Thermochimica Acta 691 (2020) 178714

Contents lists available at ScienceDirect

Thermochimica Acta
journal homepage: www.elsevier.com/locate/tca

Coconut fiber pyrolysis decomposition kinetics applying single- and multi- T

step reaction models
Fernanda Cristina Rezende Lopes, Katia Tannous*
School of Chemical Engineering, University of Campinas - 500 Albert Einstein Avenue, Campinas, SP, 13083-852, Brazil

ARTICLE INFO ABSTRACT

Keywords: This work aims to evaluate the thermal decomposition kinetics of coconut fiber in a thermogravimetric analyzer
Biomass using four heating rates (5−20 °C/min) in nitrogen atmosphere. Three decomposition stages were identified
Kinetic (dehydration, pyrolysis, and carbonization) and the pyrolysis kinetic was performed considering two reaction
Modeling models (single- and multi-step). In the first model, the apparent activation energy (94.5–210.8 kJ/mol) was
Pyrolysis
determined using five isoconversional methods, and the reaction model representative was the three-dimen-
Thermogravimetry
sional diffusion model, f(α) = (3/2) (1-α)2/3 [1 - (1-α)1/3]−1, obtained using the master plots method. Through
the linearization method of the conversion rate equation, the global activation energy found was 129.8 kJ/mol
and the pre-exponential factor (log A) was 18.8 1/s. Due to the complexity of the biomass decomposition, the
second model, three independent parallel reactions scheme (IPRS), was applied to describe each pseudo-com-
ponent (hemicellulose, cellulose, and lignin). It was possible to obtain activation energies from 79.1–196.3 kJ/
mol, pre-exponential factors (log A) from 3.1–15.0 1/s, volatilized fractions from 0.3 to 0.5, considering 1st and
2nd reaction orders. Finally, the modeling of conversions and conversion rates of both approaches was in a good
agreement with the experimental data, however, the second model proved accuracy in describing the coconut
fiber pyrolysis.

1. Introduction such as floor mats, doormats, brushes, mattresses, coarse filling mate-
rial, and upholstery [6]. Another use of coconut fiber to solve problems
Energy is the driving force for the development of the world caused by waste could be energy and fuel productions via the thermal
economy. With population growth and economic development, there is conversion of its biomass [7–12].
an increase in energy demand, making it necessary to look for alter- There are several thermochemical processes for converting biomass
native sources to non-renewable resources [1]. Biomass waste that into energy and fuels, such as liquefaction, direct combustion, gasifi-
particularly derived from the food and agribusiness sectors can be cation, and pyrolysis [13]. The pyrolysis is one of the most important
considered a promising renewable energy resource [2]. According to pathways for biomass conversion [14], in which biomass is heated
the Brazilian Energy Balance (Summary Report/year 2018), 45.3 % of under inert conditions [15,16] generating the following products: li-
the energy supply was represented by renewable sources. These re- quids (bio-oil), solids (carbon), and gases (e.g., H2, CO, CH4) [2].
newable sources were hydroelectric energy (12.6 %) and biomass, such Concerning to the biomass pyrolysis, the thermogravimetry (TG),
as sugarcane (17.4 %) and firewood (8.4 %) [3]. which measures the remained mass of material as a function of time and
Coconut residue is other biomass that could be integrated into the temperature, is the most used technique for analyzing the thermal de-
Brazilian energy matrix since that the production is 2.3 million tons per composition and its reaction kinetics [16]. Based on TG, mathematical
year [4]. Approximately 15 % of coconut is used for consumption methods with different levels of complexity have been developed to
(water and pulp), and 85 % of the material is disposed of in landfills describe the kinetics of pyrolysis through the triplet parameters, in-
[5], causing numerous problems, such as methane generation. There- cluding classical isoconversional methods (differential and integral),
fore, the utilization of coconut residue is very important to preserve the such as Friedman [17], Ozawa [18], Flynn and Wall [19], Coats and
environment from the disposal wastes. Regarding the discarded mate- Redfern [20], modified Coats-Redfern [21] and Starink [22], and ad-
rial, it is possible to take advantage of the fruit mesocarp (fiber bundle). vanced isoconversional method, such as Vyazovkin [23]. These
The use of these fibers can lead to the production of low-cost materials methods have been widely used to investigate the apparent activation

⁎
Corresponding author.
E-mail addresses: [email protected] (F.C.R. Lopes), [email protected] (K. Tannous).

https://doi.org/10.1016/j.tca.2020.178714
Received 1 April 2020; Received in revised form 2 July 2020; Accepted 7 July 2020
Available online 10 July 2020
0040-6031/ © 2020 Elsevier B.V. All rights reserved.
F.C.R. Lopes and K. Tannous Thermochimica Acta 691 (2020) 178714

energy of biomass decomposition assuming a single-step reaction standard applied. According to White et al. [33], these particles are
[11,12,24–26]. They are considered the simplest pathway for data recommended to avoid the temperature gradients and transport phe-
processing [27]. According to Vyazovkin [23], different isoconversional nomena effects. However, the study made by Rueda and Tannous [34]
methods should be applied simultaneously to guarantee more reliability for sugarcane straw with different particle diameters of 0.250 mm,
of apparent activation energy. 0.510 mm, and 1.010 mm, showed no significant difference between
Also, the literature has presented adjustment methods such as the the decomposition profiles obtained by thermogravimetry. In addition,
consecutive (CRS) and independent parallel (IPRS) reactions schemes, the studies [11,12,14,24–26,28] used a sample between 3 and 11 mg,
in which consider that biomass decomposes in multi-step reactions due the particle density and the size restriction of sample holder. Ac-
[12,14,26,28]. Regarding these methods, to facilitate the convergence cording to Vyazovkin et al. [32], a minimum sample size must be de-
of the kinetic parameters and more reliable results, the first guess of fined to ensure uniform distribution on the sample holder.
activation energy is recommended through the values obtained from Another important factor, is the use of at least three heating rates
isoconversional methods [29]. lower than 20 °C/min to obtain accuracy in the determination of the
Besides that, other models, such as deconvolution and the dis- kinetic parameters [29]. However, as can be seen in Table 1, the works
tributed activation energy (DAE) models, are found in the literature carried out by Mishra and Bhaskar [24] for rice husk (5, 10, 15, 20, 30
[30,31]. These models assume that lignocellulosic biomass decom- and 40 °C/min) and Siddiqi et al. [11,12] for coconut shell (10, 20 and
position occurs through multi-step reactions, however present greater 40 °C/min) were performed using higher heating rates.,According to
mathematical complexities and require a large computational time, White et al. [33], it can cause a temperature gradient inside the particle
when compared with CRS and IPRS. affecting, consequently, the kinetic modeling. Besides that, these lead to
Generally, regarding the practical aspect of TG in the pyrolysis the lower quality of fit in the linearizations (R2 < 0.9) obtained from
process, the literature has recommended some important procedures isoconversional methods under the apparent activation energies, as can
based on physical property (e.g., particle size) and operational condi- be noted by Siddiqi et al. [11] work. Moreover, a large standard de-
tions (e.g., sample mass, and heating rate). An inadequate application viation can be remarked for activation energy between the heating rates
of thermogravimetry can lead to errors in interpretation and calcula- in Siddiqi et al. [12] study, which applying IPRS reached values of 24
tions of the kinetic parameters [32]. Table 1 shows some researches %, 23 %, and 13 %, for hemicellulose, cellulose, and lignin, respec-
applying this technique for different biomasses closer to the coco fiber, tively. Even though there is a wide dissemination of experimental re-
case study of the present article, since that is quite rarely in the lit- sults and kinetic modeling, some gaps are still observed in the literature
erature. There is an agreement between the authors [14,24–26,28] concerning to thermogravimetric analysis for their validation related
concerning in applying of small particles size (lower than 0.51 mm), with theoretical models.
although some authors [11,12] did not present clearly the diameter Then, this study aims to characterize coconut fiber through the

Table 1
Kinetic studies considering TGA conditions, modeling, and parameters for biomass in an inert atmosphere.
Reference Biomass Diameter (mm) TG conditions Method Kinetic parameters

