Thermogravimetric , Devolatilization Rate , and Differential Scanning Calorimetry Analyses of Biomass of Tropical Plantation Species of Costa Rica Torrefied at Different Temperatures and Times

We evaluated the thermogravimetric and devolatilization rates of hemicellulose and cellulose, and the calorimetric behavior of the torrefied biomass, of five tropical woody species (Cupressus lusitanica, Dipteryx panamensis, Gmelina arborea, Tectona grandis and Vochysia ferruginea), at three temperatures (TT) and three torrefaction times (tT) using a thermogravimetric analyzer. Through a multivariate analysis of principal components (MAPC), the most appropriate torrefaction conditions for the different types of woody biomass were identified. The thermogravimetric analysis-derivative thermogravimetry (TGA-DTG) analysis showed that a higher percentage of the hemicellulose component of the biomass degrades, followed by cellulose, so that the hemicellulose energy of activation (Ea) was less than that of cellulose. With an increase in TT and tT, the Ea for hemicellulose decreased but increased for cellulose. The calorimetric analyses showed that hemicellulose is the least stable component in the torrefied biomass under severe torrefaction conditions, and cellulose is more thermally stable in torrefied biomass. From the MAPC results, the best torrefaction conditions for calorimetric analyses were at 200 and 225 ◦C after 8, 10, and 12 min, for light and middle torrefaction, respectively, for the five woody species.


Introduction
Biomass is widely available worldwide, and often used in biofuel production to help reduce the use of fossil energy reserves and mitigate problems the environmental problems caused by petroleum derived fuels [1].In addition, biomass produces lower carbon dioxide (CO 2 ) emissions, as it maintains the carbon cycle by freeing the carbon that was previously fixed during photosynthesis [2].However, despite the importance of biomass, it has not been developed into other types of energy, such as hydroelectricity, eolic, or solar energies, which are being highly exploited [3].
Despite the increase in the use of biomass as an energy source, some disadvantages still limit its optimum performance, such as difficulties in collection due to disperse distribution, irregular shape, high volume, low energy density, and storage and transportation problems [2].These problems affect the variability of the biomass's physical properties [4].Other challenges posed by biomass include its Energies 2018, 11, 696 2 of 26 low calorific power, high moisture content, and hygroscopic nature that cause economic problems and deficiencies during transportation, handling, storage, and conversion of the material [5].
Technologies that use thermal treatment of biomass could be applied to solve the mentioned difficulties and convert biomass into energy via combustion [6].Of the thermal treatments, torrefaction appears to be an effective solution [5].Studies have shown that torrefaction increases the energy density of the biomass and reduces its hygroscopicity [7,8].Biomass torrefaction occurs at temperatures from 200 and 300 • C, at atmospheric pressure, and in the presence of inert atmosphere, meaning a limited oxygen presence during the process.The advantages of torrefaction include calorific power increase, reduction of O-H and H-C, lower, moisture content, higher hydrophobicity, and better grinding capacity [9][10][11][12].
The success and quality of the torrefied biomass depend on the temperature and time of torrefaction, mainly due to the heterogeneity of the chemical composition of the biomass [13].Among the chemical components of biomass, lignin is the most thermally stable, followed by cellulose and then by [5,12,14].For average ranges of torrefaction, the component that degrades the most is hemicellulose as well as some non-structural components, such as extractives [9,12,15].
Important changes in biomass composition were observed using thermogravimetric analysis (TGA) after torrefaction [16].The curves in this analysis demonstrate the thermal stability of the components of the biomass, including the mass loss and residual mass [14,17].Previous studies confirmed that hemicellulose decreased and, consequently, the proportion of cellulose and lignin increased in the species after torrefaction [12].However, the characteristics of the torrefied biomass of tropical species have rarely been studied [12,18].Large volumes of biomass residues of tropical species in Costa Rica are constantly produced in the timber industry, so torrefaction is an option to process this raw material [12,19].The biomass of tropical timber species and the different torrefaction processes have been characterized [18][19][20][21].This study continued these studies on the biomass torrefaction of the most used species with energy potential in Costa Rica [22][23][24].
The present study aimed to evaluate the thermogravimetric behaviour, devolatilization of hemicellulose and cellulose, and the calorimetric behaviour of the torrefied biomass of five tropical woody species (Cupressus lusitanica, Dipteryx panamensis, Gmelina arborea, Tectona grandis and Vochysia ferruginea), at three temperature conditions (light, middle and severe) and three torrefaction times using simultaneous thermogravimetric and differential scanning calorimetry analyses.Then, we aimed to find the most appropriate torrefaction conditions for the different types of woody biomass using multivariate analysis of principal components (MAPC) in relation to the thermo-chemical degradation without significantly affecting the chemical composition of the material.This study will enhance the treatment of biomass to obtain renewable and viable raw material for the generation of clean energy from a lignocellulosic material [25].

Material Characteristics
The woody waste biomass of C. lusitanica, D. panamensis, G. arborea, T. grandis and V. ferruginea from fast growth plantations at different sites in Costa Rica was used.The age of the plantations ranged between 8 and 14 years.The details of the materials are available in Moya et al. [18] and Gaitán-Álvarez et al. [12,21].Sawdust from all the species was directly collected from the sawing process, conditioned to a 7% moisture content and then sieved.After sieving, the sawdust particles used were 70% of 2.00-4.00mm and 30% 0.42-2.00mm.The chemical compositions of the five species are shown in Table 1.

