Scale effects and mechanisms ruling the onset of wastewater sludges self-heating

Depósito de Investigación de la Universidad de Sevilla https://idus.us.es/ This is an Accepted Manuscript of an article published by Elsevier in Fuel, Vol. 256, on November 2019, , available at: https://doi.org/10.1016/j.fuel.2019.115876 © 2019 Elsevier. En idUS Licencia Creative Commons CC BY-NC-ND 1 Scale effects and mechanisms ruling the onset of biomass self-heating Della Zassa M.a,c, Ronda A.b, Gianfelice G.a,c, Canu P.a,c* a Department of Industrial Engineering, University of Padua, 35131, Padova (Italy) b Department of Environmental and Chemical Engineering, University of Seville 41092, Seville (Spain) c K-INN Tech, Kinetics & Innovation, 35121 Padova (Italy) *corresponding author: [email protected] Abstract We investigated the autoignition and self-heating of three different dried, wastewater treatment sludges of different origin and share of municipal and industrial waste water. We used a custom testing apparatus to study large (approx. 20 g) samples, where heatand mass transfer limitations are purposely relevant, as in real scale applications. We tested different particle size distributions, applying heating rates between 0.1 and 6.0 °C/min. We proved that i) ignition can occur in more than one step, ii) the same solid behaves very differently by changing its particle size distribution and the applied heating rate, iii) the onset is controlled by physical processes, rather than chemical, iv) the hazard is determined (and can be controlled) by modifying the particle size, the bed structure and density. We conclude that the hazard of self-ignition and heating is largely controlled (thus manageable) by the physical properties of the solids and its packing, given that its reactivity in air is always sufficiently high, once appropriate (and common) environmental conditions are met. Consistently, the selfheating hazard assessment requires investigation at a representative scale, where physical processes have an appreciable impact. Conclusions obtained on sludges are expected to apply to any biomass, with comparable properties. Keywords: Self-heating, Solids ignition, TGA, Oxidation, Smouldering, Sewage Sludge 2 1. Introduction, The expanding interest in biomass exploitation as a C-neutral source of energy calls for deeper understanding of the threads in biomass treatment and manipulation. Here we address the spontaneous self-heating up to ignition. We focus on a largely available, specific type of biomass, which is the sludge resulting from biological wastewater treatment processes. It is a complex mixture of biomass, water, chemicals that are refractory to biological degradation, and inorganic materials. Sewage sludge is being considered as a raw renewable energy source with zero-CO2-emission [1]. Previously studies proved that the sewage sludge can be used efficiently as fuel [2]. However, storage, handling and transportation of these waste requires assessing the hazards of its self-ignition propensity. The properties of these material, together with the presence of fines (due to wearing and tearing) and oxidizing environment (the atmosphere) can trigger combustion, resulting in smouldering or fires [3]. Understanding how sewage sludge chemical and physical properties affect the thermochemical processes and the risk of self-ignition is mandatory ([3], [5]). The properties of sewage sludge are extremely variable, because each sludge has specific origin and production conditions, including daily variations. In this study, three sewage sludges with different origin have been used. Syed-Hassan et al. ([2]) indicated that the characteristics of sewage sludge relies on six factors: wastewater pollutant loading; the wastewater treatment; the local environmental legislation; the water reclamation requirements; the processing steps, and the seasonal variations. Díaz et al., [3], define criteria for the risk assessment of sewage sludge self-ignition, developing characteristic risk plots; they correlated variables and parameters such as activation energy (Ea) with oxidation temperature (Tc), or the sulphur content with oxygen/hydrogen ratio, and define risk zones (low, medium, high and very high risk) to classify the sludges. In previous contributions ([6], [7]), we reported about the self-heating of a single, specific dried wastewater sludge, also at industrial scale. Here we elaborate on the hazard evaluation procedure and criteria, extending the comparison to other sewage sludges. We already showed [5] the gap between analytical techniques and measurements on larger samples. The standard TGA-based studies aim at identifying the intrinsic, chemical kinetics of biomass degradation ([8], [9], [10]), attempting to minimize the interference of transport limitation, of heat and mass. It has indeed frequently reported that kinetics may be falsified by the operation conditions [11], so that a large gap between standard, analytic TGA and larger samples’ behavior is observed. 