Eα/Ei (kJ/mol) log A n F

[11] Coconut shell Mesh < 72 5 mg FD 117 - - -

10, 20 and 40 °C/min OFW 143 - - -
room-727 °C CRM 111 - - -
ST 112 - - -
[12] Coconut shell Mesh < 72 5 mg IPRS HC: 171 HC: 11 HC: 1 HC: 0.51
10, 20 and 40 °C/min C: 149 C: 10 C: 1 C: 0.10
room-727 °C L: 93 L: 7 L: 1 L: 0.28
[14] Energy cane (Saccharum robustum Linnaeus) 0.254 10.0 mg FD 169 - - -
5, 10 and 20 °C/min OFW 167 - - -
room-900 °C VZ 168 - - -
IPRS E: 105 E: 8 E: 1.05 E: 0.13
HC: 153 HC: 11 HC: 1.31 HC: 0.19
C: 181 C: 12 C: 1.02 C: 0.43
L: 60 L: 12 L: 3.00 L: 0.25
[24] Rice straw 0.05 10 mg FD 195 - - -
5, 10, 15, 20, 30 and 40 °C/min OFW 179 - - -
room-700 °C CRM 179 - - -
VZ 180 - - -
[25] Sugarcane straw 0.51 3.0 mg FD 163 - - -
1.25, 2.5, 5.0 and 10 °C/min
room-900 °C
[26] Palm kernel shell 0.25 7.0 mg FD 185 - - -
1.25, 2.5, 5.0 and 10 °C/min VZ 184 - - -
room-900 °C IPRS HC: 107 - - HC: 0.35
C: 200 - - C: 0.37
L: 49 - - L: 0.28
Empty fruit bunch 0.25 7.0 mg FD 180 - - -
1.25, 2.5, 5.0 and 10 °C/min VZ 165 - - -
room-900 °C IPRS HC: 85 - - HC: 0.33
C:194 - - C: 0.36
L: 49 - - L: 0.31
[28] Energy cane (Saccharum spontaneum Linnaeus) 0.502 10.7 mg IPRS E: 111 E: 9.4 E: 1 E: 0.10
5, 10 and 15 °C/min HC: 119 HC: 8.6 HC: 1 HC: 0.27
room-900 °C C: 196 C: 14.0 C: 1 C: 0.33
L: 51 L:1.0 L: 3 L: 0.30

2
F.C.R. Lopes and K. Tannous Thermochimica Acta 691 (2020) 178714

chemical and thermal properties (ultimate, proximate and the higher hydrogen, nitrogen, sulfur and ash contents (from proximate analysis)
heating value analyses) following appropriate standard methods for considering 100 % of the total mass fraction. The proximate analysis
biomasses, to evaluate its thermal decomposition by thermogravimetry was obtained according to standards methods: volatile material content
(TG/DTG), and to discuss the application of two mathematical ap- (ASTM E872−82, 2019), ash content (ASTM E1755, 2015), as well as
proaches (single- and multi-step reaction models) for efficient the fixed carbon content (by difference). The higher heating value
achievement of chemical kinetics (triplet parameters). For the first (HHV) was determined using an oxygen bomb calorimeter (IKA, model
approach, four classical and one advanced isoconversional methods C200, United States), applying standard ASTM D240−2017. The lower
(Friedman, Ozawa-Flynn-Wall, modified Coats-Redfern, Starink and heating value (LHV) was determined considering the HHV, based on the
Vyazovkin) were applied, followed by master plots and linearization enthalpy of vaporization of water at 25 °C and 1 atm (2440 kJ/kg), and
methods. The application of different isoconversional methods gen- the hydrogen content [36].
erates more accuracy and reliability of activation energy results. For the As observed in Table 2, the ultimate analysis showed that carbon
second approach, the independent parallel reactions scheme was per- (46.54 ± 0.21 %) and oxygen (43.13 ± 0.01 %) contents were more
fomed taking into account the decomposition of each pseudo-compo- prevalent in coconut fiber. Furthermore, H/C and O/C atomic ratios
nent. Both models were evaluated through the conversion and con- were 1.76 and 0.70, respectively. These ratios are important for un-
version rates. The sequential application of these models is very derstanding and predicting the thermochemical conversion [37]. The
important, since the single-step model provides overall knowledge of nitrogen (0.36 ± 0.02 %) and sulfur contents (0.19 ± 0.07 %) were
the biomass decomposition as the presence or not of multiple stages and significantly lower comparable to fossil fuel. These levels suggest the
establishes the initial guesses for more complex methods promoting lowest emissions of NOx and SOx for atmosphere [38]. In addition, the
reliable results. Final results are relevant for the design of pyrolysis proximate analysis showed a significant amount of volatile matter
reactor. (79.61 ± 2.09 %), which facilitates their decomposition. Also, it pre-
sented low ash content (2.90 ± 0.36 %) comparable to rice straw,
19.74 % [24] and empty fruit bunch, 8.19 % [26]. It is desirable, be-
2. Material and methods cause this component can cause ignition problems, equipment in-
crustations, and affect the decomposition rate, reflecting in the increase
2.1. Material of carbonaceous material at the end of the thermochemical process
[39,40]. Besides that, the results obtained demonstrate that coconut
The material selected for the present study was coconut fiber (Cocos fiber has potential for energy generation.
nucifera). Coconut shells were collected in the leisure areas of The ultimate, proximate and the higher heating value analyses of
Campinas/SP, and defibrated; then, the fibers were ground in an elec- the sample are the most important analyses for recognizing the quality
tric knife mill coupled to a 1720 rpm induction motor, CV3 (Eberle, of biomass for use as feedstock in thermal conversion processes [41],
model B90 L4, Brazil) and sieving (Granutest and Bertel, Brazil). Due to but it is not directly linked to the kinetic decomposition of pyrolysis and
the elongated cylindrical shape of the sample, a low contact area with its parameters.
the sample holder surface (ASTM E1269, 2018) was observed. Thus, it
was necessary to cut the fiber manually, and one hundred particles were
measured through the APOGEO software (version 1.0) [35] corre- 2.2. Thermogravimetric analysis
spondent to 0.26 ± 0.09 mm in diameter and 3.04 ± 0.85 mm in
length. This diameter was chosen because it is small enough to avoid The experiments were performed on a TGA/DSC 1 equipment
the effects of heat and mass transfer between particles in thermo- (Mettler-Toledo, Switzerland, precision of temperature
gravimetric tests. measurement ± 0.15 °C and microbalance sensitivity ± 0.1 μg), with
The samples, in duplicate, of 0.7216 ± 0.0047 g were dried using controlled temperature and records of mass variation over time (1 s
petri dishes in a forced air oven (Quimis, model Q314M-242, Brazil) at intervals). The samples (10.36 ± 0.42 mg) were placed in a 70 μl
103.0 ± 1.0 °C for 16 h, plus an additional six intercalated hours (22 h) platinum sample holder with perforated lid and subjected to 10 min
until the final mass variation was less than 0.2 % (ASTM E871−82, isotherms at the initial (25.18 ± 0.17 °C) and final temperature
2019). (699.87 ± 0.11 °C), and four heating rates (β) of 5, 10, 15 and 20 °C/
Table 2 presents the chemical and thermal properties of coconut min, following the recommendation of the International Confederation
fiber on a dry basis. The ultimate analysis was performed on a CHNS for Thermal Analysis and Calorimetry (ICTAC) [29]. The analyses were
elemental analyzer (Model Vario MACRO Cube, Germany, ± 0,2%), the performed in an inert atmosphere using nitrogen with 99.996 % purity
oxygen content was determined by the difference between the carbon, (4.6 FID) from White Martins Gases Industriais Ltda (Praxair, Brazil) at
a flow rate of 50 mL/min. In addition, baseline experimental data were
Table 2 obtained for each heating rate to eliminate the buoyancy effects shown
Ultimate and proximate analyses and heating values of coconut as mass gain in the thermogravimetric analysis results [29].
fiber in dry basis.
Properties Values
2.3. Data analysis
Ultimate analysis [% m/m]
Carbon 46.54 ± 0.21
For treating the experimental data, initially, the thermogravimetric
Hidrogen 6.88 ± 0.29
Nitrogen 0.36 ± 0.02 curves (TG) and their derivatives (DTG) were normalized by Eq. (1) and
Sulfur 0.19 ± 0.07 Eq. (2), respectively, where mo is the mass at the initial temperature
Oxygen (by difference) 43.13 ± 0.01 (25.18 ± 0.17 °C) and m is the mass at any time t. The normalized
Ash content 2.90 ± 0.36 DTG, Eq. (2), was represented by dW/dt, where dm/dt corresponds to
Proximate analysis [% m/m]
Volatile matter content 79.61 ± 2.09
the experimental data of mass variation over time. Normalization was
Ash content 2.90 ± 0.36 performed to standardize the curves obtained for the different heating
Fixed carbon content 17.50 ± 2.43 rates (5, 10, 15 and 20 °C/min).
Heating Value, HV [MJ/kg]
Higher, HHV 17.70 ± 0.07 m
Lower, LHV 16.48 ± 0.07
W=
m0 (1)

