Effects of pyrolysis atmosphere on the porous structure and reactivity of chars from middle and high rank coals

The influence of a CO2 or N2-based atmosphere on the porous structure and microstructure of the chars obtained from the nonisothermal slow devolatilization (10 °C/min), from room temperature to 900 °C, of two coals of different rank (Semi-Anthracite (SA) and High Volatile Bituminous type C (HVBC)) and different particle size distribution was studied. Physicochemical characterization (ultimate and proximate analysis), structural and morphological characterization by Raman spectroscopy, FE-SEM, BET surface area, and volume and diameter microporous by CO2 adsorption measurements were carried out for all the chars. It was found that the kinetic parameters, the physicochemical properties, and reactivity of the chars are different, depending on the pyrolysis atmosphere. It was also determined that for the char from SA coal with particle size greater than 0.7 mm, the BET surface area increases when the atmosphere is enriched with CO2. This effect appears to be promoted by the interaction of different processes such as intraparticle side reactions (softening, nucleation and coalescence of bubbles, crosslinking, among others), differences in the thermal diffusivity of N2 and CO2, and the reactive effects of the latter. Additionally, tests of oxidative reactivity of chars showed that the char formed in a CO2 atmosphere is more reactive than that formed in N2. With the results of Raman analysis and kinetic parameters quantified, it was concluded that the reaction atmosphere determined the degree of ordering achieved by the char structure and that the thermodiffusive properties of the reaction atmosphere promoted structural differences in the char even at low heating rates.


Introduction
As a CO 2 -capturing system, oxy-fuel combustion is one of the most promising processes for burning coal due to the advantages in terms of thermal efficiency, carbon burnout, reactor size-reduction, reduction of pollutants, and easy separation of CO 2 from flue gas (Riaza et al., 2014;M. B. Toftegaard, Brix, Jensen, Glarborg, & Jensen, 2010).In this process, oxidation occurs using higher concentrations of O 2 than those used in conventional combustion.To reduce the system and flame temperatures, the flue gases -mainly composed of CO 2 -are recirculated.Additionally, a substantial reduction in NOx occurs.
The substitution of CO 2 for N 2 in the environment modifies the combustion characteristics due to the complexity of the system.(Kim, Choi, Shaddix, & Geier, 2014) and (Tolvanen, & Raiko, 2014), found that the temperature at the particle surface decreased when substituting N 2 with CO 2 even when oxygen participates in the combustion process of char.These differences were also observed in researches of Levendis' Group (Bejarano & Levendis, 2008;Khatami, Stivers, Joshi, Levendis, & Sarofim, 2012;Maffei et al., 2013), during combustion of coal and char, with variations between 150 -200 °C; the variations are attributed to the difference in heat capacity and diffusivity between N 2 and CO 2. Meanwhile, the results estimated by (Heuer et al., 2016), through CFD simulation, concerning the change in temperature were opposite, with a very low difference of only 50 °C, to which it is not possible to attribute the variations between the products obtained under N 2 and CO 2 .Since these investigations were carried out in Drop Tube Reactors (DTR), the discrepancy on the effect of CO 2 in the particle temperature could be partly reconciled when differences in the residence times and the methodology used for the measurement and estimation of the variable are considered.
However, this substitution is a source of controversy due to the contradictory results discovered by researchers Angeles G. Borrego, & Alvarez (2007) and Al-Makhadmeh, Maier, Al-Harahsheh, & Scheffknecht (2013) who carried out oxidation and pyrolysis of bituminous coals, respectively.Angeles G. Borrego, & Alvarez, (2007) studied the oxidation of bituminous coals by creating an O 2 /CO 2 and O 2 /N 2 atmosphere at 1300 °C in a drop tube reactor (DTR).Al-Makhadmeh et al., (2013) studied the pyrolysis of bituminous coals using an entrained flow reactor (EFR) at 1000-1150 in the presence of N 2 and CO 2 .The residence times were 0,3s and 1 -2,5s, respectively.In both cases, CO 2 -chars developed larger surface areas.On the other hand, each atmosphere produced opposite volatile releases.Such differences in the release of volatiles can be explained by the differences between the thermo-diffusive properties (C.Wang, Zhang, Liu, & Che, 2012) and their effects on pyrolysis and nascent porous structure, which is concordant with the explanation given by Saucedo et al. ( 2015), who attributed a lower oxy-combustion rate than traditional combustion to the low diffusivity of O 2 in CO 2 .
Regarding the reactivity, Q. Li et al. (2010) and X. Li et al.( 2010), working with coals (anthracite, bituminous and subbituminous) in DTR, found differences in reactivity between chars obtained in different reaction atmospheres.While for Brix et al.(2010) in EFR, Rathnam et al. (2009) in DTR, Gil et al. ( 2012) in EFR, these differences were insignificant.Overall, these differences can be explained from the chemical composition of coal, the reaction atmosphere, and the heating rate of the solid, which demonstrates the complexity of the oxy-combustion process and the replacement of N 2 with CO 2 .Therefore, there is permanent interest in the study of the effects of reaction atmosphere on the global process and the carbonaceous structure formed during pyrolysis.
However, despite the large number of publications that have been devoted to oxy-fuel combustion in recent years (Maffei et al., 2013;Molina & Shaddix, 2007;Murphy & Shaddix, 2006;Rathnam et al., 2009;Singer, Chen, & Ghoniem, 2013), the quantitative information about the evolution of char chemical structure in the initial stage of oxy-fuel combustion (i.e.non-pure pyrolysis in CO 2 ) is scarce and limited to particle sizes of a few micrometers (Angeles G. Borrego & Alvarez, 2007;Su et al., 2015;B. Wang et al., 2012;Zeng, Hu, & Sarv, 2008).Therefore, is important to note, that despite the apparent abundance of reports on the effects of devolatilization in oxy-combustion, the fact is that only recently paper's on the quality of pyrolysis solid products under oxy-fuel conditions, and exclusively for pulverized coal have been published.In this line of research, the works carried out by Senneca and Schiemann groups (Apicella et al., 2016;Heuer et al., 2016;Pielsticker et al., 2016;O. Senneca et al., 2016) are highlighted.
To explain the changes that occur during the coal devolatilization, regardless of reaction atmosphere, it is necessary to understand the structure of the coal and char (Xu et al., 2016).It is also important to note that the chars are more sensitive to intraparticle processes such as thermoplasticity, crosslinking, swelling, and fragmentation, among others, hindering their prediction (Khan & Jenkins, 1985;Speight, 2013).Studies have focused mainly on devolatilization during the oxy-fuel coal combustion in conventional boilers (small particle sizes) (Gil et al., 2012;Pielsticker et al., 2016;Rathnam et VALDÉS, BETANCUR, LÓPEZ, GÓMEZ, AND CHEJNE al., 2009;Su et al., 2015), but little research has addressed the implications by the change of reaction atmosphere when the particle size increases, either in fluidized bed reactors (with heating rates from 100K/s to 10 000K/s) or in some fixed bed applications (with heating rate of 100K/min) where the behavior of the oxy-combustion changes completely (Bu et al., 2015(Bu et al., , 2016)).In practical applications, devolatilization of coal is strongly influenced by the operational parameters that are dependent on the equipment and coal properties (Yu, Lucas, & Wall, 2007), making the interpretation of results for oxy-fuel coal combustion complicated.
In this research, two ranks of coal (HVBC and SA) in two reaction environments (N 2 and CO 2 ) and three particle size distributions were studied to compare the effects of pyrolysis and pyrolysis-gasification (under CO 2 ) on the carbonaceous structure formed and to identify the effect of particle size.Despite that, research was carried out at low heating rate.To explain these effects, a comparative analysis of mass losses and structural differences for each particle size of chars formed during slow devolatilization was performed.The results confirmed that the reactive effect of CO 2 (gasification) leads to higher mass losses in both coals compared to the process under N 2 .The results also provide evidence for changes in the pore structure of nascent chars as a result of the gasification reaction, intraparticle reactions during the molten phase, and the differences between the thermo-diffusive properties of these two gases.The reactivity to combustion and structural order of different coal ranks were related from the reactivity index and area ratios of the Raman bands (I D1 /I G and I G /I ALL ), It was found that the slight differences between the thermal properties of the evaluated pyrolysis atmospheres are responsible for the structural changes in the char and its reactivity.