3 That has been elaborated in purely experimental works ([3], [12]) and in experiments supplemented by modeling ([13], [14], [15]). In this work, we question the relevance of a fine characterization of the chemical reactivity of the solids in air, attempting to minimize heat and mass transfer limitations using small samples, as in standard TGA. Rather, we suggest to properly investigate the ignition on larger samples (sometimes called ‘thick biomass’), carefully investigating the heat transfer rate. 2. Materials,and,Methods, 2.1 Sewage&Sludge& Three different sewage sludges are used in this study: • sewage sludge A (labelled SS-A) is produced in an Italian municipal wastewater plant (10% industrial-90% municipal) • sewage sludge B (SS-B) is produced in an Italian industrial wastewater plant (90% industrial-10% municipal) • sewage sludge C (SS-C) is produced in a Spanish municipal wastewater plant (100% municipal) All sludges were granulated particles, from drying process at the wastewater treatment plants. Rotary driers are used, with indirect heating, based on thin film evaporation on externally heated surfaces. However, the plant suppliers, the precise settings and the detailed operations might differ, causing variation on the properties of the products, in addition to their origin. The main properties of the sludges are in Table 1; other properties and the appearance are in the Supplementary Material and in [5]. The proximate analysis is carried out using the ASTME871-82, E872-82 and D1102-84 standards, to determine the content of moisture, volatile matter and ashes respectively ([1618]). The fixed carbon (FC) is obtained by difference. The mean particle size is determined from the particle size distribution, reported and discussed in Section 3.4 for each material. The intrinsic density is determined by a 50 ml Regnault Pycnometer. The IT (Induction Temperature) and MLT (Maximum Weight Loss Temperature) are measured as in [3], in our thermogravimetric testing apparatus, with a heating rate of 2°C/min, from ambient 4 temperature up to 250°C, based on previous experience [23], and indications by DSC of the onset of reactivity. The Reactivity Index (RI= 100[min]/t) is evaluated modifying the procedure reported in [19] for the Maciejasz index; we used a Dewar flask, with 10 g of sample, adding 15 mL of H2O2 (30% water solution). The RI is defined as the time necessary for the sample to increase by 15K from the initial temperature, since we never observed much higher temperature in a reasonably short time. From our experience [6,7], we expect this test to be extremely vulnerable to the concentration of H2O2 used, the heat transfer properties of the vessel and the amount of material. Table 1: Characterization of the dry sewage sludges Parameters SS-A SS-B SS-C Proximate analysis moisture, % 23.5 13.6 12.6 volatile matter, % 65.6 66.4 64.6 fixed carbon, % 12.6 7.9 7.5 ashes, % 21.7 25.8 27.9 Atomic ratio H/C 1.79 1.73 1.87 O/C 0.90 0.50 0.70 O/H 0.50 0.29 0.37 S content, % 1.6 8.2 2.0 Mean diameter, mm 0.986 0.462 1.699 Intrinsic density, kg/m3 2577 2874 2750 Bulk density, kg/m3 504 668 636 HHV, KJ/Kg 13100 22200 17280 IT, °C 225 204 210 RI 33.3 2.5 1.3 MLT, °C 247 229 246 risk classification [3] high very high high Table 1, underlines a number of differences among the three sludges, that can affect the onset and the propagation of biomass smouldering. The water content is much higher for the SSA; moisture content is expected to delay the reactivity. Even if SS-A has also the highest FC and the lowest ashes content, it results in the lowest HV. The ashes also vary significantly among these sludges; it should be minimized for SS use as fuel, to increase the energy content and minimize furnace cleaning, with sensible heat losses that lower the overall efficiency [20]." The mean particle size varies by a factor of 3. In gas-solid reactions, the reaction rate depends on the accessible solid surface per unit volume of the porous bed, which scales with 1/dp. Therefore, the lower the particle size, the higher the expected volumetric reaction rate ([21], [22], [23]). However, a smaller particle size (and specifically 5 fines in the distribution) decreases the permeability of the bed to the co-reactant (oxygen in the air), so that the final influence on the reactivity must be evaluated. The main differences in metals content of sludges (see Supplementary Material) are due to their origin (municipal or industrial). The special case of B sludge is due to the prevailing local industrial wastewater, from tanneries [7]. In the B sludge we find a low concentration of heavy metals except chromium, largely used in tanning, reaching up to 3% wt. of solid substance (on dry basis). Iron in sludge B (up to 2 wt. %) is due to additions at different wastewater treatment stages: as FeCl3 after biological treatment, to remove phosphorous from wastewater and improve clarification, and then as FeCl2 before dewatering. According to Díaz et al., 2019 [3], the relation between S content and O/H ratio measure the self-ignition risk, thus SS-A and SS-C are classified at ‘high risk’ of self-ignition, whereas the SS-B is at ‘very high risk’, due to the larger S content and the lower O/H ratio. These reactivity parameters and ranking will be compared with the following, real ignition tests. 