Energy Conversion and Management 105 (2015) 251–259

Contents lists available at ScienceDirect

Energy Conversion and Management

journal homepage: www.elsevier.com/locate/enconman

Pyrolysis behaviors and kinetic studies on Eucalyptus residues using

thermogravimetric analysis
Zhihua Chen a, Quanjie Zhu b, Xun Wang a, Bo Xiao a, Shiming Liu a,⇑
a
School of Environmental Science & Engineering, Huazhong University of Science and Technology, Wuhan 430074, China
b
Safety Engineering College, North China Institute of Science and Technology, Beijing 101601, China

a r t i c l e i n f o a b s t r a c t

Article history: The pyrolysis behaviors and kinetics of Eucalyptus leaves (EL), Eucalyptus bark (EB) and Eucalyptus saw-
Received 11 May 2015 dust (ESD) were investigated by using thermogravimetric analysis (TGA) technique. Three stages for
Accepted 30 July 2015 EL, EB and ESD pyrolysis have been divided using differential derivative thermogravimetric (DDTG)
Available online 7 August 2015
method and the second stage is the main pyrolysis process with approximately 86.93% (EL), 88.96%
(EB) and 97.84% (ESD) weight loss percentages. Kinetic parameters of Gaussian distributed activation
Keywords: energy model (DAEM) for EL, EB and ESD pyrolysis are: distributed centers (E0) of 141.15 kJ/mol (EL),
Pyrolysis
149.21 kJ/mol (EB), 175.79 kJ/mol (ESD), standard deviations (r) of 18.35 kJ/mol (EL), 18.37 kJ/mol
Eucalyptus residues
Kinetic
(EB), 14.41 kJ/mol (ESD) and pre-exponential factors (A) of 1.15E+10 s1 (EL), 4.34E+10 s1 (EB),
Thermogravimetric analysis 7.44E+12 s1 (ESD). A new modified discrete DAEM was performed and showed excellent fits to experi-
mental data than Gaussian DAEM. According to the modified discrete DAEM, the activation energies are
in ranges of 122.67–308.64 kJ/mol, 118.72–410.80 kJ/mol and 108.39–192.93 kJ/mol for EL, EB and ESD
pyrolysis, respectively. The pre-exponential factors of discrete DAEM have wide ranges of 4.84E+13–
6.12E+22 s1 (EL), 1.91E+12–4.51E+25 s1 (EB) and 63.43–4.36E+11 s1 (ESD). The variation of activation
energy versus conversion reveals the mechanism change during pyrolysis process. The kinetic data would
be of immense benefit to model, design and develop suitable thermo-chemical systems for the applica-
tion of Eucalyptus residues.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction problems. Therefore, the reuse of Eucalyptus residues could solve

pollution problems by means of thermo-chemical conversion
Due to the decrease in reservoir of fossil fuels and the increasing processes.
concerns on environmental protection, biomass is widely regarded Pyrolysis process is a promising technology for biomass utiliza-
as a clean, sustainable, renewable and alternative energy source in tion, which converts the biomass to bio-oil, char and gasses
the future. In recent decades, special attentions have been paid to depending on the pyrolysis conditions [4]. The pyrolysis process
the conversion of residual biomass into biofuels through is preferred because of its lower temperature requirement, oxygen
thermo-chemical conversion processes, such as pyrolysis, gasifica- independent operational conditions and higher-quality oils fea-
tion and combustion [1]. Particularly in China, biomass residues tures comparing to gasification and combustion [5]. In order to
had been widely used for electric power generation over last dec- use the Eucalyptus residues for energy applications through
ade [2]. The Eucalyptus trees, which represent fast-growing species, thermo-chemical conversion processes, it is very important to
have been planted extensively in many parts of the world to meet investigate their thermal characteristic and kinetics. Furthermore,
the increasing demand for paper pulp, timber and fiber [3]. pyrolysis also represents the initial stage of gasification and com-
Annually in China, a large amount of Eucalyptus residues such as bustion, its kinetic modeling is helpful in describing practical con-
the leaves, bark and sawdust are generated during the process of version processes and design more efficient reactors [6].
paper pulp production. Such abundant of these residues is an To date, thermogravimetric analysis (TGA) is considered as a
attractive energy source. Moreover, the accumulation and disposal powerful technique for pyrolysis behaviors characterization and
of these residues also constituted a source of environmental kinetic study. In a TGA experiment, the pyrolysis behaviors (mass
loss and mass loss rate) can be measured as functions of tempera-
⇑ Corresponding author.
ture and time in an inert atmosphere [7]. By using the TGA data,
E-mail address: [email protected] (S. Liu).
many mathematical methods such as model-fit and model-free

http://dx.doi.org/10.1016/j.enconman.2015.07.077
0196-8904/Ó 2015 Elsevier Ltd. All rights reserved.
252 Z. Chen et al. / Energy Conversion and Management 105 (2015) 251–259