Torrefaction Process
Three 500 g samples of sawdust were obtained from each species.The material was then divided to apply three different torrefaction durations (8, 10 and 12 min), and the three different torrefaction temperatures (200, 225, and 250 • C), resulting in nine treatments per species.Figure 1 shows the nine treatment and their abbreviations.These durations and temperatures were selected according to a previous study [5].A modified Thermolyne Furnace 48,000 (Thermolyne, Waltham, MA, USA) was used for the torrefaction process.The furnace was sealed to prevent airflow from the manual system to maintain pressure.Every 4-5 min, the air was freed, allowing the development of the torrefaction process within an environment with limited oxygen content, adding N 2 to the furnace [19].

Torrefaction Process
Three 500 g samples of sawdust were obtained from each species.The material was then divided to apply three different torrefaction durations (8, 10 and 12 min), and the three different torrefaction temperatures (200, 225, and 250 °C), resulting in nine treatments per species.Figure 1 shows the nine treatment and their abbreviations.These durations and temperatures were selected according to a previous study [5].A modified Thermolyne Furnace 48,000 (Thermolyne, Waltham, MA, USA) was used for the torrefaction process.The furnace was sealed to prevent airflow from the manual system to maintain pressure.Every 4-5 min, the air was freed, allowing the development of the torrefaction process within an environment with limited oxygen content, adding N2 to the furnace [19].

Thermogravimetric Analysis (TGA)
To obtain the degradation curves, TGA was performed at atmospheric pressure under inert ambient nitrogen, using about 5 mg of sawdust of each species.The heating rate was 20 °C/min in a nitrogen atmosphere of ultra-high purity N2 at 100 mL/min, reaching a temperature of 800 °C.A TA Instruments (New Castle, DE, USA) thermogravimetric analyzer, model SDT Q600, was used.The TGA provided values for mass loss in relation to temperature, from which the derivative thermogravimetry (DTG) was obtained, allowing us to determine the position and temperature at which sample degradation occurred.The TGA data and their first and second derivatives (DTG and D 2 TG) were analyzed using TA Instruments Universal Analysis 2000 software.The parameters are presented in Figure 2a,b: (i) the temperature at the beginning of degradation (Ti) and the percentage of residual mass at Ti (WTi); (ii) the temperature corresponding to the maximum degradation of hemicellulose (Tsh) and the percentage of residual mass at Tsh (WTsh); (iii) the temperature corresponding to the maximum cellulose mass loss rate (Tm) and the percentage of the residual mass at Tm (WTm) and (iv) the temperature corresponding to the end of degradation (Tf) and the percentage of residual mass at Tf (WTf), when mass loss began to stabilize as the temperature increased.Additional parameters were obtained from the derivative thermogravimetry (DTG): (v) the temperature of hemicellulose degradation onset (Tonset(hc)) and residual mass at Tonset(hc) (WTonset (hc)); (vi) the end temperature of the hemicellulose degradation (Toffset(hc)) and the residual mass at Toffset(hc)

Thermogravimetric Analysis (TGA)
To obtain the degradation curves, TGA was performed at atmospheric pressure under inert ambient nitrogen, using about 5 mg of sawdust of each species.The heating rate was 20 • C/min in a nitrogen atmosphere of ultra-high purity N 2 at 100 mL/min, reaching a temperature of 800 • C. A TA Instruments (New Castle, DE, USA) thermogravimetric analyzer, model SDT Q600, was used.The TGA provided values for mass loss in relation to temperature, from which the derivative thermogravimetry (DTG) was obtained, allowing us to determine the position and temperature at which sample degradation occurred.The TGA data and their first and second derivatives (DTG and D 2 TG) were analyzed using TA Instruments Universal Analysis 2000 software.The parameters are presented in Figure 2a,b: (i) the temperature at the beginning of degradation (T i ) and the percentage of residual mass at Ti (W Ti ); (ii) the temperature corresponding to the maximum degradation of hemicellulose (T sh ) and the percentage of residual mass at T sh (WT sh ); (iii) the temperature corresponding to the maximum cellulose mass loss rate (T m ) and the percentage of the residual mass at Tm (W Tm ) and (iv)  Once the decomposition start points for hemicellulose and cellulose were obtained, the thermostability of these components was evaluated using the model described in Equation (2), which was obtained from the linearized model in Equation (1) according to Sbirrazzuoli et al. [26].The differential was the conversional method used by Friedman.The objective was to calculate the activation energy of the decomposition for each component of the materials being studied (hemicellulose and cellulose): * (1) where α is the degraded mass, is the percentage of the degraded sample per unit time, A is the pre-exponential factor, Ea is the energy of activation, and T is temperature.Once the decomposition start points for hemicellulose and cellulose were obtained, the thermostability of these components was evaluated using the model described in Equation (2), which was obtained from the linearized model in Equation (1) according to Sbirrazzuoli et al. [26].The differential was the conversional method used by Friedman.The objective was to calculate the activation energy of the decomposition for each component of the materials being studied (hemicellulose and cellulose):

Devolatilization Variation
where α is the degraded mass, dα dt is the percentage of the degraded sample per unit time, A is the pre-exponential factor, Ea is the energy of activation, and T is temperature.