2.2 Experimental&procedure& Figure 1 sketches the configuration of the experimental set-up used. Figure 1: Schematic representation of the macro TGA The smouldering reaction occurs in a gas-permeable cubic container, with a side length of 3.7 cm, made of a stainless-steel net (mesh size 500 µm). The crucible is hanging on a scale and it is placed inside a static oven, to evaluate the weight loss during biomass heating and 6 ignition. The heating policy was a unique temperature ramp from ambient temperature up to 250 ºC, with four different heating rates (HR), 0.1, 0.2, 2 and 6 °C/min. Note that HRs are very low, compared to standard TGA. The reason will become clearer later; it is required to isolate different steps in the ignition process. Furthermore, the oven is mainly used to compensate for the heat losses from the sample, particularly at the reaction onset, when the heat of reaction is still weak and thermal extinction extremely likely. The oven thus allows to reduce the misleading effect of the sample size. Small samples appear to ignite at a higher temperature simply because they lose heat more easily. Again, the procedure proposed here to assess the ignition hazard aims at reducing the uncertainty in the evaluation of the risk due to an inappropriate scale of testing. The oven atmosphere is not controlled. It is air with significant infiltration (thus O2 renewal), driven by the convective mass losses from the top hole. At the beginning of each test, a sample of approx. 20 g (see Table 2) is loaded into the crucible and placed inside the oven, before heating starts. The temperature of the sample, of the oven, and the weight of the sample are continuously recorded. When the temperature of the oven reaches 250°C, it is kept constant until the sample weight settles to the final value. Therefore, the runtime may vary considerably, up to days. This is consistent with the time scale of industrially relevant solids ignition incidents. Finally, the resulting sample is collected to be analyzed. A total of 16 test were performed to analyze in detail the smouldering of three SS and the mechanism of the process, as well as the effect of particle size and the reproducibility of results. In a sense, this apparatus resembles a standard TGA on large samples. Contrary to the aim of the analytical instruments, where the sample size is purposely minimized to limit the effect of heatand mass transfer resistances, we deliberately investigated larger sample. That is consistent with actual applications; in addition, it highlights that in real applications the intrinsic reactivity of the material is hindered by overwhelming physical processes, such as gas permeation in the bed and heat dissipation across the bed. 3. Results,, Table 2 collects all the experiments used in the following discussion, with the corresponding operating conditions and main results, in term of Weight Loss (WL), defined as: 7 𝑊𝐿 = $ !"!#$""#$%&% "! ∗100 (1) where W0 is the initial and Wfinal is the final sample weight. ID is our unique identifier of each single test. Table 2: Table of experiments Run ID Sample HR, ºC/min W0, g WL, % Notes #1 133 SS-A1 0.1 19.80 46.7 Lowest HR #2 122 SS-A2 0.2 19.70 78.9 #2Bis 127 SS-A3 0.2 20.80 76.6 Reproducibility #3 151 SS-A4 0.2 20.58 41.0 Pellets #4 140 SS-A5 2.0 18.50 81.5 Large HR #5 114 SS-B1 0.1 24.00 75.5 Lowest HR #6 120 SS-B2 0.2 24.30 75.7 #6Bis 112 SS-B3 0.2 24.00 73.8 Reproducibility #7 128 SS-B4 0.2 20.63 73.3 Fine fraction #8 158 SS-B5 0.2 20.59 20.6 Pellets #9 166 SS-B6 0.2 23.38 32.8 Stop after I peak #10 169 SS-B7 0.2 24.22 38.1 Stop at II peak #11 168 SS-B8 0.2 23.00 75.0 Stop after II peak #12 115 SS-B9 2.0 23.67 75.3 Large HR #13 142 SS-C1 0.2 21.56 34.8 #14 143 SS-C2 0.2 22.28 45.0 Milled #15 141 SS-C3 2.0 20.80 64.8 #16 139 SS-C4 6.0 25.95 54.2 Largest HR 3.1 The&ignition&pattern&& The slow heating of a large sample produces a typical pattern, summarized in Figure 2 for SS-B, which shows the most interesting behavior. The figure (and several in the following) has 4 panels; panel a) shows the evolution of temperature in the sample and in the oven; panel b), c), and d) show the same gravimetric information (% of the initial weight), as a function of sample temperature, time, and oven temperature, respectively. Although panels are different combinations of the measured weight, time, oven and sample temperature, they allow to highlight specific features, as discussed in the text. Specifically, Fig. 2 also quantifies the reproducibility of the test, following our procedure. The reproducibility of ignition on SS-A is in Figure SI-1. Since tests are