3
F.C.R. Lopes and K. Tannous Thermochimica Acta 691 (2020) 178714

Table 3
Differential, f(α), and integral, g(α), reaction models applied to describe solid-state kinetics (adapted from Khawam and Flanagan [44]).
N° Reaction Model Symbol f(α) g(α)

Nucleation
1 Power Law P1 4 3/4 1/4

2 Power Law P2 3 2/3 1/3

3 Power Law P3 2 1/2 1/2

4 Power Law P4 2/3 1/2 3/2

5 Avrami-Erofe’ev (m = 2) A2 2(1 )[ ln (1 )]1/2 [ ln (1 )]1/2

6 Avrami-Erofe’ev (m = 3) A3 3(1 )[ ln (1 )]2/3 [ ln (1 )]1/3
7 Avrami-Erofe’ev (m = 4) A4 4(1 )[ ln (1 )]3/4 [ ln (1 )]1/4
Geometrical contraction models
8 Contracting area R2 2(1 )1/2 1 (1 )1/2
9 Contracting volume R3 3(1 ) 2/3 1 (1 )1/3
Reaction order
10 First order F1 (1 ) ln (1 )
11 n-th Order Fn (1 )n [1 (1 )(1 n ) ] (1 n)

Diffusion phenomena
12 One-dimensional D1 1/2 2

13 Two-dimensional D2 [ ln (1 )] 1 + (1 ) ln (1 )
14 Three dimensional Jader equation D3 (3/2)(1 )2/3 [1 (1 )1/3] 1 [1 (1 )1/3]2
15 Three dimensional Ginstling–Brounshtein equation D4 (3/2)[(1 ) 1/3 1] 1 (1 2 /3) (1 )2/3

dW
=
dm 1 the pre-exponential factor (1/s), E is the activation energy (kJ/mol), R
dt dt m 0 (2) is the universal gas constant (8.314 J/Kmol), and T is the absolute
temperature (K).
To evaluate the experimental conversions (α)exp and conversion
rates (dα/dt)exp, the Eq. (3) and Eq. (4) were used, respectively, where d E
( )theo = k (T ) f ( ) = A [f ( )][exp( )]
Wo, W, and Wf are the initial normalized mass, the mass at time t, and dt RT (5)
the final mass, respectively. The thermal decomposition analysis and
kinetic modeling were performed using Microsoft Excel Office 365 MSO
(version 16.0.12130.20232). 2.4.1. Isoconversional methods
Wo W The thermal decomposition kinetics of solid fuels can be analyzed
exp =
Wo Wf (3) by applying different isoconversional methods (differentials and in-
tegrals) that can solve Eq. (5) under non-isothermal conditions, which
d dW 1 provides the apparent activation energy (Eα) for each conversion level
=
dt exp dt (W0 Wf ) (4) (α).
The differential method developed by Friedman [17] solves the
conversion rate, Eq. (5), using the natural logarithm, resulting in Eq.
2.4. Kinetic approaches (6). In the latter equation, Tα,i corresponds to the respective tempera-
ture at each conversion level (α) for the different heating rates, re-
The region with remarkable mass loss generally characterizes the presented by sub-index “i”.
thermal decomposition of biomass. This effect can be represented by a
global reaction (single-step) as shown in scheme (1.a), where V and S ln
d
= ln[A f ( )]
E
represent the volatile matter and solid components generated, respec- dt ,i RT , i (6)
tively.
The integral methods are obtained from the rearrangement and
Biomass (Solid) V+S (1.a) integration of Eq. (5), expressed in relation to heating rate, β (°C/min),
However, this region can also be described by parallel and in- thus generating Eq. (7).
dependent reactions of each pseudo-component as shown in scheme d A
( ) AE
f Tf E
(1.b), where HC, C, and L are hemicellulose, cellulose, and lignin, v is g( ) = = exp RT dT = p (x )
0 f( ) T0 R (7)
the volatile matter fraction formed by condensable and non-con-
densable gases, and s is the solid fraction generated by each component, The left side of Eq. (7), g(α), expresses the integral form of the re-
respectively. action model (Table 3). The right side represents the temperature in-
HC vHC + sHC tegral and has no analytical solution, being solved via numerical ap-
Biomass (Solid) C vC + sC proximations, p(x), where x = E/RT. Several approaches to the integral
L vL + sL (1.b) solution have been proposed in the literature, namely: the Ozawa-
Flynn-Wall [18,19], the modified Coats-Redfern [21] and the Starink
Therefore, in this work, isoconversional methods (single-step reac- [22] methods, represented by Eqs. (8), (9), and (10), respectively.
tion model) and the independent parallel reactions scheme (multi-step
reaction model) were used to describe the decomposition kinetics of the log = [log (
A E
) log[g ( )] 2.315] 0.4567
E
i
biomass. R RT , i (8)
To define the thermal decomposition kinetics using these meth-
odologies, the conversion rate (dα/dt)theo was estimated according to A R E
i
Eq. (5), where k(T) is the conversion rate constant, usually expressed by ln = ln
T 2, i g( ) E RT , i (9)
the Arrhenius equation [33], f(α) is the differential reaction model, A is

4
F.C.R. Lopes and K. Tannous Thermochimica Acta 691 (2020) 178714

i A R E more than 10 % of the values obtained by the isoconversional methods

ln [ ] = ln[ ] 1.0008
T1.92
,i g ( )E RT , i (10) used to determine the reaction model [43].

For each conversion, the slope of the straight lines ln (dα/dt), log β, d
dt E
ln = lnA
ln (β/T2) and ln (β/T1.92) were defined in relation to the inverse ab- f( ) RT (14)
solute temperature (1/T). The apparent activation energies (Eα) for
each conversion level were obtained by the slopes (m) as follows: Eα = In the same way applied to the isoconversional and master plots
-mαR, Eα = -mαR/0.4567, Eα = -mαR, and Eα = -mαR/1.0008, for methods, the linearization method was conducted considering the 19
Friedman (FD), Ozawa-Flynn-Wall (OFW), modified Coats-Redfern conversion levels (0.05−0.95) and three heating rates of the thermo-
(CRM) and Starink (ST) equations, respectively. To evaluate the quality gravimetry (5, 10 and 15 °C/min). After obtaining the kinetic para-
of fit of the linear regressions of each conversion in all methods, the meters, a comparison between the experimental, Eq. (4), and theore-
determination coefficient (R2) was determined. tical data, Eq. (5), was performed by simulating the theoretical
In addition to these methods, Vyazovkin [23] proposed an advanced conversions and conversion rates using the fourth order Runge-Kutta
method, in which the activation energy is obtained from the direct method.
application of the numerical integral, according to Eq. (11), where I(Eα,
Tα,i) = (Eα/R)[p(x)]. 2.4.4. Independent parallel reactions scheme, IPRS
n n The previously discussed isoconversional methods predict that
j I (Ea, T , i )
= = min biomass decomposes in a single-step reaction, characterized as a global
i=1 j i iI (Ea, T , j ) (11) method. For a better description of pyrolysis kinetics, the independent
parallel reactions scheme, IPRS, was applied. In IPRS (multiple reac-
To solve Vyazovkin method (VZ), the minimization function, Ω, was
tions model) the theoretical conversion, (α)theo, and conversion rate,
used by applying the SOLVER optimization tool with the Generalized
(dα/dt)theo, in relation to the main components of biomass, namely
Reduced Gradient method of Microsoft Excel Office 365 MSO (version
hemicellulose (HC), cellulose (C), and lignin (L) were described by Eq.
16.0.12130.20232). For the temperature integral expression I(Eα,Tα,i)=
(15) and (16), respectively.
(Eα/R)[p(x)], the 8th degree rational function approximation, Eq. (12),
was used, in which x=E/RT [42]. n
( )theo = Fi ( )i = FHC HC + FC C + FL L
p (x ) =
exp ( x )
i=1 (15)
x
x7 + 70x 6 + 1886x 5 + 24920x 4 + 170136x 3 + 577584x 2 + 844560x + 357120 n
x 8 + 72x7 + 2024x 6 + 28560x 5 + 216720x 4 + 880320x 3 + 1794240x 2 + 1572480x + 403200
d d d HC d C d L
= Fi = FHC + FC + FL
dt theo dt dt dt dt (16)
(12) i=1 i