Devolatilization Variation
Several methods can be used to measure the degree of biomass devolatilization [18,27].According to Grønli et al. [27], the total volatiles released during devolatilizatiion include mass fractions, whose dynamics are described by first-order kinetics.In this research, the devolatilization behavior during the thermal degradation of the biomass components in different samples was evidence by the percentage of devolatilized mass relative to time, and a subsequent analysis of the devolatilization rate (D rate ).The D rate behavior with different T T and t T was first described.Next, we determined the maximum devolatilization rate (D max ) and the time at which D rate was obtained.Figure 2d shows where these parameters were determined in the first time derivatives of the mass fraction with respect to time.

Statistical Analysis
The experiment had a two-level factorial design.Level one corresponded to the t T of the biomass at three different times: 8, 10, and 12 min.The second factorial level was T T at three temperatures: 200, 225, and 250 • C.This design was applied to each species studied (C.lusitanica, D. panamenisis, G. arborea, T. grandis and V. ferruginea).We worked with three samples for each treatment per species.Secondly, a multivariate analysis of the principal components (MAPC) was performed, including all the variables of the TGA and the determined devolatilization parameters.Two main components were selected.This analysis was performed with SAS software (SAS Inc., Cary, NC, USA).Significance level was established at 5%.

TGA-DTG Analysis
The thermogravimetric decomposition behavior of the torrefied biomass for the five species showed the same pattern with different T T and t T (Figure 3a-h).However, the DTG curve showed some differences in biomass decomposition (Figure 3a-h).For the TGA curve, five important stages were observed.A predominant signal appeared in the first stage prior to 100 • C. The second stage showed a pronounced peak between 290 • C and 330 • C, the third stage occurred between 340-380 • C.And the fourth stage appeared between 400-500 • C, where the speed of mass loss mas lower compared with the two previous decomposition stages.Finally, few changes in the sample occurred as temperature continued to increase.
Overall, the TGA and DTG curves (Figure 3a-h) showed small visually noticeable differences in the thermal behaviour of the torrefied biomass at various T T .For all species studied, the biomass torrefied at 250 • C at the three t T were thermally different.First, the TGA curves show that the biomass torrefied at 250 • C, or severe torrefaction, behaved differently compared to the rest of the T T .After 340-380 • C, mass loss was less than for biomass torrefied at 250 • C (Figure 3a-h).Second, the DTG curves showed a strong signal at 290 • C, but it appeared as a small shoulder along that at 250 • C (Figure 3a-h).This signal was more visible in D. panamensis (Figure 3c) and V. ferruginea (Figure 3i), whereas this shoulder was not present in C. lusitanica (Figure 3a), G. arborea (Figure 3e), and T. grandis (Figure 3g) in biomass torrefied at 250-12.untorrefied biomass (Table 2).Tonset(hc) was lower in the torrefied biomass of D. panamensis, G. arborea, T. grandis, and V. ferruginea compared to untorrefied biomass.The Tsh and Toffset(hc) of the torrefied biomass of these species were higher than the untorrefied biomass (Table 2).Tables 2 and 3 show the detailed analyses of the temperatures and mass loss for the various species, where the main changes in the degradation of the different chemical components of the torrefied biomass during the TGA occurred.In the evaluation of the decomposition T i in C. lusitanica, the T i of biomass torrefaction increased with respect to untorrefied biomass, except for the 250-8 condition.For the remaining four species under all torrefaction conditions, decomposition T i increased (Table 2).As for W Ti at 250 • C, torrefaction was greater in the torrefied biomass compared with untorrefied biomass for all species (Table 3).Conversely, T i tended to increase in torrefied biomass under the light torrefaction condition (200-8) to the middle torrefaction condition between 225-10 and 225-12, depending on the species, and decreased under 250-10 or 250-12 conditions.For D. panamensis, G. arborea, T. grandis, and V. ferruginea, the decomposition T f was lower for the torrefied biomass than the untorrefied biomass under any condition of T T and t T .T f increased at 200 • C in the torrefied biomass of C. lusitanica compared to untorrefied biomass.The remaining conditions (225 and 250 • C) displayed lower T f compared to the untorrefied biomass (Table 2).The behavior of W Ti , W Tf , and the residual mass differed among T i and T f conditions for all species, as W Ti and W Tf increased with increasing T T and t T (Tables 2 and 3).
C. lusitanica behaved differently with respect to hemicellulose parameters compared with the other species.T onset(hc) and T sh were higher in the torrefied biomass compared to the untorrefied biomass (Table 2), whereas the T offset(hc) was lower in all torrefied biomasses compared to the untorrefied biomass (Table 2).T onset(hc) was lower in the torrefied biomass of D. panamensis, G. arborea, T. grandis, and V. ferruginea compared to untorrefied biomass.The T sh and T offset(hc) of the torrefied biomass of these species were higher than the untorrefied biomass (Table 2).
The different torrefaction conditions had varying effects on the hemicellulose of C. lusitanica compared to the other four species.The T onset(hc) of the torrefied biomass of C. lusitanica increased as T T and t T increased, whereas T sh and T offset(hc) decreased with increasing T T and t T .The T onset(hc) also increased in the torrefied biomass of the remaining species (D. panemensis, G. arborea, T. grandis and V. feruginea) under light 200-8) to medium (between 225-10 or 225-12 depending on the species), torrefaction conditions, then decreasing under the 250-10 or 250-12 conditions.For the T sh and T offset(hc) parameters, their values decreased with increasing T T and t T , whereas some irregularities were observed in this behaviour in C. lusitanica and T. grandis.
