that the charring had started to reach the rear side of the sample during this period of time. At 780 s, Figure 7C(d), the charring had grown larger but did not yet cover the entire sample backside. Figure 8 shows the SEM micrographs of pristine spruce wood and char after the cone calorimeter tests. Pristine spruce wood shows a porous structure with intact lumina and pits ( Figure 8A). Spruce wood after the second/end HRR peak, presented in Figure 8B, shows that most of the wood microstructures are still intact, which is most likely due to the liquid nitrogen halting the combustion process.  26 However, in Figure 8B, two major features are visible; white particle-like agglomerations and fibre-like layers on the edges of the cell lumens. According to Udoeyo et al. 27 these agglomerations and fibres are ash generated from burning the wood sample. Spruce wood before and after the second/end HRR peak are presented in Figure 8C,D. For the sample, after the second/end HRR peak, the crack is at least three times larger than the sample before the second/end HRR peak. The widening of the crack is attributed to the progression of heating, which is known to cause expansion. The formation of wider cracks increases the surface area of the entire sample, which facilitates increased burning. Moreover, ash formation is observed on the sample surface after the second/end peak, seen as white interconnecting fibres.

| The effect of moisture content
In order to analyse the effect of moisture on the HRR, samples of 10, 20, and 30 mm with moisture ratios ranging from 11.2% to 20.8% were dried at 100 C for 1 week, to attain a moisture content of 5%-8%. The samples were subjected to radiation from the cone calorimeter until flameout and HRR was measured for each, see

| The effect of moisture transport
According to Bartlett et al.,22 upon exposure of the wood samples to the irradiation from the cone calorimeter, some water vapour moves away from the heat source and recondenses at the rear side. Hence, moisture transport towards the rear, as a result of incident heat from above, for 10, 20, and 30 mm wood samples with Kaowool were tested in the form of relative humidity (RH), which is an indirect indication of the moisture transport. To keep a constant air volume, the size of the samples and insulation materials was kept constant, and the humidity sensor was placed at the same location, that is, the centre of the rear side of the wood. Three samples were analysed for each thickness and the results are shown as mean value and standard deviation in Figure 11A. All the samples were tested until the humidity sensor reached its operational limit of 60 C, and as a result, the relative humidity could not be measured during the entire burning process. It was observed that the highest humidity of 95 % RH was