destructive, replicas use different samples 8 of the same type of sludge. The reproducibility is very good in terms of stoichiometry; the moisture and final residues are consistent with proximate analysis, where much longer treatment time is used. The kinetics (thus timing) of the process is not so precisely reproducible. Although the onset of reactivity is well comparable, the progress of the reactive (then thermal) wave inside the large sample is not so uniform to yield perfectly overlapping temperature profiles. Still, we believe that the technique is well reproducible for real wastes that are often poorly homogeneous, to reliably develop an investigation on supporting conditions for self-heating. From Figure 2 we see that the sample temperature lags below the oven temperature during drying, because of the endothermic evaporation. The moisture lost (Fig. 2b-d) matches the amount determined by the proximate analysis, to confirm that a slow heating effectively releases all the moisture, even at low temperature. Following desiccation, we observe a clear, 2-steps ignition in SS-B, at two temperatures. Approx. 20% of the initial weight is lost in the first ignition, which is unable to self-support. A further increase in temperature produces a second ignition where an additional 40% of weight is lost, up to the inorganic residues identified by the proximate analysis (see Table 1). The loss of weight appears totally different when compared to the sample or to the oven temperature (panels on the right). In term of oven temperature, we might conclude that the onset of reactivity is sharply located at about 180°C for the first step and 230°C for the second. In practice, the sample temperature suggests a dramatic increase of temperature at about 85% of the initial weight, before any significant loss of organic mass is measured, before the first ignition. That is consistent with strongly exothermic, homogenous (gas-phase) reactions of pyrolysis products. Oxidation of CO is the most likely one, as confirmed by the emissions analysis, Figure 3. 15 a rate of progress controlled by the heat dispersion across the sample. Once more, we conclude that the degradation progress is ruled by the heatand mass-transfer rates, rather than the intrinsic, chemical reactivity of the solid, once the latter is significant enough (and that is quite common in air). As frequently reported, increasing the rate of heating generate a false perception of the ignition temperature, if referred to the oven temperature. That is well spotted by the two panels on the right; at 6°/min we conclude that ignition take place approaching 250°C in the oven, while decreasing the rate of heating we reproducibly measure that the ignition starts well below, just above 170°C. Such a large difference in ignition temperature is not evident from the weight loss vs. sample temperature plot (panel on right, above). It is a clear consequence of the increasing difference of temperature between sample and environment at large heating rate. Figure 7: Onset and progress of reactivity in SS-B at 0.1, 0.2, and 2°C/min. In the case of sludge B, Figure 7, the highest (2°/min) heating rate somehow hides the twosteps ignition. The first ignition is still noticeable, but the sample has not enough time to 16 extinguish the first process that the second one begins. The final degradation degree is perfectly reproducible, but the degradation history appears very different in terms of sample or oven temperature. While it is very reproducible when analyzed in terms of temperature of the sample, Fig. 7b, it can be very misleading if the environment temperature is considered, Fig. 7d. Figure 8: Onset and progress of reactivity in SS-C at 0.2, 2, and 6°C/min. In the case of sludge C, Figure 8, we applied the largest HRs, up to 6°/min, because of the apparent difficulty of igniting SS-C. Indeed, SS-C did not ignite at 0.2ºC/min (see Figure 8) like the other sludges (Figure 6 and Figure 7). The behavior of SS-C at the highest HR appears similar to SS-A, with an almost single ignition step. With SS-C the slowest HR (0.2 °/min) does not lead to a vigorous ignition, but still causes almost 40% loss of the initial mass. Again, the rate of heat dissipation competes with the rate of heat delivered by the reaction; however, the degradation without local overheating is observed at 0.2°/min in SSC, and 0.1°/min in SS-A. It suggests that the rate of heat dissipation across the sample is larger in SS-C, perhaps because of a larger particle size. That is consistent with the observation that the minimum HR required to observe the onset of a self-sustained 17 degradation, with distinguished local overheating compared with the environment, follows the sequence of the average particle size: SS-B: Dmean= 0.462 mm HRmin = 0.1°C/min SS-A: Dmean= 0.986 mm HRmin = 0.2°C/min SS-C: Dmean= 1.699 mm HRmin = 2°C/min A heating rate in excess of the minimum provides a rate of heat supply to the sample that overcomes its rate of heat dissipation. These concepts can be summarized semiquantitatively with the energy balance of the sample: 𝑚𝑐&'( ') = $ 𝑄+*+,-)./0 + 𝑄+1+,).02 − 𝑄+'.33.