In biomass torrefied at any T T or t T , W Tonset(hc), W Tsh , and W Toffset(hc) had higher values than in untorrefied biomass in all species (Table 3).The values of T T and t T varied under different torrefaction conditions.In general, the values of W Tonset(hc), W Tsh and W Toffset(hc) for all species increased with increasing T T , particularly in biomass torrefied at 250 • C. Few changes were observed in t T at the same T T in the W Tonset(hc) and W Tsh values for all t T of the different species.However, for W Toffset(hc), for 8 and 10 min, the parameter values were similar, whereas under condition 250-12, a significant increase in W Toffset(hc) was observed in all species (Table 3).
For the cellulose decomposition parameters, biomass torrefaction increased T onset(c) compared to the untorrefied biomass in C. lusitanica, G. arborea, T. grandis, and V. ferruginea, whereas in D. panamensis, T onset(c) increased from the 200-8 to the 250-10 condition, and then decreased under the most severe condition (250-12) (Table 3).Conversely, T m increased in the torrefaction of the biomass of C. lusitanica and D. panamensis from the least severe condition (200-8) to condition 225-10.Also, under conditions 225-12 and T T at 250 • C, the torrefied biomass had a lower T m than the untorrefied biomass.The torrefied biomass of G. arborea had a higher T m value compared to the untorrefied biomass, except under condition 250-8.In the biomass of T. grandis, torrefaction increased T m compared to untorrefied biomass, except for condition 250-12.In the biomass of V. ferruginea, torrefaction reduced T m under conditions 200-12, 225-8, and 225-10, whereas T m was higher under the rest of the torrefaction conditions (Table 3) compared with untorrefied biomass.Lastly, torrefaction of the biomass of the five species decreased T offset(c) compared with untorrefied biomass (Table 3).With respect to the different torrefaction conditions, the increase in T T and t T decreased T onset(c) in C. lusitanica and D. panamensis.Conversely, in G. arborea, the increase in T T and t T decreased T onset(c), except under condition 225-8.As for T. grandis and V. ferruginea, no trend was found in T onset(c) with either an increase or decrease of T T or t T (Table 3).T m and T offset(c) decreased in all species as T T or t T increased (Table 3).
The evaluation of the residual mass of the different biomasses showed that torrefaction decreased the W Tonset(c) value in C. lusitanica and T. grandis, whereas in D. panamensis, G. arborea and V. ferruginea, torrefaction decreased the W Tonset(c) value, except under the most severe condition (250-12) (Table 3).Torrefaction increased T m in all species, except under condition 200-8 for D. panamensis and T. grandis and conditions 200-8 and 200-10 for V. ferruginea, where T m decreased.Lastly, W Toffset(c) was higher in torrefied biomass than in untorrefied biomass for all species (Table 3).The evaluation of the residual mass under the different torrefaction conditions showed that all parameters related to residual mass (W Tonset(c), W Tm , and W Toffset(c) ) increased their values with increasing T T and t T of torrefaction in all species (Table 3).
Table 4 shows the kinetic parameters of hemicellulose and cellulose decomposition in torrefied biomass observed with TGA.In the torrefied biomass of C. lusitanica, the activation energy (Ea) value of hemicellulose increased with the increase in T T and t T up to 225 • C; however, in torrefaction at 250 • C for 10 and 12 min, the Ea values were lower.The Ea cellulose values for the biomass of C. lusitanica increased with the increase in T T and t T , requiring more energy to degrade the cellulose in the biomass.For Ea, the torrefied biomasses at 200 • C and 225-8 had lower values than the untorrefied biomass; then Ea increased with T T and t T , decreasing again under the 250-12 condition.
The torrefaction increased the pre-exponential factor (A) and Ea of the hemicellulose in the D. panamensis biomass compared to the untorrefied biomass.The A and Ea increased from 200-8 to 225-12, and decreased at 250 • C. For cellulose, the A and Ea in the torrefied biomass decreased with the increase in T T and t T , but any torrefaction produced lower values of A and Ea compared to the untorrefied biomass.
The G. arborea torrefied biomass had lower A and Ea hemicellulose values than the untorrefied biomass.Torrefaction at 200 • C had high Ea hemicellulose values, whereas the Ea decreased significantly with T T and t T above 225 • C. For cellulose, Ea increased as T T and t T increased, especially under condition 225-10 (Table 4).The torrefied biomass had higher Ea values compared to untorrefied biomass, except for conditions 200-8 and 200-10.In general, torrefaction Ea increased with increasing T T and t T .
With T. grandis, the hemicellulose A and Ea increased up to torrefaction condition 225-12, with values greater than those for untorrefied biomass.Beyond these torrefaction conditions, A and Ea decreased in the torrefied biomass at 250 • C, with values lower than found for the untorrefied biomass.For cellulose, Ea was lower for the different types of torrefied biomass, except under condition 250-12.Ea was also lower in torrefied biomass under 225-8, 225-10, and 250-8 conditions.Ea increased with T T for the other temperatures.However, Ea was greater for all torrefied biomasses, increasing as T T and t T increased (Table 4).
The torrefied biomass of V. ferruginea had lower Ea values for hemicellulose than the untorrefied biomass.Ea in hemicellulose decreased with decreasing T T for the different torrefaction conditions, whereas at the same temperature, Ea decreased at 10 and 12 min.For cellulose, the Ea value was lower as T T, decreased except for conditions 225-10 and 250-12, in which Ea was higher in the torrefied biomass.For the EA value for cellulose for different T T , Ea increased with T T , excluding conditions 200-10 and 225-8, which had a low Ea value.
Notably, the correlation coefficients (R 2 ) for all torrefaction conditions remained close to 0.99, with the exception of the cellulose models for G. arborea, which were low.