| DISCUSSIONS
From Section 3.1.1, it was observed that char formed on the surfaces of the samples immediately after the first peak, which acted as an insulating barrier and therefore decreased HRR. All the samples had similar initial peaks and decrease in HRR, and the difference in HRR began when the HRR curves started reaching towards the second/ end PHRR. Figure 3A showed that a 10 mm sample experienced the quickest and highest second/end PHRR of 160 kWm À2 whereas the slowest and lowest corresponding values were observed in the 30 mm samples of 78 kWm À2 . It is clear that an increased thickness results in a lower, broader and delayed second/end PHRR indicating that an infinitely thick wood specimen would most likely not experience a second/end peak. This is because, for an infinitely thick sample, the heat wave will never reach the rear material (discussed subsequently). From Figure 3B, it is seen that the temperature kept rising after the second PHRR occurred. This indicated that the pyrolysis and char zone continued downwards through the specimen, which increased the temperature around the thermocouple the closer it got.
Moreover, it was observed that a thicker wood specimen delayed the increase in temperature. The 10 mm sample experienced its maximum temperature around 650 s, while 20 and 30 mm samples reached their maximum temperatures at 880 and 2340 s, respectively ( Figure 3B).
From the results reported in Section 3.1.2 and Figure 4C Figure 4C) and this is because of the conductive properties of steel. A thicker piece of steel will conduct the heat from the sample with better efficiency than a thin piece, and this reduces the heat available for combustion, resulting in no second PHRR or charring on the sample backside.
When comparing the effect of different rear materials, whose results are detailed in Section 3.1.4 and Figure 6, it was observed that the highest second/end PHRR was achieved by Kaowool, followed by aluminium foil-wrapped wood and 8 mm steel. This outcome was an effect from the thermal properties of the rear materials. Kaowool is an insulating material meaning that when heat approaches, it will prevent heat absorption, steel, however, is known for its conductive properties and heat will therefore travel through it (i.e., heat sink). This resulted in a higher recorded second/end PHRR from samples with Kaowool and a lower second/end PHRR from samples with steel.
Analysing the results of the 30 mm samples, Figure 6C, it was seen that the samples with aluminium foil had the highest second/end HRR peak of 88 kWm À2 and not Kaowool, which had a second/end HRR of 78 kWm À2 . Based on the results of the 10 mm samples (see Figure 6A), Kaowool was expected to have the higher second/end PHRR, however, the difference of 10 kWm À2 is considered insignificant. The increase in second/end PHRR might be a response from combustion and escape of volatile gases from the wood wrapped inside the aluminium foil.
From Figure 6B, it can be seen that Kaowool had the highest and   gases, the wood wrapped inside the aluminium foil started to combust that increased both the temperature and the HRR. Figure 7C shows the char progression and rear sides of 10 mm wood samples with 8 mm steel at various timesteps around the second/end PHRR. This sample combination was of particular interest since it was the only one of the different steel combinations that showed a second/end PHRR (see Figure 6A). Similar to the char progression of the 10 and 20 mm Kaowool samples (see Figure 7A,B), it was seen that charring reached the backside of the sample, especially on the corners. The occurrence of a second/end PHRR in this sample ( Figure 7C) was due to the rear material being thinner than, for example, a 20 mm steel plate, which understandably consisted of more conductive material; thus, a thicker plate is a more efficient heat sink.
A thin steel plate would encounter air underneath, which is an insulator and may facilitate heat accumulation in the sample, relatively faster than a thick steel plate. Another parameter to consider may also be that the lid of the sample holder was not tight enough around the sample allowing easy air access through the corners of the lid. A 20 mm steel plate took up more space in the sample holder, which tightens the wood to the lid. These findings support that the second/ end PHRR is strongly affected by charring of the rear side.
From Figure 8, it was observed that after the second/end PHRR, the cracks on the char were at least three times wider, which is attributed to heating, that is known to cause expansion. Wider cracks contribute to the increase of the surface area, which in turn facilitates intensified burning.
This implies that the formation of larger cracks and breakdown of char could contribute to increased HRR. However, it was observed that the time of the second/end PHRR is delayed with thicker samples (see Figure 3A).
This refutes char cracking as the major reason for the time of occurrence of the second/end PHRR. Cracks in char form regardless of sample thickness and the peak should therefore have occurred earlier. Moreover, Figure 3D proves that cracks become wider and deeper when a sample is thicker, which means that HRR would increase as the surface area increases. This is, however, not observed in the cone calorimeter tests, that is, a thinner sample experienced a more intense HRR than a thicker sample.
This could be explained by the same assumption that a thicker sample burns for a longer period of time; hence, most of the fuel inside the sample is consumed by the end of the test, thereby reducing the second/end PHRR. In conclusion, char cracking contributes less to the time of occurrence of the second/end PHRR and more on the intensity of the overall HRR. Without char cracking, there would be a less exchange of combustible gases, which in turn would markedly reduce the HRR.
Comparing low moisture content samples and normal samples, see Figure 10, it is clear that drier samples experience more intense HRR peaks, and that less time is needed to reach flameout. Upon heating, water starts to evaporate inside the wood and most water vapour will leave through the sample surface, while some will migrate deeper into the material where it recondenses. Significant energy is required to undergo a phase change, leaving less energy accessible for pyrolysis. Therefore, fire spread rate is quicker in dry wood since all energy is used for pyrolysis. This is supported by Figure 11A where it is shown that RH rose at the back of the sample during analysis. The 10 mm sample had a fast RH rise at the rear side as more water was pushed in that direction. The 20 and 30 mm samples had lower RH at the rear by the end of analysis (see Figure 11A). This was due to the wet zone having a larger volume when removed from the cone calorimeter, as shown in Figure 11B. Hence, the wet zone can contain more moisture. However, it was observed that RH of the thicker samples was decreasing prior to removal from the cone calorimeter; therefore, it cannot be determined if all the thicknesses would have reached the same level of RH by the time of flameout.
Based on the findings from the current study, a table (Table 3) has been formulated that describes each event and the cause during the combustion of wood, in a cone calorimeter, and the subsequent heat release. Although the study used spruce wood, it is envisaged that the overall trend of the findings will remain universal.