Samples preparation
Two Colombian coal rank samples were used in this study: Highly Volatile Bituminous C coal (HVBC) was extracted from the prolific mining region called "Cuenca del Sinifana" located in the Department of Antioquia, while Semi-anthracite coal (SA) was extracted from a mining region located in the Department of Santander.For physicochemical characterization, raw coal samples were crushed and sieved to a particle size less than 0,25 mm.
A 6500 rpm tilt hammer mill was used to reduce the size of the coal samples used in the experiments.Afterwards, they were separated into three homogeneous particle size distributions for testing in the horizontal reactor: 2,36 to 2,00 mm (named HVBCR10 and SAR10), 0,71 to 0,60 mm (named HVBCR35 and SAR35), and 45 to 38 mm (named HVBCR400 and SAR400).The chars of the carbon samples were named in the same way, adding at the end the environment under which they were obtained.
Table 1 presents the results of the physicochemical characterization of raw coal samples which were carried out following the standard methods listed in the bottom row.The error shown here represents the population standard deviation (sd) of different analyses.(Thommes, 2010).Both coals have similar mesoporous distributions (see Figure 1a and b) and present the bimodal PSD (with slightly higher pore volume for coal of lesser rank).This follows the results of Prinz, Pyckhout-Hintzen, & Littke (2004) and Sakurovs et al. (2012), who determined that the PSD's are strongly affected by coal rank.In addition, Sakurovs et al., (2012) showed that pores in the size range 2 to 25nm are less common in coals of higher rank.
No major differences exist in the microporosity of the coals, comprised of micropores of diameter less than 0,7nm (also named ultramicroporos (Thommes, 2010)).
The microporous volume values are similar (see Figure 1c  and d), which justifies the similarity in their behavior under adsorption, despite their differences in coal rank.