Devolatilization
Figure 4 displays the devolatilization rate of the torrefied and untorrefied biomasses.Table 5 shows when D max was reached and the D max values.For the C. lusitanica biomass, torrefaction at 250-10 and 250-12 had lower D rate values (Figure 5a) and reached D max more quickly (Table 5), whereas D max increased between 200-8 and 225-12, and then decreased in biomass torrefied at 250 • C.
The D. panamensis biomass torrefied at 250 • C at the three temperatures had the lowest devolatilization rate.In addition, the shoulder in the devolatilization curve at 13 min disappeared in the biomass torrefied at 250 • C (Figure 4b).The time to reach D max showed no significant variation, except for condition 250-12 where the time required was shorter and D max increased with T T and t T , except for condition 250-12, where again the value was low (Table 5).G. arborea had the lowest devolatilization rate for all torrefactions, and especially for 225-10, 225-12, and 250 • C in the three t T (Figure 4c).The time to reach D max was approximately 16 min in the different types of biomass.However, the shortest time was obtained with 250-10 (Table 5).The D max value increased at 225 • C, but decreased at 250 • C. The T. grandis torrefied biomass had a D rate above 20 dw/dt, whereas torrefactions at 250-10 and 250-12 displayed no inflexion at 13 min (Figure 3d).The maximum devolatilization was reached at 17 min for the untorrefied biomass and 200 • C and 250-12 for torrefied biomass.Under other torrefaction conditions, the time exceeded 18 min.D max increased with increasing T T and t T of torrefaction, with the exception of 250-12, which had a low D max value (Table 5).
Under conditions 200-8, 200-10, and 225-8, the torrefied V. ferruginea had a lower D rate relative to the torrefied and untorrefied biomass under the other conditions (Figure 4e).The time to reach D max was close to 17 min, with a slight increase in the value of D max with increasing T T and t T of torrefaction (Table 5).
Notably, in all species, once D max was reached, the slope of the curve became more severe, with a steeper slope (Figure 4a-e).The T. grandis torrefied biomass had a Drate above 20 dw/dt, whereas torrefactions at 250-10 and 250-12 displayed no inflexion at 13 min (Figure 3d).The maximum devolatilization was reached at 17 min for the untorrefied biomass and 200 °C and 250-12 for torrefied biomass.Under other  Notably, in all species, once Dmax was reached, the slope of the curve became more severe, with a steeper slope (Figure 4a-e).

Differential Scanning Calorimetry Analyses
Figure 5 displays the DTG curves of the calorimetric analysis of the reactions that occurred during the TGA.In all torrefaction conditions and for all woody species, the first endothermic peaks occurred at 100 and 300 °C, whereas exothermic peaks were observed between 350 and 400 °C, with some variations among the species and torrefaction conditions.All torrefaction conditions demonstrated endothermic processes in the C. lusitanica biomass.However, untorrefied biomass for

Differential Scanning Calorimetry Analyses
Figure 5 displays the DTG curves of the calorimetric analysis of the reactions that occurred during the TGA.In all torrefaction conditions and for all woody species, the first endothermic peaks occurred at 100 and 300 • C, whereas exothermic peaks were observed between 350 and 400 • C, with some variations among the species and torrefaction conditions.All torrefaction conditions demonstrated endothermic processes in the C. lusitanica biomass.However, untorrefied biomass for condition 200-8 had a more pronounced exothermic peak between 350 and 450 • C (Figure 5a).In the D. panamensis biomass, torrefaction at 200-8, 200-12, 225-8, and 250-12 demonstrated exothermic processes between 350 and 450 • C, whereas endothermic peaks of greater magnitude were observed in biomass torrefied at 225-12, 250-10, and 250-12 (Figure 5c).Exothermic reactions occurred between 300 and 400 • C in G. arborea untorrefied biomass and with torrefaction at 200-8 and 225-10, whereas the biomass under the other torrefaction conditions only showed endothermic reactions (Figure 4c).For the T. grandis biomass, the exothermic reactions between 350 and 450 • C appeared in untorrefied biomass and under conditions 200-8 and 225-10 (Figure 5d).The opposite occurred in the biomass of V. ferruginea, as torrefied biomass presented exothermic peaks under severe torrefaction conditions (225-10, 250-10, and 250-12) (Figure 6e).