&,)./0 (2) The rate of heat production by the reaction is proportional to the enthalpy and the rate of reaction, DHR and R respectively: 𝑄+*+,-)./0 = 𝑅∆𝐻4= $𝑘(𝑇)$𝐶56 37*8$∆𝐻4 (3) Then, the reaction is supported by the temperature and oxygen availability at the solid surface, inside the granular bed. O2 at the solid interface is a compromise between its rate of consumption and supply from the environment, through the bed, facilitated by larger porosity (larger particle size). The rate of heating by the oven is proportional to the HR, 𝑄+1+,).02~𝑑𝑇 𝑑𝑡 ⁄ while the heat dissipation rate is ruled by the heat transfer rate in the bed 𝑄+'.33.&,)./0~$𝜆+88∇𝑇, through an effective bed thermal conductivity, 𝜆+88, where again the bed porosity is critical. It is concluded that a self-sustained degradation in a sample of significant size is more affected by the physical properties of the solids and its assembly, that rule the heatand mass-transfer rate (such as the particle size and the bed porosity), than the chemical composition of the material ([3], [5], [25]). That is revealed only by testing at a realistic sample size. Eventually, the reactivity reflects a ranking of permeability to oxygen and heat, and that is more evident at higher HR. Also, increasing the HR we may conclude that the temperature threshold to reach selfsustained oxidation is higher than that actually measured in the sample. Although this is a known effect of HR [27], it conveys a false hazard prediction. 18 3.4 Effect&of&the&particle&size& The particle size distribution of the materials, as received and after milling or pelletization, is shown in Figure 9. It has been obtained by sieving. Figure 9: PSD of sludges A, B, C as received, fines of B, milled C and Pellet of A and B 19 The average diameters, Table 3, suggest an order SS-B<SS-A<<SS-C. However, the sludges differ markedly by PSD, with SS-B and SS-C quite monodispersed, while the other materials show a wide range of dimensions. Fines, quite abundant in SS-A and SS-B, dramatically affect the bed porosity. In SS-B, 98% are particles with a diameter < 1 mm, better described as powder than granules. However, SS-A and SS-C look like granular material because of a significant fraction at larger size, approaching 2 mm. The appearance of all the materials tested is shown in Figures SI-1 to SI-4 of Supplementary Material. Table 3: Mean dimensions and specific area of raw and modified sewage sludges Parameters Units SS-A SS-B SS-C B fines C milled A pellets B pellets Mean Diameter mm 0.98 0.46 1.70 0.28 0.28 5.00 5.80 Mean Length mm 11.87 14.89 Specific Area mm-1 6.12 13.04 3.53 21.43 21.43 0.97 0.82 In Table 3 we also calculated the ideal specific surface, assuming spherical (cylinders for pellets), non-porous particles. While the number is just a lower estimation of the true value, the relative ranking is meaningful if the internal porosity of the granules is comparable. In any case, it allows to better understand the results. In the results discussed so far, we speculated that the difficulty of SS-C in developing a selfsustained oxidation could be a consequence of its larger size, supporting a convective heat dissipation through the bed. To prove such hypothesis, we milled SS-C. Its ignition is shown in Figure 10, and compared to SS-A and SS-B as received. 20 Figure 10: Onset and progress of reactivity in SS-A, SS-B, SS-C, and milled SS-C at 0.2 °C/min. The results confirm the hypothesis; a much finer SS-C shows a 2-steps ignition, quite similar to SS-B. Milled SS-C would now be classified as more reactive than SS-A, according to the results of Figure 10. Despite the chemical nature of SS-C being the same, a different PSD causes a considerable difference of apparent reactivity. It means that the reactivity due to the chemical composition is the same, but the environment at the reaction locus is different, and specifically the local temperature and oxygen availability. A finer material reduces the heat and masstransfer rate through the bed. The first limitation allows a higher local temperature to develop, pushing on the rate of reaction. The second limitation (reduction of permeability to oxygen) is a penalty for the reaction. In addition, the smouldering is a heterogeneous, gassolid reaction; as such, the greater the surface of solids exposed to the air, the greater the rate of reaction per unit volume. Indeed, milled C has a specific surface, Table 3, one order of magnitude larger than the original material. According to this, grinding is expected to 21 increment the hazard of storages of the same type of biomass, as intuitive and already reported [28]. The final result is a compromise between the contrasting effects above. Indeed, we observe an earlier ignition in milled SS-C, but the final conversion (approx. 42%) is well below the expectation (approx. 