Equipment and experimental procedure
Slow pyrolysis was carried out in a horizontal reactor using a heating rate of 10 ºC/min (see Figure 2a).This reactor consisted of a quartz tube, external diameter of 3,5 cm (ID: 3,0 cm) and length of 50 cm, inserted into a horizontal tube furnace that is heated by PID-controlled electric resistance using two type-K thermocouples that measure the reactor and sample temperatures.A sample of 3,0g was placed into the reactor in a quartz sample holder that was 2,0 cm in diameter and 13,5 cm long.The thermal treatment was applied from room temperature and increased at constant rate up to a maximum temperature of 900 °C under N 2 or CO 2 environments with analytic grade gases (purity = 99,999 %) flowing through at 50 ml/min.Before thermal treatment, the equipment was purged for 40 minutes using purely N 2 or CO 2 atmosphere for each respective test.Char reactivity and pyrolysis tests for coal samples were performed using a TGA balance (see Figure 2b).Then, for chars obtained from the pyrolysis of both coal samples and both reaction environments with particle size distributions between 45 and 38 mm, (Figure 2b) combustibility tests were implemented.The measurements were performed under non-isothermal conditions following the same procedure described by the FWO method (Flynn-Wall-Ozawa) (Ozawa, 1965;Flynn & Wall, 1966) as described by Valdés, Marrugo, Chejne, Román, & Montoya (2016).A particle bed height of less than 2 mm was fixed, which corresponded to 10mg of sample in the crucible placed in the TGA.To dry the char samples, they were heated from room temperature to 110 °C with a heating rate of 10 °C/min and held 30-min under N 2 (50 ml/min).Then, coal samples were pyrolyzed in N 2 (500 ml/min) and chars samples were heated under the mixing N 2 /O 2 (50/13 ml/min).Pyrolysis tests such as combustibility were carried out from 110 °C to 900 °C at rates of 2, 5, and 10 °C/min until they were held for 60min for complete burnout.
Slow pyrolysis and reactivity tests were performed following a factorial experimental design with 22x3 triplicates.Likewise, the products of each test were characterized in triplicate.The error bars in each figure correspond to the calculation of the standard deviation of

VALDÉS, BETANCUR, LÓPEZ, GÓMEZ, AND CHEJNE
the population of tests realized for the same sample.The obtained calculations follow the uncertainty propagation rules for the parameters on DAF basis (Dry Ash Free).In order to compare the reactivity of each char sample, the sample temperature at 35 % conversion was used as the reactivity index, ensuring an index that it is not affected by the residual free moisture in samples.

Coals and chars samples analysis
Coal and char samples were reduced to particle sizes less than 0.25 mm and characterized in an elemental analyzer (EXETER, model CE-440) using the ASTM D5373 standard.Then, the proximate analysis was carried out using a thermogravimetric analyzer (TGA), employing the method described by Shi, Liu, Guo, Wu, & Liu (2013).At the end of each experiment, residual mass was removed from the sample holder and determined using an analytic balance Mettler Toledo brand, XP205 model, with precision ±0,01 mg, with that the mass loss was calculated.This was used to calculate the mass loss percent for each test.The characterization of the carbonaceous microstructure was performed by Raman spectroscopy, on a Horib Jobin Yvon brand, Labram high-resolution model, laser He/Ne 17MW 633nm and 785nm diode laser 80mW.The porous structure of the char particles was characterized through the thermal process: samples were previously dried for 24 hours at 120 °C and raw coal samples were degassed at 150 °C for one hour (for chars samples, the degassing is carried out at 300 °C).Subsequently the samples were characterized with the mean diameter (md) and mean volume (mv) of a pore, using the Horvath-Kawazoe (HK) method and pore size distribution (PSD) through N 2 adsorption (BET -Brunauer-Emmett-Teller) and CO 2 using a Micromeritics TriStar II Plus with precision ±0,02 m 2 /g.Finally, the morphological visualization of the carbonaceous structure of coals and chars was performed in a FESEM, brand JEOL JSM 7100F Model.