TGA-DTG Analysis
In general, TGA trends for torrefied and untorrefied biomass of the different woody species were similar, which is consistent with previous reports [18,20,21].The DTG curves support this finding, where important stages were defined (Figure 3a-e) and differences were clarified.
During thermogravimetric analyses, the initial decrease in mass is attributed to the release of the moisture in the samples [28].This water release is lower for biomasses torrefied under more severe conditions, consistent with a higher drying temperature and an increase in the hydrophobicity related to such conditions [12].Higher temperatures enable the decomposition of the polymers present in the samples.Hemicellulose degradation occurs between 230 and 330 • C [29].This degradation mainly tends to disappear under severe torrefaction conditions for all five species because a higher percentage of hemicellulose has already been eliminated during torrefaction [4,28,30,31].Then, cellulose decomposition occurs at temperatures between 305 and 380 • C [28,32], which appears in all the biomasses analyzed considering that torrefaction processes affect this biopolymer less than hemicellulose.Temperatures between 400 and 500 • C cause the final decomposition of cellulose and most of the lignin [33].During this stage, the decomposition rate slows and then continues to a period of limited change as temperature increases.
The torrefied biomass displays the four decomposition stages of the well-defined components (Figure 3a-e).The first signal in the DTG curves before 150 • C is attributable to the removal of moisture in the samples, since moisture decreases with T T [12].The next signal or decomposition stage between 230 and 330 • C is due to hemicellulose degradation [29]; however, contrary to the untorrefied biomass, this signal tends to disappear under severe torrefaction in all five species (Figure 3).This is because high percentages of hemicelluloses have already been eliminated in the process prior to torrefaction [28,30].This result agrees with the work reported by Bach et al. [31] and Ren et al. [4], who torrefied the biomass of conifers under temperatures above 250 • C and found that the signal decreased in the TGA curve.The next peak in the curve is related to cellulose decomposition, which occurs in the range of 305 to 380 • C [28,32].This curve occurs in all torrefaction conditions and in untorrefied biomass, with differences in the magnitude of the peak, evidenced by weight loss.Lastly, in the final stage between 400 and 500 • C, the rate of decomposition slows, which is attributable to the final decomposition of cellulose and most of the lignin [33].
Using the parameters for material degradation (T i , W i ) and hemicellulose degration (T onset(hc) , W Tonset (hc ), T sh , T offset(hc ), and W Toffset(hc )), the evaluation of stages two and three of the TGA curve shows that an increase of T T and t T increase T i (Tables 2 and 3), indicating that torrefied biomass is more thermally stable than untorrefied biomass, which agrees with results found by Lee et al. [34] and Islam et al. [35] when evaluation some tropical species (Dyera costulata, Esdospermun diadenum, Paraserianthes moluccana, Hevea brasiliensis, and Alstonia pneumatophora).This result also indicates that an increase in T T and t T stabilizes the biomass, leading to a reduction in the mass loss values (W Tonset(hc) and W Toffset(hc) ) of this component (Table 4).Nevertheless, this behaviour should be viewed cautiously, as some authors indicated that this relationship is the result of the content of extractives in the wood, and the volatile material [27], affecting the combustion process [36].In fact, Gaitán-Alvarez et al. [12] showed that, with these same woody species, weight loss during torrefaction is correlated with the type and content of extractives.
As for the cellulose degradation parameters, the evaluation showed that the most important differences were in temperature and residual mass at different T T and t T (Tables 2 and 3).The temperature parameters (T onset(c) , T m , and T offset(c) ) increased as T T and t T increased, except in G. arborea (Table 2), again indicating that this biomass component has higher thermal stability than in torrefied biomass, causing a reduction in weight loss values in the different stages of the evaluated cellulose decomposition evaluated (W Tonset(c), W Tm , and W Toffset(c) ).Some of the differences found in the species, or in the behaviour of the parameters evaluated in the decomposition of cellulose in torrefied biomass between species (Tables 2 and 3), show that the decomposition process of cellulose is complex in both torrefied and untorrefied biomass.The thermal stability of cellulose is related to the natural variation in the material, having a more chemically complex structure than hemicellulose [32,37].Therefore, the parameters evaluated in the five species differ among them, even under the torrefaction conditions (Tables 3 and 4).
The activation energy was expected to increase with increasing T T [38].However, each structural component of the biomass has its own behaviour due to its chemical nature [39].In the first stage, early degradation of hemicellulose appears during the torrefaction process [39].Here, low activation energy is required to initiate hemicellulose degradation compared to activation energy of cellulose [38].Hemicellulose starts to decompose at low temperatures, between 180 and 350 • C [40].Ramos [41] indicated that xylan, a type of hemicellulose, depolymerizes and reduces hemicelluloses into smaller molecules with lower molecular weight, which are more sensitive to pyrolysis [42].Thus, with an increase in the torrefaction conditions in T T and t T , the hemicellulose decomposes in monosaccharides and volatilizes more rapidly [42].For this reason, torrefied biomass has a low percentage of hemicellulose.Within the same woody species, Gaitán-Alvarez et al. [12] found that the weight loss of hemicellulose increases with T T and t T .Then, given the low hemicellulose content in the biomass torrefied under severe conditions, the activation energy is lower as T T and t T increase (Table 3).This result coincides with the studies of Bach et al. [43] on Norway spruce, Bobleter [44] on plants, and Wyman et al. [45] on biomass.These authors found that an increase in T T significantly decreases the Ea in hemicellulose.
Cellulose degradation requires higher energy [38].Biomass torrefaction increases the Ea value for cellulose (Table 4), since the thermal process increases the order of the regions of cellulose [40].This means that heat transportation is more difficult [46], so the thermal stability of the biomass is greater [38].This behaviour was observed in the biomasses studied, where the Ea for cellulose increased with increasing T T and t T , particularly under severe torrefaction conditions (Table 4).