73%), according to the proximate analysis. We further checked such speculations comparing the behavior of SS-B collected in different parts of the drying plant, the regular output of the rotary drier and the filter-bags cleaning the exhaust gases. Once again, the chemical nature is the same, but the SS-B from filters is much finer. The comparison is reported in Figure 11. Figure 11: Onset and progress of reactivity in SS-B and fines of SS-B at 0.2 °C/min. 22 The reactivity of finer SS-B is much higher; the broad PSD increased the bed density and thus the heat insulation, allowing the smouldering front to propagate since the first ignition, up to the total consumption of the organics. On the other hand, pelletization is a process that increases the particles size, while producing much denser solids. We evaluated the effect on both SS-A and SS-B. Results are shown in Figure 12. Figure 12: Onset and progress of reactivity in SS-A and SS-B before and after pelletization, at 0.2 °C/min. The pelletization of sludge significantly limited the reactivity onset and progress. The local temperature rise above the oven temperature is unperceivable in pelletized SS-B and quite small in SS-A, now showing a 2-steps ignition. The final loss of mass was much lower than for granular samples, suggesting that a large fraction of organic matter did not react. We 23 explain the results in terms of a dramatic increase of intraparticle diffusive limitations, making extremely difficult the supply of oxygen to the core of the solids. At the same time, pellets produce coarser bed, where the interparticle heat transfer is facilitated, also by convection, reducing the bed thermal insulation effects. We believe that the difference between pellets of SS-A and SS-B are only partially due to their different chemical nature; more likely, since the self-heating phenomenon is superficial in nature, small differences in the thermal degradation can be attributed to the different value of pellet specific area (A/V), see Table 3, and the pellets length distribution. Also, small variations of pellet compaction are quite likely (density varies from 600 to 700 kg/m3) and the bed arrangement with larger particles is less uniform and reproducible in small vessels. At any rate, results on pellets confirm that materials compaction is an effective mean to mitigate the sludge intrinsic reactivity ([28], [29]). 3.5 Reaction&mechanism& We attempted to collect more evidences of the mechanism formulated above, based only on temperature and weight loss measurements. We tried to look closer at the material while it is degrading. More specifically, we tried to investigate the nature of the 2-steps progress that is shown by all the materials, under different circumstances. It shows as a sequence of 2 peaks in the sample temperature, and correspondingly 2 steps of mass loss. We focused on the most reactive sludge, SS-B, which shows the 2-steps process with the original material, at 0.1 and 0.2 °/min. The mechanism was experimentally studied by stopping the process at 3 sequential stages, highlighted in Figure 13. In addition to the initial, unreacted sludge (1), we collected a sample right after the first ignition (2), one during the vigorous second ignition (3), and a last sample (4) after the second ignition has terminated. Intermediate samples are extracted from the oven and poured on a flat surface, so that the sample cools quickly, stopping its degradation. Because breaking the course of the experiment implies a quenching of the material under degradation, each intermediate sample comes from an independent replica of the test, from the beginning. The proximate analysis of each samples has been carried out, and results added to Figure 13. It is now very clear that heating rapidly removes all the moisture; after the first ignition, a share of volatiles is largely depleted, as a result of both devolatilization (but the loss of weight is not so relevant) and charring. Indeed, the amount of fixed carbon, now including 24 char, is growing (samples 2 and 3). Volatiles total consumption occurs over the second important ignition, together with the oxidation of char. The final residue is mainly ashes (90%), but a 10% of residual char is found, as prove of diffusional limitations to the particle core. Figure 13: Intermediate sampling and their proximate analysis. SS-B, at 0.2°/min The appearance of the material after each ignition peak (i.e. samples 2 and 4) is shown in Figure SI-5 of Supplementary Material. It is immediately evident the dominance of char after the first ignition, and of ashes after the second degradation. The appearance in between (sample 3), right after the vigorous onset of the second ignition is enlightening, Figure 14. It proves that the process evolves in the core of the bed, with large temperature gradients in the bed. A layer of ashes is left at the boundary of the core, possibly affecting the heatand mass-transfer to and from the core. The survival of the core is