To determine the apparent activation energy (Eα) using the iso- in which, Fi refers to the volatilized fractions of each component; and
conversional methods, the experimental data for the conversions from (dα/dt)i, given by Eq. (5), the conversion rates of each component
0.05 to 0.95 were considered, with 0.05 steps (totaling 19 conversion (hemicellulose, cellulose, and lignin).
levels). To apply the Eqs. (15) and (16), it was necessary to assume the
reaction model that better represented the thermal decomposition. For
2.4.2. Determination of reaction model this work, Fn-order reaction model, f(α)=(1-α)n, was assumed, re-
To describe the thermal decomposition reaction model (Table 3), presenting the reaction rate in relation to the amount of remaining
the integral master plots method recommended by the International reagents [44].
Confederation for Thermal Analysis and Calorimetry (ICTAC) [29] was The theoretical conversions (α)theo and conversion rates (dα/dt)theo
used. The method is based on the normalization of Eq. (7), generating were determined by applying the fourth-order Runge-Kutta method,
Eq. (13), where g(α)/g(0.5) are the theoretical master plots and p(x)/p and theoretical data were adjusted to the experimental data using the
(x0.5) are the experimental master plots [43]. The reference conversion least squares method. The curves of each heating rate (5, 10 and 15 °C/
(α) of 0.5 was chosen because it is the intermediate point of the con- min) were separately modeled by varying the activation energy (Ei),
version range (0.05 to 0.95). pre-exponential factor (Ai), reaction order (ni), and composition (Fi) of
E A
each component.
g( ) R
p (x ) p (x )
= E A
=
g (0.5) p (x 0.5 ) p (x 0.5 ) 2.4.5. Quality of fit
R (13)
The evaluation of the reaction model, g(α), that better described the
For solving the function p(x), the approximation of the temperature single-step thermal decomposition (Section 2.4.2 Determination of re-
integral presented in Eq. (12) was used. In the master plots method, it is action model), and the kinetic parameters obtained by the linearization
necessary to know the activation energy before estimating p(x) to en- method (Section 2.4.3. Determination of pre-exponential factor), as well
sure that it does not vary significantly with the biomass conversion as the fit of the theoretical data (Section 2.4.4. Independent parallel re-
[29]. Thus, for the determination of x=E/RT, it was used the average of actions scheme, IPRS) in relation to the experimental data, were de-
the activation energies provided by isoconversional method in- termined based on the average percentage deviation (APD), Eq. (17).
dividually for each conversion level. SQR is the least squares method, given by Eq. (18). In Eqs. (17) and
(18), N represents the number of experimental data obtained in the
2.4.3. Determination of pre-exponential factor thermogravimetric analysis and, (y)i,exp and (y)i,theo correspond to the
After defining the reaction model (Section 2.4.2. Determination of experimental and theoretical data, respectively, related to p(x)/p(x0.5)
reaction model) compatible with the pyrolysis reaction, it was possible and g(α)/g(0.5), conversion (α), and conversion rate (dα/dt). The
to determine the pre-exponential factor (A) and the global activation maximum experimental values, ymax, are[p(x)/p(x0.5)]max, αmax, and
energy (E) that would allow defining the biomass decomposition ki- (dα/dt)max.
netics in a single-step model. For this, the linearization of ln[(dα/dt)/f
(α)] was applied in relation to the inverse absolute temperature, 1/T, SQR
N
obtained from Eq. (14). The intercept represents ln A, and the slope ADP (%) = 100
defines the global activation energy (E). This energy should not vary
ymax (17)

5
F.C.R. Lopes and K. Tannous Thermochimica Acta 691 (2020) 178714

modeling.

3.1.2. Volatile release

The second stage was characterized by the release of volatiles
caused by the breakdown of hemicellulose and cellulose, and partial
decomposition of lignin [8,9,25,46]. Different gaseous compounds are
released being precursors of the liquid fraction [15] by condensation. In
Fig. 1 (a), it is observed that the volatilization stage, represented by the
greatest mass loss (55.00 ± 0.00 %), occurred between
192.42 ± 20.12 °C and 382.58 ± 8.49 °C. In the normalized DTG
curve (Fig. 1b), this stage was characterized by the presence of two
decomposition peaks, the first peak (-7.45*10−4 ± 3.78*10-4 1/s) cor-
responding to the decomposition of hemicellulose (272.78 ± 9.13 °C),
and the second peak (-1.59*10-3 ± 6.81*10-4 1/s), to the decomposi-
tion of cellulose (335.99 ± 9.43 °C). Lignin decomposes smoothly over
a wide range, 200–600 °C, with no characteristic decomposition peak
observed in the volatilization stage. This means that when hemi-
cellulose and cellulose react, the lignin also breaks down, but does not
govern the overall reaction [2,25].

3.1.3. Carbonization
The third stage, which occurred from 382.58 ± 8.49 °C to the final
temperature of 699.87 ± 0.11 °C, was characterized by the carboni-
zation of the remaining carbonaceous material [26]. A mass loss of
13.08 ± 1.74 % was observed. At this stage, it is believed that gases
such as carbon dioxide (CO2) and water (H2O) are released. The final
residue, consisting of minerals and coal fraction, was 24.90 ± 1.73 %.
In general, a similar normalized mass (W) and DTG (dW/dt) profile
was observed in Fig. 1(a) and (b), respectively, for the different heating
rates evaluated (5, 10, 15 and 20 °C/min). However, it should be noted
that the W (Fig. 1a) showed a different behavior for the heating rate of
20 °C/min, in which it intersected other curves and overlapped the
heating rate of 15 °C/min. It was also found that the initial temperature
of the volatilization stage increased with the increase in heating rate,
Fig. 1. (a) Normalized mass, W, (b) and normalized DTG, dW/dt, as a function
but for the rate of 20 °C/min, it was observed that it was lower than the
of temperature of coconut fiber decomposition for heating rates of 5, 10, 15 and
temperatures found for the other rates analyzed. This may have oc-
20 °C/min.
curred because the increase in heating rate may lead to the influence of
transport phenomena (mass and heat) that were not observed for lower
N
SQR = [(y )i, exp (y )i, theo ]2 heating rates.
i=0 (18) Based on the experimental data, it was decided to define the pyr-
olysis kinetics considering only the data obtained for 5, 10 and 15 °C/
min, since the heating rate of 20 °C/min showed atypical behavior and
3. Results and discussion may influence the decomposition kinetic results.

3.1. Thermal decomposition analysis 3.2. Analysis of kinetic parameters by the single-step reaction model

The thermogravimetry of coconut fiber was performed in an inert 3.2.1. Apparent activation energy analysis
atmosphere to evaluate the decomposition profile and biomass pyr- The global isoconversional methods, Friedman [17], Ozawa-Flynn-
olysis kinetics for the four heating rates (5, 10, 15 and 20 °C/min). Wall [18,19], modified Coats-Redfern [21], Starink [22] and Vyazovkin
Fig. 1(a) and 1(b) show the normalized mass (W) and DTG (dW/dt) [23], were applied to obtain the apparent activation energy for the
curves in relation to temperature. The thermal decomposition of bio- decomposition of coconut fiber in relation to the conversion. For the
mass showed three characteristic stages: (i) dehydration, (ii) volatili- first four methods mentioned, the apparent activation energies were
zation, and (iii) carbonization. These three stages were also observed obtained from the linearization of ln (dα/dt), log β, ln (β/T2) and ln (β/
for other biomasses such as coconut copra [10], empty fruit bunch and T1.92), respectively, in relation to the inverse absolute temperature
palm kernel shell [26], and coconut shell [11,12] with particularities in shown in Fig. 2.
the temperature ranges. When analyzing Fig. 2(a–d), a different behavior was observed in
the linearization plots obtained using Friedman method (FD). This can
3.1.1. Dehydration be attributed to the fact this is the only one that depends directly on the
The first stage was characterized by the reactions of dehydration conversion rate (dα/dt). In Fig. 2(a), a change in the concavity among
and release of compounds with low molecular weight (extractives) the lines was observed, with two conversion zones, the first resulting
present in the sample [26,45]. This stage occurred in the temperature from the hemicellulose decomposition, between 0.15 ≤ α ≤ 0.35
range between 25.18 ± 0.17 °C and 192.42 ± 20.12 °C, with a mass (E , HC = 171.4 ± 11.6 kJ/mol), and the second from the cellulose
loss (normalized TG) of 7.01 ± 0.04 %. The normalized DTG was re- decomposition, between 0.40 ≤ α ≤ 0.85 (E , C = 178.3 ± 12.0 kJ/
presented by a first peak of -1.47*10−4 ± 7.45*10-5 1/s at mol). In addition, the change in the concavity may be indicative of the
66.36 ± 9.11 °C. In this work, the released mass in this stage was occurrence of multiple reactions [47].
considered negligible; therefore, it was not considered in the kinetic In Fig. 2(b–d), apparent parallelism among the linear regressions