Effect of pyrolysis atmosphere by TGA
Experiments were carried out in TGA balance (see Figure 2b) to determine the kinetic parameters of the devolatilization of the two coal samples, with particle diameters between 45 and 38 µm, to elucidate the kinetic differences and similarities during pyrolysis in N 2 or CO 2 .Figure 3 shows the DTG curves obtained from the mass loss tests under N 2 and CO 2 -atmosphere in both coals (See Figure 3a, c and e to SA coal and Figure 3b, d and f to HVBC coal).For both coal rank and independently of the heating rate, it was found that at higher heating rates, there is a greater mass loss rate, as reported by other studies (Niksa, Heyd, Russel, & Saville, 1985;Strezov, Lucas, Evans, & Strezov, 2004;Valdés et al., 2016).In the mid-rank coal (HVBC) (see Figure 3b, d and f), the stage of the primary devolatilization represented in the first peak between 400 and 460 °C was clearly observed, which is related to its lower rank, its higher volatiles content and higher reactivity.For SA coal, its lower volatile content, harder structure and lower pore volume (see Figure 1a) leads to a higher primary devolatilization temperature (> 550 °C).
In SA samples, above 650 °C mainly in N 2 -atmosphere and at the lower heating rate, small peaks were observed, which correspond to secondary reactions and they are more frequent with the increase of the coal rank.This is related to the low volatiles content, the low local permeability and longer duration of the plastic phase of these coals (Gavalas, 1982;Oh, 1985;Perera et al., 2011).This shows that secondary reactions are difficult to prevent in TG experimental equipment.Under CO 2atmosphere above 700 °C, a grand peak was observed, at which point CO 2 can oxidize the solid surface (CO 2gasification reaction).
The analyses of mass loss and mass loss rate were used to calculate the activation energy, describing the thermal devolatilization mechanism of both coals during the pyrolysis processes with different conversion rate (a), using a method for various heating rates (more methodological details can be consulted in Valdés et al. (2016)).The results presented in Figure 4 suggest that since the porous structure of both coals is similar (See Figure 1), the physicochemical composition and the reaction environments are responsible for the variations in the devolatilization kinetic rates.The Figures 4a and 4b show that for temperatures below 378 °C (where devolatilization of HVBC coal had already started), in HVBC coal the Ea is lower than in SA coal, in which, the devolatilization is still insignificant.In HVBC coal samples at conversions lesser of 0,35 (410 °C), the Ea under N 2 -atmosphere is minor than under CO 2 -atmosphere.However, a higher reaction rate was exhibited under CO 2atmosphere, which can be explained through decarbonation reactions (CaCO3 and MgCO3 in the ashes) by thermal decomposition (Johnson, Rostam-Abadi, Mirza, Stephensn, & Kruse, 1986), and degradation of the coal matrix.The latter starts with the CO 2 production by decomposition of carboxyl group at lower temperature and then the split of ethers, ketones and oxygen containing heterocycle (S.Li, Ma, Liu, & Guo, 2016;Miao, Wu, Li, Meng, & Zheng, 2012), pyrolysis water and cracking of the few released aliphatic in the narrow porous structure developed up to that moment (Arenillas et al., 2003).Therefore, diffusive type restrictions of volatile matter released in CO 2 can be irrelevant to the reaction rate at these temperatures.
At conversions between 0,35 and 0,55 (410 to 472 °C), the Ea and the rate of reaction during the devolatilization of the HVBC is higher in N 2 than that in CO 2 , which is consistent with the intervals in which the primary devolatilization, where the higher mass loss rate was reached.These results can be explained through the greater diffusive type restrictions that volatiles have in the presence of CO 2 (Molina & Shaddix, 2007;Valdés et al., 2016), which implies a minor interaction between the CO 2 -atmosphere and char (less occurrence of secondary processes between CO 2 and char at low temperatures).Therefore, a reduction in the values of the Ea and reaction rate can be expected.
Given the lower content of volatile matter for SA coal and that its devolatilization reaction is significant at higher temperatures (greater at 525 °C, see Figure 4a), the Ea and reaction rate for the different stages coexisting during devolatilization (pyrolysis and gasification) of SA samples, were lower than that for HVBC coal until a 55 % conversion is reached (see Figure 4).This implies a larger energy barrier during the primary devolatilization; nevertheless, the higher reaction rate exhibited by the lowest coal rank is determined by the higher content of released volatile material and not because of the reaction atmosphere.This can be explained by the reactive effects of the atmosphere on some of the volatile species which are released and the development of porous structure (Yin & Yan, 2016).Also, the high temperature (above 600 °C) favors the occurrence of the CO 2 -gasification reaction (Su et al., 2015;Valdés et al., 2016;Xu et al., 2016); which provided a slightly higher reaction rate when the coals were pyrolyzed under a CO 2 -atmosphere compared to N 2 (see Figure 4c and d).
The results presented in Figure 4 also confirm that given the higher content of volatile matter in the lower rank coal under the experimental conditions, the HVBC is ten times more reactive during pyrolysis (reaction rate HVBC >> SA coal), despite requiring a greater activation energy.However, HVBC coal have a higher pre-exponential factor (C. Wang et al., 2012), which, by definition, indicates a higher probability for reaction occurrence.Additionally, it was observed that during the pyrolysis of HVBC coal, the reaction rate is higher in N 2 than in CO 2 until temperatures around 600 °C.Above this temperature, the CO 2 takes the already demonstrated oxidant properties that increase the reaction rate in this atmosphere.The results for temperatures below 600 °C can be explained through the differences between the thermo-diffusive properties of CO 2 and N 2 , see Valdés et al. (2016).In CO 2 , due to a lower rate of diffusion of volatiles released, intraparticle secondary reactions can be promoted (Molina & Shaddix, 2007; C. Wang et al., 2012).Further, CO 2 can react with the carbonaceous structure (Gonzalo-Tirado & Jiménez, 2015; Gonzalo-Tirado, Jiménez, & Ballester, 2013; Hecht, Shaddix, Geier, Molina, & Haynes, 2012).While, when N 2 atmosphere is used, its lower volumetric energy density (kJ/ m 3 ) alters the heat transfer between phases, causing the most thermally severe pyrolysis in this environment.

Effect of pyrolysis atmosphere on char structure evolution
The total mass loss of the chars prepared in the N 2 and CO 2 atmospheres until 900 °C in the horizontal reactor are shown in Figure 5.The mass loss was higher for the coal sample of lower rank (HVBC) than for the higher rank (SA) sample.Additionally, the mass loss was higher in the CO 2 atmosphere than in the N 2 atmosphere, concordant with the results of other research (Su et al., 2015;B. Wang et al., 2012;Zeng et al., 2008).The mass loss in all samples of HVBC coal pyrolyzed in N 2 and CO 2 atmospheres was greater than the sum of the volatile content and moisture for the proximate analysis reported (see