Devolatilization
D max is associated with the activation energy of cellulose decomposition.Higher Ea makes the polymer decomposition process more difficult, which is reflected in the lower D max values and vice versa (Tables 4 and 5).Likewise, the decrease in devolatilization rates at higher temperatures at 250-10 and 250-12 (Figure 4a-e, Table 5) is attributed to the fact that at these T T , a high proportion of hemicellulose has been degraded [27,46], leaving a low percentage of hemicellulose and less cellulose to devolatilize when the biomass is used for energy production.Likewise, a reduction in the devolatilization rate at the higher temperatures of 250-10 and 250-12 (Figure 4a-e, Table 5) is attributed to the degradation of a high proportion of hemicellulose at these T T [27,46], leaving a low percentage of hemicellulose and less cellulose for devolatilization when the biomass is used for energy production.
The differences found in the devolatilization and D max values among the various species (Figure 4a-e, Table 5) are associated with the proportion and nature of the hemicellulose and cellulose contained in the biomass, since each species has its unique behaviour and chemical structure, and therefore its own pyrolysis characteristics [18].
Chen et al. [5] showed that an increase in T T and t T affects D max , without affecting the time to reach maximum devolatilization, with differences of approximately 2 min (Table 5).This behaviour indicates that, in torrefied biomass, the decomposition of cellulose (the component associated with maximum devolatilization) and the time to reach maximum devolatilization are maintained, whereas thermal stability of the torrefied biomass causes values of D max to vary.

Differential Scanning Calorimetry Analyses
At temperatures below 200 • C, all DSC curves of the five species show endothermic values, which is linked to the energy biomass needs to absorb to remove moisture [32].Later, the exothermic peaks at 275 • C correspond to degradation of hemicellulose, while yhe peak at 365 • C corresponds to ligninn [32,47].The endothermic peak close to 355 • C corresponds to degradation of cellulose [32].Figure 4a-e clearly shows the processes previously described.
Although all torrefied biomasses of the different species display the exothermic processes of hemicellulose and lignin and the endothermic process of cellulose [48,49], the different behavior of each species with respect to torrefaction conditions are evident.For the torrefied biomass of C. lusitanica (Figure 4a), the endothermic peaks at 375 • C are more pronounced than for the other species, indicating the greater stability of the cellulose in this species [38].Conversely, in the torrefied biomass of D. panamensis and T. grandis (Figure 5b,d), the endothermic peaks at 375 • C are small or less pronounced, occurring mainly in the torrefaction conditions above 225-12, meaning that under these torrefaction conditions, cellulose is less stable [38].
The exothermic peaks at 275 • C corresponding to hemicellulose [48,49], are less pronounced or absent in some torrefied biomasses (Figure 5a-e), and especially in the biomass of G. arborea under all torrefaction conditions (Figure 5c).In the remainder of the species, this behaviour mainly appears under torrefaction conditions above 225-10 (Figure 5a,b,d,e).This is because at those T T , part of the hemicellulose was removed during the torrefaction process [5].Therefore, the exothermic peak with severe torrefaction is unclear [35].

Multivariate Analysis
The variables related to hemicellulose (T onset(hc) , T offset(hc) , W Tonset(hc) , W Toffset(hc) , T sh , and W Tsh ) form PC1, whereas PC2 is related to the cellulose parameters (Table 6).This demonstrates that the behaviour of torrefied biomass at different T T and t T can be classified relative to the content of these components.In addition, these two components reflect the thermal stability of the torrefied biomass, as these were statistically and significantly reflected in the principal components (Table 3).However, the relationships between the principal components and the hemicellulose or cellulose parameters may not always be significant under some torrefied biomass conditions, likely due to the nature and quantity of these components in the biomass [18,21].
The scores of the components of the different types of biomass under torrefaction conditions display the different T T and t T conditions of hemicellulose and cellulose (Figure 6a,e, respectively).C. lusitanica, D. panamensis, and T. grandis, likely due to greater thermal stability under severe torrefaction conditions (250-10 and 250-12), form a unique group, different from the group formed with biomasses torrefied under light and middle torrefaction, which have similar conditions amongst the two groups.The torrefied biomass of G. arborea and V. ferruginea of form a group with biomasses torrefied under these conditions, different from biomass torrefied at 200 and 225 • C probably due to lesser thermal stability under condition 225-10 [5,18,21].Then, the group formed by the different types of biomass torrefied under light and middle conditions, at 200 and 225 • C, respectively, indicate that these are the appropriate torrefaction conditions for those species, since they have the most appropriate parameters for combustion, such as positive correlation with D max thermal stability (Table 6).