6
F.C.R. Lopes and K. Tannous Thermochimica Acta 691 (2020) 178714

Fig. 2. Linearization plots for the isoconversional methods: (a) Friedman, (b) Ozawa-Flynn-Wall, (c) modified Coats–Redfern and (d) Starink.

was verified, no change in concavity has been observed. The integral

methods of OFW, CRM and ST are based on heating rate, but not on the
biomass conversion rate, being less sensitive to changes in the concavity
among linearization plots. In this analysis, two groups of lines were also
found. In the first group (0.15 ≤ α ≤ 0.45), due to the decomposition
of hemicellulose, the lines were further apart, and the average apparent
activation energy in this range was E , HC = 164.3 ± 15.7 kJ/mol,
E , HC = 163.6 ± 16.2 kJ/mol, and E , HC = 163.8 ± 16.2 kJ/mol, re-
spectively. In the second group (0.50 ≤ α ≤ 0.90), due to the de-
composition of cellulose, the lines were closer to each other, and the
average apparent activation energies were E , C = 182.2 ± 2.8 kJ/mol,
E , C = 181.6 ± 2.9 kJ/mol, and E , C = 181.8 ± 2.9 kJ/mol, respec-
tively. The quality of fit of the linear regressions obtained by iso-
conversional methods was calculated by determination coefficients,
where R² > 0.99, for all methods and conversion levels considered.
The activation energies determined in relation to the conversion,
based on the slope of the lines of the isoconversional methods (FD,
OFW, CRM and ST) are shown in Fig. 3. In addition, the activation Fig. 3. Activation energy and conversion rate as a function of conversion ap-
energy was determined according to the VZ method. The activation plied for different isoconversional methods (β = 10 °C/min).
energies used to solve x = E/RT in Eq. (12) were the energies (Eα)
obtained by the isoconversional CRM method. As previously observed, (46.5 %), 140.4–188.3 kJ/mol (34.2 %), 138.8–187.8 kJ/mol (35.3 %),
the energies obtained by the four integral methods were very similar, 139.0–188.0 kJ/mol (35.2 %), and 139.0–188.0 kJ/mol (35.2 %), re-
and it was possible to choose any of them. Thus, the CRM method was spectively.
chosen to facilitate convergence in the minimization of Eq. (11). According to Di Blasi [15], the increase in activation energy as a
In Fig. 3, the apparent activation energy evolution in relation to the function of thermal conversion is caused probably by polymeric dif-
conversion obtained by the FD method showed slightly different be- ferences. The initial range of conversion of lignocellulosic biomass is
havior compared to the other methods (OFW, CRM, ST, and VZ). In related to the depolymerization of hemicellulose, which has low energy
general, an increase in activation energy was observed with the in- requirements. The cellulose is formed by very stable links, thus re-
crease in the biomass conversion (0.15 < α < 0.90), with variation quiring more energy for the breaking and decomposition [25]. This can
between the minimum and maximum energies of 157.1−230.2 kJ/mol be confirmed by the peaks of conversion rates (dα/dt) related to these

7
F.C.R. Lopes and K. Tannous Thermochimica Acta 691 (2020) 178714

components, as shown in Fig. 3, in which the hemicellulose decom-

posed between 0.15 ≤ α≤0.45 and cellulose for 0.50 ≤ α≤0.90. Si-
milar behavior was obtained for palm kernel shell, empty fruit bunch
[26], and sugarcane straw [25].
When activation energy varies significantly with the conversion
(> 30 %), the process can be characterized possibly by multiple reac-
tions [29]. Nevertheless, data from the single-step reaction kinetics
gives us a prior insight into the pyrolysis process. Thus, it was necessary
to obtain the reaction model, f(α) (Section 3.2.2. Reaction model analysis)
and the pre-exponential factor, A (Section 3.2.3. Pre-exponential factor
analysis).

3.2.2. Reaction model analysis

The knowledge of the reaction model is essential in the kinetic
studies since it helps to describe and improve the understanding of the
pyrolysis. For biomass, it was commonly assumed a first or nth-order
reaction model [14,25]. However, it is recommended a prior verifica- Fig. 4. Theoretical and experimental master plots as a function of conversion at
tion for choosing appropriate reaction model [29]. heating rates of (o) 5 °C/min, (Δ) 10 °C/min, (◊) 15 °C/min.
In this work, the reaction model, f(α), representative of volatiliza-
tion stage, was defined applying the integral master plots method. The
rates of 5, 10 and 15 °C/min, respectively. It is emphasized that despite
theoretical curves were constructed based on the fifteen reaction
the slightly high deviations applying the other isoconversional methods
models presented in Table 3. For the Fn-order reaction model, f(α)=(1-
(FD, OFW, CRM, and ST), an agreement was maintained the same dif-
α)n, the exponent n was varied from 2 to 7. To estimate the experi-
fusional reaction model.
mental data, the average apparent activation energies
Biomass is a complex material that can present different phenom-
(0.15 < α < 0.90) obtained by each isoconversional method were in-
enological characteristics throughout its decomposition. Thus, the re-
dividually used: 179.4 kJ/mol, 174.4 kJ/mol, 173.7 kJ/mol, 173.9 kJ/
action models are just a mathematical representation of such behavior,
mol, and 169.2 kJ/mol for FD, OFW, CRM, ST, and VZ, respectively.
which allows an approximation of the real one [26].
Table 4 shows the average percentual deviations (APD) between
theoretical and experimental master plots. It can be observed high APD
for Fn-order reaction model superior of 4, assigning difficult physical 3.2.3. Pre-exponential factor analysis
meaning [48], and therefore, they were disregarded in this analysis. After determining the reaction model, the pre-exponential factor (A)
According to the literature (Table 1), the biomasses are usually de- was defined based on the linearization method presented in Section
scribed between first-order and third-order reaction models. Fig. 4 2.4.3 (Determination of pre-exponential factor). Thus, assuming the three-
presents the most representatives theoretical master plots, g(α)/g(0.5), dimensional conversion function of Jader D3 (Table 3), the linear
related to the diffusional reaction model (D1, D2, D3, and D4), and the function obtained to represent the global pyrolysis reaction of coconut
experimental master plots, p(x)/p(x0.5), as a function of conversion fiber is given by Eq. (19).
considering the average activation energy from Vyazovkin method, d
1
since these presented the lowest APD. Comparing the master plots ln
dt
= 15608.0 + 18.70 R2 = 0.9637
(Table 4 and Fig. 4), the best fitting reaction model was the three-di- f( ) T (19)
mensional diffusion of Jader D3, f(α)=(3/2)(1-α)2/3[1-(1-α)1/3]−1 and
The pre-exponential factor (1.5 * 108 1/s) was obtained based on the
g(α)=[1-(1-α)1/3]2, with APD of 7.5 %, 6.0 %, and 5.5 % for the heating
intercept. In addition, a new global activation energy (129.8 kJ/mol)
was defined, given by the slope, differing 38.3 %, 34.4 %, 33.9 %, 34.0
Table 4
%, and 30.4 % in relation to the average apparent activation energies
Average Percentage Deviation (APD) between theoretical and experimental
obtained for the FD, OFW, CRM, ST and VZ methods, respectively
master plots for the heating rates of 5, 10 and 15 °C/min.
(Section 3.2.1. Apparent activation energy analysis). The differences
Reaction Model Symbol Average Percentage Deviation (APD). found were greater than those recommended by the literature [43],
which states that the energies between both methodologies should be
5 °C/min 10 °C/min 15 °C/min
lower than 10 %.
Nucleation
P1 25.6 25.9 26.5
P2 25.4 25.7 26.3 3.2.4. Evaluation of kinetic parameters by single-step reaction model
P3 24.9 25.2 25.8 Fig. 5(a) and (b) show the theoretical (α)theo and experimental (α)exp
P4 21.5 21.4 21.8 conversion, and the theoretical (dα/dt)theo and experimental (dα/dt)exp
A2 23.3 23.4 23.9 conversion rate, respectively, obtained for single-step reaction model
A3 24.4 24.7 25.2
and three heating rates (5, 10 and 15 °C/min).
A4 24.9 25.2 25.8
Geometrical contraction models According to Fig. 5(a), the theoretical curves as a function of the
R2 21.7 21.6 22.0 temperature show good agreement with the experimental results, with
R3 20.8 20.6 21.0 average percentage deviation (APD) of 3.2 %, 2.7 %, and 3.5 % for the
Reaction order
heating rates of 5, 10 and 15 °C/min, respectively. In Fig. 5(b), on the
F1 18.2 17.8 18.0
F2 9.3 12.4 13.5 other hand, agreement was quite satisfactory between the theoretical
F3 203.2 224.7 233.3 and experimental data, with average percentage deviation (APD) of
Diffusion phenomena 12.5 %, 13.3 %, and 13.2 %, respectively. Although the results were
D1 19.0 18.6 18.9 satisfactory, the theoretical conversion rates were characterized by a
D2 15.5 14.8 14.8
single peak of decomposition related to the single-step reaction model,
D3 7.5 6.0 5.5
D4 13.3 12.4 12.3 differently from the experimental data. The number of points (N) con-
sidered in the analysis was 2245, 1041, and 688 for each heating rate,