VALDÉS, BETANCUR, LÓPEZ, GÓMEZ, AND CHEJNE
Additionally, for samples of HVBC coal, mass loss in CO 2 atmosphere was marked significantly greater with increases of the particle size, whereas the samples processed under N 2 did not show observable differences.This is also attributed to the differences in thermo-diffusive properties between CO 2 and N 2 (Molina & Shaddix, 2007; C. Wang et al., 2012;Valdés et al., 2016).
For SA coal, overall mass loss was always slightly lower than the sum of the content of volatile matter and moisture.This could be a consequence of the strong influence of the thermodeactivation by thermal annealing or by thermoplastic properties of coal, which were much more relevant for high rank coals.This has been studied by Gavalas (1982), Peter R. Solomon, Serio, & Suuberg (1992), and Perera et al. (2011) who found that pore number and permeability decreases as coal rank increases.The presence of the thermoplastic phase favors the occurrence of intraparticle secondary reactions and encapsulation of volatile matter (Hayhurst & Lawrence, 1995;Larsen, 2004).
10, respectively.This means that the process is governed by the release of primary devolatilization products, regardless of the reaction atmosphere since the process is in the kinetically controlled regime.Additionally, this is consistent with studies of scales characteristic times of the phenomena occurring during thermochemical transformations, carried out by Dufour, Ouartassi, Bounaceur, & Zoulalian (2011) and Stark (2015), with biomasses.
With the aim of validating the reactive effects of CO 2 and the implications of the change in the particle size on the chemical composition and the porous structure developed during the process of pyrolysis, char samples were characterized through proximate and ultimate analysis, as shown in Table 2 in DAF basis.The mean values of VM, FC and C for the chars reflect CO 2 -volatile and CO 2 -char reactions.Regardless of rank and particle size, the FC was always lower under the CO 2 atmosphere than under N 2 .Similar effects of CO 2 have been observed by other researchers in gasification studies (Gonzalo-Tirado et al., 2013;Rathnam et al., 2009;Su et al., 2015).The VM content in chars was always higher for those obtained in a CO 2 atmosphere.This also can be explained by the lower diffusion of volatiles, occurrence of secondary intraparticle reactions, and changes in the local permeability of the chars due to the atmosphere.It is important to highlight that researchers such as Khan & Jenkins (1985) and Larsen (2004) have provided great evidence of the plasticizing effect of CO 2 in carbonaceous materials.It has been shown that under pressurized conditions (Khan & Jenkins, 1985;Larsen, 2004) and/or The differences in mass loss are not significant for particle sizes between 45 and 38 mm.In the case of samples treated under N 2 atmosphere, a possible explanation, despite the large differences between the heating rates used, is thermodeactivation of the char by thermal annealing, as shown by Osvalda Senneca, Salatino, & Masi (2005).This prevents, in the case of the tests under CO 2 , the gasification reaction of Boudouard from having an impact on the process.The previous explanation does not agree with the results of another research by Osvalda Senneca & Cortese (2012), where they show that for a heating rate and coal of similar range to the one used in the present investigation, the contribution of the gasification reaction is significant; however, such a discrepancy may be due to the large difference between the ash contents of both coals.
Another possible explication for both atmospheres may be due to the smaller particle size: there is increased exposure of volatile material to the surrounding environment, coupled with the low gas velocity (equivalent to 1,2x10 -3 m/s).The Biot and Pyrolysis numbers are less than 0,1 and greater than thermal gasification conditions (Khan & Jenkins, 1985), dissolution of CO 2 in coals is possible.This phenomenon manifests itself as slight swelling or shrinkage of the particles by reordering of their physical structure.CO 2 also reduces the softening temperature of the coal (Khan & Jenkins, 1985).Therefore, the result of the chemical characterization of chars and coals shown in Table 2 gives evidence that during devolatilization, CO 2 enhances thermoplastic behaviors in both coals.Thermoplastic behavior contributes to chars with a higher content of volatile matter due to encapsulation of volatiles in the plastic phase (Hayhurst & Lawrence, 1995;Larsen, 2004).It consequently leads to more intraparticle secondary reactions.This is consistent with many proposed models of devolatilization, including one by P. R. Solomon, Hamblen, Carangelo, Serio, & Deshpande (1988), which shows that the viscosity of coal in its plastic state has an inverse relationship with the diffusion coefficient of the pyrolysis products.
The physical structures of the chars are also affected by the different reaction conditions.Measurement the porous surface area was performed via adsorption/desorption with N 2 (see Figure 6).It is evident that the BET surface area (Figure 6b) increased with particle size for HVBC in both atmospheres compared to the raw samples (Table 1).However, the surface area does not increase for the smaller particle size regardless of the reaction environment.Additionally, a higher development of BET surface area for the HVBC chars obtained under an atmosphere of CO 2 versus N 2 is reflected in the larger pore volume (see Figure 7d and f) and is consistent with the results found by Zeng et al. (2008), which used particle sizes between 106-125 mm.
In this research, for particle size between 45-38 mm, neither coals (HVBCR400 and SAR400) nor environments (N 2 and CO 2 ) showed variation in the BET surface area.This is consistent with the results reported by Brix et al. (2010), who worked with similar ranking coals but with particle sizes between 90 and 105 mm.
Brix et al., found a similar morphology of the residual char and a similar volatile release in both atmospheres.A possible explanation is that with small particle sizes, the volatiles are more exposed to the surrounding atmosphere, favoring their release without causing significant evolution of BET surface area by intraparticle diffusive processes or reaction of the surrounding environment with the carbonaceous structure.However, the volatiles release from the coal microstructure favors the development of small pores (less than 0,7nm).
The PSDs for chars named SAR400 and HVBCR400 (see Figure 7a and b), in both atmospheres, are practically unchanged in comparison to raw coals (the mesoporous distribution continues to be bimodal) particularly for HVBCR400 samples independent of reaction atmosphere.
Mesopores with diameters between 2-10 nm appeared in SAR400CO 2 char samples (see Figure 7a), evidenced by a slight increase in BET surface area (see zoom in Figure 7a) and an increase of micropore volume with a decrease in the average diameter when the particle size decreased (see Figure 8).The authors hypothesized that variations in surface area with different particle sizes are due to the combined effect of the carbonaceous structure and reaction environment on interactions with volatiles and char.This has been explained and demonstrated for devolatilization in N 2 (Borah, Ghosh, & Rao, 2008;Ross, 2000), where increased residence times of volatiles in the particle led to more char.In this research, when the volatile matter content was high as in the case of coal HVBC, it was very likely that during transport of the products of primary devolatilization, bubble-explosions generated pore volume.This increased porosity (see Figure 7b), when the devolatilization atmosphere was CO 2 , it could have occurred because in this environment, the migration of volatiles was more complex.This is due to the lower diffusion coefficient of volatiles in CO 2 , which causes bubbles to nucleate and subsequently explode, leading to large pore development.Therefore, as observed for char obtained under CO 2 atmosphere, there was a greater generation of pore volume and an increase in surface area.This is consistent with the results reported by A.G. Borrego, Garavaglia, & Kalkreuth (2009), who concluded that CO 2 promotes cross-linking reactions and the formation of more pore volume.
gasification reactions occurred (Bu et al., 2016;Xu et al., 2016).Gasification is likely responsible for the variation of surface area in these samples (see Figure 6a).In coal of high rank (SA), having a low content of volatile material and a more organized structure, BET surface area is only developed for the large particle sizes (SAR10, > 2 mm), as shown in Figure 6a.When the coals are pyrolyzed in a CO 2 atmosphere, the CO 2 -char interactions (gasification) produce a slight development of BET surface area.This is also evidenced in the PSD of SAR10CO 2 samples (see Figure 7e), where there is considerable evolution of mesoporosity between 2 and 10nm.
To complement the morphological study, all coal and char samples were visualized using a FE-SEM.The photomicrographs in Figures 9 and 10 illustrate the differences between samples of coal and chars prepared under different atmospheres.
The first column in figures 9 and 10 (a, d and g) shows the unpyrolyzed coals (R10, R35 and R400) for SA and HVBC.For both, the surfaces have irregular edges along fractures and the internal surfaces are rough; this is more apparent in SA than in HVBC.These fractures likely come from the milling process and are sites where volatiles can be released during the thermochemical processing.
SAR10 chars have well defined edges but the internal walls are smoother under N 2 than under CO 2 (Figures 10b and  c; respectively).This could be caused by a more violent release of volatiles and plastic transformation during pyrolysis under N 2 .The highest heating rate is evidenced in this environment because of lower energy density (r • Cp) in comparison to the energy density of CO 2 (Kim et al., 2014;Molina & Shaddix, 2007;Tolvanen & Raiko, 2014).Also, the presence of spongy masses of micronic diameter (1 to 2 micrometers), similar to those found by Heuer et al. (2016), and characterized in detail by Apicella et al. ( 2016) was visible.In agreement with the work of Heuer et al., more of these micronic particles were found in the char obtained under atmosphere of CO 2 than under N 2 .In general, chars obtained under CO 2 have a rougher surface and provide evidence of the chemical etching which implies that  and c) show char samples obtained under N 2 and CO 2 ; respectively.It is observed that due to the thermochemical processing, holes were developed by release of volatiles with more holes present in the char from CO 2 atmosphere (Riaza et al., 2014).
In accordance with the effects on the char with reduced particle size, chars from both R35N 2 and R400N 2 coals (subplots e and h, respectively, Figures 9 and 10) have smooth surfaces with very few holes.However, chars with particle size R35 and R400 (f and i) obtained from both coals under CO 2 atmosphere show that for smaller distributions, there was not much superficial exfoliation; this behavior is contrary to R10CO 2 (Figures 9c and 10c).The presence of smooth, rounded edges was also common in both atmospheres, but more marked in the N 2 atmosphere.These effects are characteristic of a material that passed through a plastic phase.Therefore, it is likely that as the particle size is reduced, the differences between the two environments are not relevant to the primary devolatilization reactions.With the low particle size, the thermal gradients are minimal.However, char produced under CO 2 still exhibits softening effects compared to N 2 .