| CONCLUSIONS
The purpose of this study was to provide an understanding of the HRR curve of wood when exposed to heating in the cone calorimeter.
Fire development is often characterised in terms of HRR as a function Between 1st and 2nd PHRR Between the two PHRRs, the HRR decreases rapidly. This outcome is due to char formation on the surface of the sample. Char acts as a protective barrier that prevents mass transport of volatile gases and oxygen.
2nd PHRR Near the end of analysis, the HRR increases to a second/end peak. The second/end PHRR becomes increasingly delayed with thicker samples and varies in intensity depending on the material beneath the wooden sample. The second/end PHRR is a response to sample burn through, meaning that the heat gradient reaches the rear side of the sample. Kaowool, which is an insulator, hinders heat escape through conduction and facilitates heat accumulation, which raises the temperature and pyrolysis rate of the entire sample, consequently increasing the HRR and causing the second/end PHRR. Steel, however, is a heat sink, which will conduct heat away from the sample, resulting in a lower PHRR.
After 2nd PHRR After the second/end PHRR, there is a distinct decrease in HRR as most of the fuel is now exhausted. The sample smoulders until the remaining flames extinguish.
of time. However, before flameout happens, there is usually another peak in the HRR, and the reasons for this were revealed in the current study in a systematic manner. The investigation, thus, determined the effect of various factors that can cause this second/end peak. One of these factors was the type of rear material over which the wood is placed during testing, which included insulating Kaowool, conducting steel plates, and aluminium foil wrapped around wood that has similar thermal properties to that of the tested wood without facilitating further combustion. The other factors were thickness of the tested wood (10, 20, and 30 mm), char progression, and its microstructure, moisture content in wood and its transport during combustion.
From the results obtained in this study, it can be stated that the second/end PHRR is mainly due to sample burn through, that is, when the heat wave reaches the rear side of the wooden specimen the heat losses will be affected. The significance of this effect, and thus, also the intensity of the PHRR, is heavily dependent on the type of rear material that is used. Kaowool, which is an insulator, conducts less heat than, for example, steel, which is a conductor. Additionally, thicker samples delay the occurrence and intensity of the second/end PHRR. The formation of char does not have a significant impact on the second/end PHRR but more so on the overall HRR during combustion. Char has insulating properties that obstruct the flow of oxygen and combustible gases, resulting in supressed burning and low HRR. When the char layer cracks, the protective ability deteriorates and the HRR increases. Therefore, robust char with minimal cracking will delay the time it takes to reach burn through and flameout, which in turn preserves the load-bearing capacity of wood for a longer period of time. It was, however, difficult to determine to what degree charring and char cracking affects the second/end PHRR. The time of occurrence of the second/end PHRR is also delayed by high moisture content, as a high amount of energy is required for water to undergo a phase change. Therefore, wood containing a normal amount of water will reach the second/end PHRR at a later stage than dry wood, as more water needs to evaporate.
This study was based on examining the occurrence of a second/end peak on the HRR curve of wood, which is experiencing radiative heating. The comprehension of the fundamental burning behaviour, especially, the mechanisms that contribute towards another peak in heat release towards the end of the combustion cycle is beneficial for fire scientists and rescue services.

ACKNOWLEDGMENTS
The authors express their gratitude towards Forskningsingenjör Erik Andersson from Structural Engineering research subject of Luleå University of Technology for his assistance with cutting the steel plates and arranging test set-up to measure relative humidity. The authors also immensely appreciate the constructive comments of the three anonymous peer reviewers, who helped elevate the quality of this article.

CONFLICTS OF INTEREST
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.