8
F.C.R. Lopes and K. Tannous Thermochimica Acta 691 (2020) 178714

activation energy varies from 40 to 150 kJ/mol [26], and the reaction
order could be represented between 1 and 3 [26,50]. Thus, activation
energy of 60 kJ/mol and reaction order of 3 were considered as the
initial parameters. For the initial estimation of the volatilized fractions
(Fi) of each pseudo-component, the chemical compositionof hemi-
cellulose (19.8 ± 5.5 wt.%), cellulose (42.0 ± 5.5 wt.%), and lignin
(35.6 ± 0.8 wt.%) established in previous work by Higuchi and Tan-
nous [51], were considered.
The conversion and conversion rates modeled as a function of the
temperature for each heating rate of 5, 10 and 15 °C/min and tem-
perature range of 150−700 °C are shown in Fig. 6. According to the
theoretical data, the thermal decomposition of the pseudo-components
(HC, C, and L) occurred in the temperature ranges between
154.3 ± 4.0 °C to 322.2 ± 116 °C, 242.5 ± 0.9 °C to
367.6 ± 11.1 °C, and 245.9 ± 9.8 °C to 483.7 ± 15.7 °C, respec-
tively. The lignin reaction is prolonged at higher temperatures, but
above of 483.7 ± 15.7 °C, the mass loss of biomass was very small.
These ranges are in agreement with Di Blasi [15], who mentions that
the reaction temperature range of hemicellulose is between 200 °C and
300 °C, cellulose between 250 °C and 350 °C and lignin between 200 °C
and 500 °C for different lignocellulosic biomasses.
Table 5 shows the kinetic parameters and the average percentage
deviation (APD) calculated for the conversion and conversion rate of
the thermal decomposition of coconut fiber. The number of points (N)
considered in the analyses was 7161, 3883, and 2790.
It was observed that the activation energy found for hemicellulose,
HC (100.1 kJ/mol) was lower than the range obtained using the in-
tegral isoconversional methods in Section 3.2.1 (Apparent activation
energy analysis) (159.7–166.0 kJ/mol), while that for cellulose, C
(196.3 kJ/mol) it was greater than the established range
(182.2–184.6 kJ/mol). These results are in agreement with the litera-
ture for other biomasses (Table 1), next to the upper limits of variation
(HC:85−171 kJ/mol and C:149−200 kJ/mol). For lignin (79.4 kJ/
Fig. 5. Conversions and conversion rates as a function of temperature obtained mol), none activation energy could be defined using the isoconversional
by single-step reaction model - experimental (symbols), theoretical (dashed methods, however the result obtained by IPRS matches from 49 to
lines). 93 kJ/mol (Table 1).
In the Table 5, it is also possible to highlight that the standard de-
respectively. viation of the activation energies was null for all pseudo-components.
Although the analysis of activation energy using isoconversional According to Rueda-Ordóñez et al. [26], this value provides more re-
methods (Section 3.2.1. Apparent activation energy analysis) and linear- liable results and can be used for different heating rates.
ization method (Section 3.2.3 Pre-exponential factor analysis) did not In the simulation, the pre-exponential factors are used as adjustment
represent the pyrolysis of coconut fiber with high precision, these parameter, thus, the curve was displaced by changing log Ai to obtain
methods are simple and easy to apply, making them suitable for iden- the smallest average percentage deviation (APD). The log Ai values
tifying the presence of multiple reactions [29]. In addition, the global found in this work for thehemicellulose (7.5 ± 0.1 1/s), cellulose
kinetic parameters are recommended as first guess in the application of (15.0 ± 0.0 1/s) and lignin (3.3 ± 0.2 1/s) are within the limits es-
more complex schemes, giving support for the convergence of the re- tablished in the literature (Table 1). Besides that, in this analysis, it was
sults. also observed there are no significant change of log Ai values at dif-
Next, the investigation of decomposition kinetics using the in- ferent heating rates.
dependent parallel reactions scheme (IPRS) was developed in order to It is noteworthy that the activation energy (Ei) and pre-exponential
obtain better details on the thermal decomposition of coconut fiber, factor (log Ai) for the coconut fiber are within the established ranges for
taking into consideration the main components (hemicellulose, cellu- other lignocellulosic biomass shown in Table 1, providing greater re-
lose and lignin) present in the biomass. liability of parameters in the decomposition process.
Also, the results presented in Table 5 show that the decomposition
reactions of hemicellulose and cellulose were well-described by the
3.3. Determination and evaluation of kinetic parameters by the multi-step first-order reactions, while the lignin reaction was well-represented by
reaction model (independent parallel reactions scheme, IPRS) the second-order reaction.
The final volatilized fractions of each pseudo-component, Fi (30 %
To analyze the decomposition of coconut fiber using the in- HC, 40 % C and 35 % L) differed significantly to the chemical com-
dependent parallel reactions scheme (IPRS), three reactions related to position of coconut fiber for hemicellulose established by Higuchi and
the decomposition of hemicellulose (HC), cellulose (C), and lignin (L) Tannous [51] (19.8 w.% HC, 42.0 w.% C, and 35.6 w.% L).
were considered. The first guess of activation energies were based on Finally, the results of the APD of the conversion (α) and conversion
the average values obtained for hemicellulose (E , HC = 164 kJ/mol) rate (dα/dt) obtained using IPRS (multi-step) were shown to be lower
and cellulose (E , C = 181 kJ/mol) using the integral isoconversional than 3.5 % and 3.0 % (Table 5), respectively, for the heating rates of 5,
methods of OFW, CRM and ST, and these components were represented 10 and 15 °C/min. Despite the higher precision of kinetic parameters
by first-order reactions [49]. Lignin has a very complex chemical obtained by the independent parallel reactions scheme (IPRS), it should
composition and a wide thermal decomposition temperature range. The be noted the importance to consider this in conjunction with the

9
F.C.R. Lopes and K. Tannous Thermochimica Acta 691 (2020) 178714

Fig. 6. Conversions and conversion rates as a function of temperature obtained by multi-step reaction model (IPRS) for four heating rates, (a,b) 5 °C/min, (c,d) 10 °C/
min, and (e,f) 15 °C/min.