Impact of pyrolysis atmospheres on char micro-structure evolution
Raman spectroscopy analysis was performed for samples of raw coal and chars obtained under both reaction atmospheres (N 2 , CO 2 ) and three different particle size distributions.Raman spectra were corrected with the same baseline, then smoothed and normalized with respect to the total area of each Raman analysis.In Figure 10, the spectra of samples of coal (HVBC and SA) and chars are presented for the particle size denoted as R400.In all cases it is possible to observe bands D 1 and G, characteristic of carbonaceous materials, positioned around 1350 and 1580 cm-1; respectively (Reich & Thomsen, 2004;Tuinstra, 1970).For more information on the structural characteristics of the tested samples, the deconvolution was performed in 6 bands for raw materials and 4 bands for chars using Lorentzian functions.
It is observed in Figure 11a and 11d that the spectra of the coal samples are different, but characteristic of their respective ranks.Since the intensity of the graphitization band for SA coal is greater than for HVBC coal, greater organization is present in the higher-ranking SA coal.An increased intensity of D 2 and D 4 bands observed for HVBC coals, which is characteristic of materials with a very poor organization, is consistent with them being lowerranking coals.Among the chars, independent of coal and devolatilization atmosphere, no major differences were observed.This may be because the low heating rate of thermochemical processing.The slight differences showed the effects of reaction atmosphere on microstructure at a high heating rate.
The differences are expected to be more remarkable with a higher heating rate.In addition, the spectra validate that during the thermochemical transformation of coal to char, migration occurs in long-chain aliphatic structures, alky-aryl ether, C-C and methyl groups associated with aromatic rings denoted sp 3 , according to Table 3 and grouped in the band SL (1250 cm -1 ), as described by X Li, Hayashi, & Li (2006).
Since the qualitative differences are very small, the area ratios of the bands and all Raman spectra were quantified.This are referred to as I D1 /I G , I D3 /I G , I D4 /I G and I G /I ALL .Figure 12 shows the different area ratios for all coals and chars devolatized in both atmospheres.Figures 12a) and 12c) show changes in the relative intensities of the bands I D1 and I D3 with respect to I G .It can be seen that for SA coal, the ratios I D1 /I G , I D3 /I G are slightly lower for thermal treatment under CO 2 atmosphere compared to N 2 treatment, while for HVBC coal (see Figures 12c and d) no significant differences were found. .CharSA-N 2 ; c).CharSA-CO 2 ; d).HVBC; e).CharHVBC-N 2 ; f).CharHVBC-CO 2 .