Conclusions
Based on our results, we conclude that the best torrefaction temperatures and times for the tested species are 200 • C for 8, 10, and 12 min and 225 • C for 8, 10, and 12 min, classified as light and medium torrefaction.Under these conditions, optimum thermo-chemical degradation is achieved for using biomass as an energy source, without significantly affecting the chemical composition of the material.In all species, severe torrefaction at 250 • C produced important degradation of the material, especially hemicellulose and part of the cellulose.As such, we do not recommend the use of this temperature in the biomass torrefaction of tropical species.However, behaviour among species presents some differences C. lusitanica, D. panamensis and T. grandis showed higher thermal stability that G. arborea and V. ferruginea.

Figure 1 .
Figure 1.Temperature and time for the torrefaction of the biomass of five fast-growth plantation species of Costa Rica.Note: the numbers in parentheses indicate the abbreviation of this torrefaction condition.

Figure 1 .
Figure 1.Temperature and time for the torrefaction of the biomass of five fast-growth plantation species of Costa Rica.Note: the numbers in parentheses indicate the abbreviation of this torrefaction condition.
to the end of degradation (T f ) and the percentage of residual mass at T f (W Tf ), when mass loss began to stabilize as the temperature increased.Additional parameters were obtained from the derivative thermogravimetry (DTG): (v) the temperature of hemicellulose degradation onset (T onset(hc) ) and residual mass at T onset(hc) (W Tonset (hc) ); (vi) the end temperature of the hemicellulose degradation (T offset(hc) ) and the residual mass at T offset(hc) (W Toffset (hc) ); (vii): the temperature of cellulose degradation onset (T onset(c) ) and the residual mass at T onset(c) (W Tonset(c) ); (viii): the end temperature of cellulose degradation (T offset(c) )and the residual mass at T offset(c) (W Toffset(c) ). Figure 2c shows the DTG curve representing the different temperature.MAgicPlot 2.5.1 software was used to obtain these values.Energies 2018, 11, x 4 of 25

Figure 2 .
Figure 2. (a,c) Derivative thermogravimetry (DTG) and (b) second derivative (D 2 TG) parameters for the different woody biomasses analyzed; (d) Devolatilization rate measured by first time derivates of the mass frcationas a function of time.Note: tbd is the start time of the maximum devolatilization rate and Dmax is the maximum devolatilization rate [18].

Figure 2 .
Figure 2. (a,c) Derivative thermogravimetry (DTG) and (b) second derivative (D 2 TG) parameters for the different woody biomasses analyzed; (d) Devolatilization rate measured by first time derivates of the mass frcationas a function of time.Note: t bd is the start time of the maximum devolatilization rate and D max is the maximum devolatilization rate [18].

Figure 4 .
Figure 4. Devolatilization rate measured by the first derivative of the mass fraction with respect to time for the biomasses of for Cupressus lusitanica (a), Dipteryx panamensis (b), Gmelina arborea (c) and Tectona grandis (d) and Vochysia guatemalensis (e), torrefied at different times and temperatures.

Figure 4 .
Figure 4. Devolatilization rate measured by the first derivative of the mass fraction with respect to time for the biomasses of for Cupressus lusitanica (a), Dipteryx panamensis (b), Gmelina arborea (c) and Tectona grandis (d) and Vochysia guatemalensis (e), torrefied at different times and temperatures.

Figure 6 .
Figure 6.Relationship between the auto-vector of components 1 and 2 of the multivariate analysis by means of principal components of for Cupressus lusitanica (a), Dipteryx panamensis (b), Gmelina arborea (c) and Tectona grandis (d) and Vochysia guatemalensis (e), torrefied at different times and temperatures of five fast-growth plantations in Costa Rica.

Table 1 .
Chemical composition of five fast-growth plantation species in Costa Rica.

Table 2 .
Thermogravimetric analysis (TGA) temperatures of biomasses of five fast-growth plantation species in Costa Rica torrefied at different times and temperatures.

Table 3 .
TGA residual masses of the biomasses of five fast-growth plantation species in Costa Rica torrefied at different times and temperatures.

Table 4 .
Activation energies and pre-exponential factors for the thermal decomposition of hemicellulose and cellulose observed in biomasses of five fast-growth plantation species in Costa Rica torrefied at different times and temperatures.

Table 5 .
Time to reach the maximum devolatilization rate determined by thermogravimetric analysis (TGA) experiments of the biomasses of five fast-growth plantation species in Costa Rica torrefied at different times and temperatures.

Table 6 .
Matrix of the multivariate analysis correlations for all variables evaluated of biomass torrefied at different times and temperatures of five fast-growth plantations species in Costa Rica.