Table 5 isoconversional methods due to the numerous combinations of results

Kinetic parameters of the coconut fiber thermal decomposition obtained by that the fitting method provide.
independent parallel reactions scheme.
Heating Pseudo- Ei (kJ/ log ni (-) Fi (-) APDα (%) APDdα/
4. Conclusions
rate component mol) Ai dt (%)
(°C/min) reaction (1/s)
The use of pyrolysis could be a good solution to solve environmental
5 HC 100.1 7.4 1 0.3 3.0 3.0 problems of coconut residue due its improper disposal. According to the
C 196.3 15.0 1 0.4 ultimate and proximate analyses, and the higher heating value, coconut
L 79.4 3.1 2 0.3
10 HC 100.1 7.5 1 0.3 2.8 2.9
fiber showed to have energetic potential, presenting significant amount
C 196.3 15.0 1 0.4 of volatile matter (79.61 ± 2.09 wt.%) and low ash content
L 79.4 3.3 2 0.3 (2.90 ± 0.36 wt.%). A low nitrogen and sulfur contents were observed
15 HC 100.1 7.6 1 0.3 2.9 2.8 providing lower emissions of oxides. The thermal decomposition of
C 196.3 15.0 1 0.4
biomass in an inert atmosphere showed three stages (dehydration, vo-
L 79.4 3.4 2 0.3
latile release, and carbonization) at the three heating rates applied. The
investigation of the apparent activation energy, corresponding to the

10
F.C.R. Lopes and K. Tannous Thermochimica Acta 691 (2020) 178714

volatile release stage, by four isoconversional methods revealed the potential, Renew. Energy 148 (2020) 756–771, https://doi.org/10.1016/j.renene.
presence of multiple reactions since it varied significantly with con- 2019.10.162.
[12] H. Siddiqi, U. Kumari, S. Biswas, A. Mishra, B.C. Meikap, A synergistic study of
version between 0.15−0.90 (> 30 %). Besides that, after evaluation of reaction kinetics and heat transfer with multicomponent modelling approach for
kinetic triplet parameters by single-step reaction model, it was observed the pyrolysis of biomass waste, Energ. 204 (2020) 117933, https://doi.org/10.
that the conversion rates showed average percentage deviation (APD) 1016/j.energy.2020.117933.
[13] M. Balat, M. Balat, E. Kirtay, H. Balat, Main routes for the termo-conversion of
higher than 10 %. The three independent parallel reactions scheme biomass into fuels and chemicals. Part 1: Pyrolysis systems, Energ. Convers.
(multi-step reaction model) was applied to improve the quality of fit, Manage. 50 (2009) 3147–3157, https://doi.org/10.1016/j.enconman.2009.08.014.
obtaining APD lower than 3.5 %. The two kinetic approaches well-de- [14] V.S. Carvalho, K. Tannous, Thermal decomposition kinetics modeling of energy
cane Saccharum robustum, Thermochim. Acta 657 (2017) 56–65, https://doi.org/
scribed the thermal decomposition of biomass. However, it was ob- 10.1016/j.tca.2017.09.016.
served that the IPRS generated more reliable data, when applied to- [15] C. Di Blasi, Modeling chemical and physical processes of wood and biomass pyr-
gether with the information obtained from the single-step method. olysis, Prog. Energ. Combust. Sci. 34 (2008) 47–90, https://doi.org/10.1016/j.pecs.
2006.12.001.
[16] Y.J. Rueda-Ordóñez, K. Tannous, E. Olivares-Gómez, An empirical model to obtain
CRediT authorship contribution statement the kinetic parameters of lignocellulosic biomass pyrolysis in an independent par-
allel reactions scheme, Fuel Process. Technol. 140 (2015) 222–230, https://doi.org/
Fernanda Cristina Rezende Lopes: Conceptualization, Formal 10.1016/j.fuproc.2015.09.001.
[17] H.L. Friedman, Kinetics of thermal degradation of char-forming plastics from
analysis, Investigation, Writing - review & editing. Katia Tannous: thermogravimetry Application to a phenolic plastic, J. Polym. Sci. 6 (1964)
Resources, Investigation, Writing - review & editing, Supervision. 183–195, https://doi.org/10.1002/polc.5070060121.
[18] T. Ozawa, A new method of analyzing thermogravimetric data, Bull. Chem. Soc.
Jpn. 38 (1965) 1881–1886, https://doi.org/10.1246/bcsj.38.1881.
Declaration of Competing Interest [19] J.H. Flynn, L.A. Wall, General treatment of thermogravimetry of polymers, J. Res.
Nat. Bur. Stand. A. Phys. Chem. 70 (1966) 487–523, https://doi.org/10.6028/jres.
070A.043.
The authors declare that they have no known competing financial [20] A.W. Coats, J.P. Redfern, Kinetic parameters from thermogravimetric data. II, J.
interests or personal relationships that could have appeared to influ- Polym. Sci., Part B. 3 (1965) 917–920, https://doi.org/10.1002/pol.1965.
ence the work reported in this paper. 110031106.
[21] R.L. Braun, A.K. Burnham, J.G. Reynolds, J.E. Clarkson, Pyrolysis kinetics for la-
custrine and marine source rocks by programmed micropyrolysis, Energ. Fuel 5
Acknowledgements (1991) 192–204, https://doi.org/10.1021/ef00025a033.
[22] M.J. Starink, The determination of activation energy from linear heating rate ex-
periments: a comparison of the accuracy of isoconversion methods, Thermochim.
The authors thank the financial support received from the
Acta 404 (2003) 163–176, https://doi.org/10.1016/S0040-6031(03)00144-8.
Coordination for the Improvement of HigherEducation Personnel, [23] S. Vyazovkin, Advanced isoconversional method, J. Therm. Anal. 49 (1997)
CAPES (PROEX process number 0566/2017), the National Council for 1493–1499, https://doi.org/10.1007/BF01983708.
Scientific and Technological Development, CNPq (Grant No. 142361/ [24] G. Mishra, T. Bhaskar, Non isothermal model free kinetics for pyrolysis of rice
straw, Bioresour. Technol. 169 (2014) 614–621, https://doi.org/10.1016/j.
2016-0), Brazil, and also for Espaço da Escrita/Pró-Reitoria de biortech.2014.07.045.
Pesquisa/UNICAMP by the language services provided. [25] Y.J. Rueda-Ordóñez, K. Tannous, Isoconversional kinetic study of the thermal de-
composition of sugarcane straw for thermal conversion processes, Bioresour.
Technol. 196 (2015) 136–144, https://doi.org/10.1016/j.biortech.2015.07.062.
Appendix A. Supplementary data [26] Y.J. Rueda-Ordóñez, C.J. Arias-Hernández, J.F. Manrique-Pinto, P. Gauthier-
Maradei, W.B. Bizzo, Assessment of the thermal decomposition kinetics of empty
Supplementary material related to this article can be found, in the fruit bunch, kernel shell and their blend, Bioresour. Technol. 292 (2019) 121923,
https://doi.org/10.1016/j.biortech.2019.121923.
online version, at doi:https://doi.org/10.1016/j.tca.2020.178714. [27] C.N. Arenas, M.V. Navarro, J.D. Martínez, Pyrolysis kinetics of biomass wastes
using isoconversional methods and the distributed activation energy model,
References Bioresour. Technol. 288 (2019) 121485, https://doi.org/10.1016/j.biortech.2019.
121485.
[28] H.R. Guimarães, K. Tannous, Influence of torrefaction on the pyrolysis of energy
[1] R.E. Guedes, A.S. Luna, A.R. Torres, Operating parameters for bio-oil production in cane: a study on thermal properties and decomposition kinetics, J. Therm. Anal.
biomass pyrolysis: a review, J. Anal. Appl. Pyrol. 129 (2018) 134–149, https://doi. Calorim. (2019), https://doi.org/10.1007/s10973-019-08584-z.
org/10.1016/j.jaap.2017.11.019. [29] S. Vyazovkin, A.K. Burnhamb, J.M. Criado, L.A. Pérez-Maqueda, C. Popescu,
[2] D.I. Aslan, B. Özoğul, S. Ceylan, F. Geyikçi, Thermokinetic analysis and product N. Sbirrazzuoli, ICTAC Kinetics Committee recommendations for performing kinetic
characterization of Medium Density Fiberboard pyrolysis, Bioresour. Technol. 258 computations on thermal analysis data, Thermochim. Acta 520 (2011) 1–19,
(2010) 105–110, https://doi.org/10.1016/j.biortech.2018.02.126. https://doi.org/10.1016/j.tca.2011.03.034.
[3] Empresa de Pesquisa Energética, Brazilian Energy Balance: Summary Report/year [30] J. Cai, W. Wu, R. Liu, G.W. Huber, A distributed activation energy model for the
2018, (2019) < http://www.epe.gov.br/pt/publicacoes-dados-abertos/ pyrolysis of lignocellulosic biomass, Green Chem. 15 (2013) 1331–1340, https://
publicacoes/balanco-energetico-nacional-2019 > . doi.org/10.1039/C3GC36958G.
[4] Food and Agriculture Organization of the United Nation, (2019) < http://www. [31] M. Hu, Z. Chen, S. Wang, D. Guo, C. Ma, Y. Zhuo, J. Chen, M. Laghari, S. Fazal,
fao.org/faostat/en/#data/QC&gt;. B. Xiao, B. Zhang, S. Ma, Thermogravimetric kinetics of lignocellulosic biomass
[5] A.S. Moura, R. Demori, R.M. Leão, C.L.C. Frankenberg, R.M.C. Santana, The in- slow pyrolysis using distributed activation energy model, Fraser-Suzuki deconvo-
fluence of the coconut fiber treated as reinforcement in PHB (polyhydroxybutyrate) lution, and iso-conversional method, Energ. Convers. Manage. 118 (2016) 1–11,
composites, Mater. Today Commun. 18 (2019) 191–198, https://doi.org/10.1016/ https://doi.org/10.1016/j.enconman.2016.03.058.
j.mtcomm.2018.12.006. [32] S. Vyazovkin, K. Chrissafis, M.L. Lorenzo, N. Koga, M. Pijolat, B. Roduit,
[6] Y. Yan, Developments in fibers for technical nonwovens, in: G. Kellie (Ed.), N. Sbirrazzuoli, J.J. Suñol, ICTAC Kinetics Committee recommendations for col-
Advances in Technical Nonwovens, Elsevier, Shaston, 2016, pp. 19–96, , https:// lecting experimental thermal analysis data for kinetic computations, Thermochim.
doi.org/10.1016/B978-0-08-100575-0.00002-4. Acta 590 (2014) 1–23, https://doi.org/10.1016/j.tca.2014.05.036.
[7] T. Siengchum, M. Isenberg, S.S.C. Chuang, Fast pyrolysis of coconut biomass - an [33] J.E. White, W.J. Catallo, B.L. Legendre, Biomass pyrolysis kinetics: a comparative
FTIR study, Fuel 105 (2013) 559–565, https://doi.org/10.1016/j.fuel.2012.09.039. critical review with relevant agricultural residue casa studies, J. Anal. Appl. Pyrol.
[8] I. Ali, H. Bahaitham, R. Naebulharam, A comprehensive kinetics study of coconut 91 (2011) 1–33, https://doi.org/10.1016/j.jaap.2011.01.004.
shell waste pyrolysis, Bioresour. Technol. 235 (2017) 1–11, https://doi.org/10. [34] Y.J. Rueda-Ordóñez, K. Tannous, Thermal decomposition of sugarcane straw, ki-
1016/j.biortech.2017.03.089. netics and heat of reaction in synthetic air, Bioresour. Technol. 211 (2016)
[9] C.G. Mothé, I.C. Miranda, Decomposition through pyrolysis process of coconut fiber 231–239, https://doi.org/10.1016/j.biortech.2016.03.035.
and rice husk and determination of kinetic parameters according isoconversional [35] K. Tannous, F.S. Silva, Particles and geometric shapes analyzer APOGEO, Mehdi
methods, J. Therm. Anal. Calorim. 131 (2018) 601–609, https://doi.org/10.1007/ Khosrow-Pour. (Org.). Encyclopedia of Information Science and Technology, 3rd
s10973-017-6377-2. ed., Hersley:IGI Global, 2015, pp. 3568–3579, https://doi.org/10.4018/978-1-
[10] V. Balasundram, N. Ibrahim, R.M. Kasmani, M.K.A. Hamid, R. Isha, H. Hasbullah, 4666-5888-2.ch349.
R.R. Ali, Thermogravimetric catalytic pyrolysis and kinetic studies of coconut copra [36] C. Finet, Heating value of municipal solid waste, Waste Manag. Res. 5 (1987)
and rice husk for possible maximum production of pyrolysis oil, J. Clean. Prod. 167 141–145, https://doi.org/10.1016/0734-242X(87)90047-4.
(2017) 218–228, https://doi.org/10.1016/j.jclepro.2017.08.173. [37] C.M. Santos, L.S. Oliveira, E.P.A. Rocha, A.S. Franca, Thermal conversion of de-
[11] H. Siddiqi, M. Bal, U. Kumari, B.C. Meikap, In-depth physiochemical character- fective coffee beans for energy purposes: characterization and kinetic modeling,
ization and detailed thermo-kinetic study of biomass wastes to analyze its energy Renew. Energy 147 (2020) 1275–1291, https://doi.org/10.1016/j.renene.2019.09.