Source: Authors
The lower values for the I D1 /I G ratio in SA chars under CO 2 atmosphere and three particle sizes can be explained by the increase in the condensation of aromatic rings as was first noted by X Li et al. (2006).While not necessarily leading to the decrease of defects in the structure, the high condensation of amorphous carbon due to random arrangement could generate more defects in the structure.

Source: Authors
The results also agree with the works reported by Sheng (2007), Zhu & Sheng (2010) and C. Wang et al. (2012).Using lower rank coals, they found that the heat treatment leads to an ordering of the carbonaceous structure with respect to increasing temperature (see Figure 12b and d for HVBC chars).This is corroborated with the decrease of I D3 /I G , I D4 /I G , and I D1 /I G and the increase of I G /I ALL for chars,

Impact of pyrolysis atmospheres on char reactivity
Since the overall mass losses for both coals with particle sizes between 45 and 38 mm were not strongly affected by the reaction atmosphere and that BET surface area remained unchanged, testing of the oxidation reactivity of chars obtained under different atmospheres was performed to identify the chemical changes.Reactivity was measured by TGA using the FWO method (Valdés et al., 2016), under an oxygen environment (O 2 /N 2 = 21.6/78.4%v/v).
As shown in Figure 13, the reactivity to combustion was greater for chars obtained under CO 2 than under N 2 .This is consistent with the results of intensity of the Raman bands and intensity band analysis explained above.Therefore, it is confirmed again that the chars produced under a CO 2 environment are more disordered than those obtained under N 2 atmosphere.Differences in reactivity can also be explained by the surface area and the PSD of the chars obtained in both atmospheres: a greater surface area leads to higher reactivity.This demonstrates that reactivity is dependent on the structure of the coal chars produced under both environments.These results are in agreement with other research on coals of similar rank (Angeles G. Borrego & Alvarez, 2007;C. Wang et al., 2012;Zeng et al., 2008) and they disagree with those obtained by Senneca and Schiemann groups (Apicella et al., 2016;Heuer et al., 2016); However, the differences between experimental conditions are the main cause of results divergence.