11
F.C.R. Lopes and K. Tannous Thermochimica Acta 691 (2020) 178714

052. S.A. Rahimuddin, M. Gull, Pyrolysis, kinetics analysis, thermodynamics parameters

[38] M. Wu, Y. Wu, M. Wang, Energy and emission benefits of alternative transportation and reaction mechanism of Typha latifolia to evaluate its bioenergy potential,
liquid fuels derived from switch grass: a fuel life cycle assessment, Biotechnol. Prog. Bioresour. Technol. 245 (2017) 491–501, https://doi.org/10.1016/j.biortech.2017.
22 (2006) 1012–1024, https://doi.org/10.1021/bp050371p. 08.162.
[39] R. Saidur, E.A. Abdelziz, A. Demirbas, M.S. Hossain, S.A. Mekhilef, A review on [46] C.G. Mothé, I.C. Miranda, Characterization of sugarcane and coconut fibers by
biomass as a fuel for boilers, Renew. Sustain. Energ. Rev. 15 (2011) 2262–2289, thermal analysis and FTIR, J. Therm. Anal. Calorim. 97 (2009) 661–665, https://
https://doi.org/10.1016/j.rser.2011.02.015. doi.org/10.1007/s10973-009-0346-3.
[40] S. Ceylan, Y. Topçu, Pyrolysis kinetics of husk using thermogravimetric analysis, [47] E.G. Baroni, K. Tannous, Y.J. Rueda-Ordóñez, L.K. Tinoco-Navarro, The applic-
Bioresour. Technol. 156 (2014) 182–188, https://doi.org/10.1016/j.biortech.2014. ability of isoconversional models in estimating the kinetic parameters of biomass
01.040. pyrolysis, J. Therm. Anal. Calorim. 132 (2016) 909–917, https://doi.org/10.1007/
[41] E.R. Zanatta, T.O. Reinehr, J.A. Awadallak, S.J. Kleinübing, J.B.O. Santos, s10973-015-4707-9.
R.A. Bariccatti, P.A. Arroyo, E.A. Silva, Kinetic studies of thermal decomposition of [48] R.L. Braun, A.K. Burnham, Analysis of chemical reaction kinetics using a distribu-
sugarcane bagasse and cassava bagasse, J. Therm. Anal. Calorim. 125 (2016) tion of activation energies and simpler models, Energ. Fuel 1 (1987) 153–161,
437–445, https://doi.org/10.1007/s10973-016-5378-x. https://doi.org/10.1021/ef00002a003.
[42] L.A. Pérez-Maqueda, J.M. Criado, The accuracy of Senum and Yang’s approxima- [49] J.J.M. Órfão, F.J.A. Antunes, J.K. Figueiredo, Pyrolysis kinetics of lignocellulosic
tions to the Arrhenius integral, J. Therm. Anal. Calorim. 60 (2000) 909–915, materials - three independent reactions model, Fuel 78 (1999) 349–358, https://
https://doi.org/10.1023/A:1010115926340. doi.org/10.1016/S0016-2361(98)00156-2.
[43] B. Janković, Kinetic analysis of the nonisothermal decomposition of potassium [50] J.J. Manyà, E. Velo, L. Puigjaner, Kinetics of biomass pyrolysis: a reformulated
metabisulfite using the model-fitting and isoconversional (model-free) methods, three-parallel-reactions model, Ind. Eng. Chem. Res. 42 (2003) 434–441, https://
Chem. Eng. J. 139 (2018) 128–135, https://doi.org/10.1016/j.cej.2007.07.085. doi.org/10.1021/ie020218p.
[44] A. Khawam, D.R. Flanagan, Solid-state kinetic models: basics and mathematical [51] L.A. Higuchi, K. Tannous, Katia Tannous (Ed.), Avaliação da produtividade de
fundamentals, J. Phys. Chem. B 110 (2006) 17315–17328, https://doi.org/10. biomassas lignocelulósicas e potencial de utilização para fins energéticos,
1021/jp062746a. Universidade Estadual de Campinas, Campinas, 2019, , https://doi.org/10.20396/
[45] M.S. Ahmad, M.A. Mehmood, S.T.H. Taqvi, A. Elkamel, C.G. Liu, J. Xu, ISBN978-85-918766-0-0.

12