Correlation between structural changes and char reactivity
A correlation between the reactivity and the area ratio of bands from Raman spectrum characteristics of carbonaceous materials was found.These studies seek to demonstrate the existence of a relationship between the decrease in the reactivity and increase in the order of the microstructure by thermal treatment.This is reflected in the independent of reaction atmosphere and particle size.The decrease of I D3 /I G , I D4 /I G , and I D1 /I G have been confirmed by other researchers (Sheng, 2007;C. Wang et al., 2012;Zhu & Sheng, 2010).The decreasing trend of the ratio I D1 /I G suggests an increase in the mean size of graphitic microcrystals during the thermal treatment, while the decrease of the ratios I D4 /I G and I D3 /I G gives information about release of volatile matter, the removal of active sites, and imperfections of amorphous carbon crystals.Regarding the effect of the CO 2 atmosphere on the level of structural arrangement achieved by HVBC chars, an increase for I G /I ALL was observed, especially for the R10 particle size, (contrary to what occurs in the char obtained under N 2 ).This is related to the greater influence of intraparticle reactions to this size, as evidenced in the BET and visualization FESEM analyses.A more disordered char structure is obtained under pyrolysis in CO 2 due to the oxidant effects that favor the concentration of structural defects in various forms in coal, and promote greater influence of plastic processing (Sheng, 2007;Su et al., 2015;C. Wang et al., 2012).
For SA chars evaluated in both environments (see Figure 12a and c), the development of ratios I G /I ALL and I D3 /I G decreased with the thermochemical treatment, regardless of particle size and reaction atmosphere.This shows that during the thermochemical processing, a disruption of char structure occurs according to the ratio I G /I ALL .Therefore, the carbonaceous char microstructure from SA coal exhibited less order than the raw coal.This effect was lower for smaller char particles and also while using the CO 2 atmosphere.This is explained by the strong thermo-plastic effects in high-ranking coals, especially with larger particle size (P.R. Solomon et al., 1988).Additionally, given the oxidant effect of CO 2 and the low content of volatile matter, the interaction between environment-char is stronger than that in HVBC.
increased value of the area ratio of the bands I G /I ALL and decrease in the area ratios of the bands I D1 /I G , I D2 /I G , I D3 /I G , and I D4 /I G .
The results are presented in Figure 14 for both coals and chars.The coals were evaluated under reaction atmospheres of CO 2 and N 2 , while for chars the oxidation reactivity to an O 2 /N 2 atmosphere was evaluated.Figures 14c and 14d present the relationship between the reactivity index and Raman ratios (I D1 /I G , I G /I ALL ) for HVBC coal and chars.The results showed an increase in reactivity index of the carbonaceous structures when the ratios I D1 /I G and I G /I ALL increased, which is consistent with the most ordered structure exhibited by the char obtained under N 2 .
Figures 14c and 14d also reveal that the reactivity index due to oxidation exhibited by the structure of the char from HVBC coal is greater than that determined for raw coal during devolatilization in either atmosphere.These relationships explain that HVBC coal is more reactive than its respective chars, because of its middle-ranking, lessordered structure which cause their lower reactivity index under devolatilization.Additionally, the results reveal that HVBC chars obtained under N 2 atmosphere are less reactive than those from CO 2 , which is consistent with the results presented in Figure 13 (dashed lines) and obtained by Zhang et al. (2015), for a similar rank coal.Those researchers attributed this difference to a high concentration of defects in the crystal structure caused by the reaction atmosphere, differences in heating rate, and to the greater energy density of CO 2 compared to N 2 , which leads to a lower value of the I G /I ALL relationship in the structure of the char obtained under CO 2 .
obtained under N 2 atmosphere is more reactive than the SA coal, decreasing its reactivity index.Meanwhile those obtained under CO 2 atmosphere lessen their reactivity index under devolatilization but remain more reactive to oxidation than those obtained under N 2 .The increase in the ratio I D1 /I G and decrease in the ratio I G /I ALL agrees with the findings by X Li et al. (2006) and Zhu & Sheng (2010).They explained that these trends are due to the increased concentration of aromatic rings, having 6 or more fused benzene rings as a result of the hydrogenation of aromatics.This behavior is possible due to the occurrence of annealing phenomena under N 2 atmosphere, given its lower energy density, which inhibits the evolution of the structure to a more ordered system.The reactive effects of CO 2 are so complex that while improving the de-hydrogenation of hydro-aromatics and promoting the growth cluster of aromatic rings, they can also introduce functional groups with oxygen, which reflect in a more disordered structure.

Conclusions
In this study, a higher reactivity was observed for HVBC chars prepared under a CO 2 atmosphere in comparison to those prepared in N 2 .However, the difference in reactivity between SA chars formed under N 2 and CO 2 was not significant.The dependence of these effects is explained by gasification reactions, thermo-diffusive properties of the volatiles released in surrounding atmosphere, and the grade of influence of thermoplastic properties of coals.The latter is more relevant in high-ranking coals.In the same sense, it is explained that the surface area development is greater for the chars obtained under CO 2 atmosphere than those with N 2 atmosphere due to bigger residence times for contact between the gases and the carbonaceous structure.
In regards to microstructure, the lower value of I D1 /I G ratio, obtained for HVBC chars compared to coal samples and the higher reactivity during coal pyrolysis, is related to highly disordered structures.These structures can be more abundant in the chars obtained under the presence of CO 2 than under N 2 due to the gasification reactions.
Additionally, the increased presence of volatile matter and oxygen in this material led to an increase in porosity during the thermochemical process.Thus, a decrease in the crystalline domain size of the carbonaceous network is favored.Finally, the effects on the nascent char in CO 2 atmosphere are more complex, to the point that it is also likely that new structures formed contain a higher concentration of hetero-atoms and oxygen functional groups than could have been introduced by the gasification reaction.This also explains the increase in Raman intensity as oxygen increases the dispersion, confirming that more disordered char structure is formed under CO 2 .

Figure 1 .
Figure 1.Pore size distribution of coal samples (NLDFT model) a) and c) SA; b) and d) HVBC.

Figure 5 .
Figure 5. Mass loss of the samples pyrolyzed in horizontal reactor under N 2 and CO 2 atmospheres compared to combined value of volatile matter and moisture.Source: Authors

Figure 9 .
Figure 9. SEM micrographs of SA coal and chars formed in different atmospheres (CO 2 and N 2 ).Source: Authors

Figure 12 .
Figure 12.Ratio of band peak areas of different chars prepared in N 2 and CO 2 atmospheres.a) SA in N 2 ; b) HVBC in N 2 ; c) SA in CO 2 and d) HVBC in CO 2 .■ I G /I ALL ; ■ I D3 /I G ; ■ I D4 /I G ; ■ I D1 /I G Source: Authors

Figure 14 .
Figure 14.Correlations between the reactivity index and the band area ratios (a and b) SA and (c and d) HVBC ■ under CO 2 atmosphere ■ under N 2 atmosphere.Source: Authors The analysis of the evolution of I D1 /I G and I G /I ALL ratios for SA coal and chars presented an opposite behavior to that shown by coals of lower rank.The results of Figures 14a and 14b indicate that with the increase of the ratio I D1 /I G and decrease of the ratio I G /I ALL , the structure of the char

Table 3 .
Summary of Raman peak/band assignment