were widely used in corresponding studies [8–10]. However, these with Arrhenius equation [12]. The differential form of such reac-
methods are limited to use in biomass pyrolytic kinetics investiga- tion rate equation can be expressed as:
tion due to the high heterogeneous (such as catalysis of alkalis and  
da A E
secondary reactions) of the pyrolysis process [11,12]. And the dis- r¼ ¼ exp  f ðaÞ ð1Þ
tributed activation energy model (DAEM) to treat the biomass
dT b RT
pyrolysis kinetics is even better. where r is the reaction rate, 1/K; a is the conversion, dimensionless;
The DAEM assumes that infinity irreversible first-order reac- T is absolute temperature, K; A is apparent pre-exponential factor,
tions happen during pyrolysis process. Such activation energy is 1/s; E is the apparent activation energy, J/mol; R is ideal gas con-
thought to be not a constant value but obeys a kind of distributed stant, 8.314 J/(mol K); b is the linear heating rate, K/s; f(a) is conver-
probability density function (PDF) such as Gaussian distribution sion dependence function (or reaction model) and its independent
[11–13]. Many attractive methods can be used to treat the DAEM variable, a, can be defined as
calculation, such as fitting method [14,15], iso-conversional
m0  mt
method [16] and discretization method [17–20]. Among these a¼ ð2Þ
m0  mf
methods, the discrete DAEM which discretize the infinity
first-order reaction as many (but finite) parallel first-order reac- where m0 (%), mt (%) and mf (%) is the initial mass, instant mass and
tions is even better for the DAEM calculation [18–20]. final mass of the sample recorded by TGA instrument, respectively.
So far, to our knowledge, there are few studies on the pyrolysis Integral form of Eq. (1) is given as
characteristics and kinetics of Eucalyptus residues, which is impor- Z Z  
a T
tant to evaluate biomass feedstock for fuel or chemical production, 1 A E A
gðaÞ ¼ da ¼ exp  dT ¼ wðE; TÞ ð3Þ
reactor design and control of thermo-chemical processes. In this 0 f ðaÞ b T0 RT b
study, the pyrolysis behaviors and kinetics of Eucalyptus residues,
where g(a) is the integral form of f(a); w(E,T) is the temperature
Eucalyptus leaves (EL), Eucalyptus bark (EB) and Eucalyptus sawdust
integral which has not an analytic solution but always can be
(ESD), were investigated by using thermogravimetric analysis. A
replaced with approximable expressions. A high accuracy equation
new modified discretize DAEM method which based on Starink
(Eq. (4)) to approximate the w(E,T) was given by Starink [21].
temperature integral [21] and Scott algorithm [19] was carried
 1:92  
out for the corresponding kinetics studies. As for a comparative E TR E
study, the Gaussian DAEM was also used in this study. wðE; TÞ ¼ exp 1:0008  0:312 ð4Þ
R E RT

2. Methods According to first-order reaction, f(a) = 1  a, and Eq. (3)

becomes
2.1. Material  
A
x ¼ 1  a ¼ exp  wðE; TÞ ð5Þ
b
The used Eucalyptus residues Eucalyptus leaves (EL), Eucalyptus
bark (EB) and Eucalyptus sawdust (ESD) were obtained from a saw- where x denotes the remaining conversion.
mill which locates on Guiping City, Guangxi province, China. The
fresh EL, EB and ESD were firstly dried under the sun for 30 days 2.3.2. Gaussian DAEM
to gain the air-dried basis (ADB) samples. ADB samples were fur- The distributed activation energy model (DAEM) assumes infi-
ther crushed, milled and sieved to achieve a particle size below nite irreversible first-order reactions happen during the whole
0.074 mm. The undersize ADB samples were used to perform prox- pyrolysis process [11,12,23]. All such reactions share a constant
imate analysis. The undersize ADB samples were then dried at pre-exponential factor (A) and the activation energy (E) is further
105 °C for 48 h to get dried basis (DB) samples. The DB samples assumed as a continuous distribution [23]. Then the conversion
were used to conduct ultimate and TGA analysis. function of DAEM during the whole temperature range is given as
Z 1  
A
2.2. Physicochemical characterization and thermogravimetric analysis x¼1a¼ exp  wðE; TÞ f ðEÞdE ð6Þ
0 b
The C, H, N, S and P elements weight contents of EL, EB and ESD where f(E) is the probability density function (PDF) of the activation
were analyzed using an elemental analyser (Vario Micro cube, energy distribution. Several PDFs such as Gaussian, Weibull, and
Elementar). The weight percent of oxygen was determined by Gamma etc. were reported in the literature [12]. Among these
difference. PDFs, the Gaussian PDF according to Eq. (7) is the most widely used.
Poximate analysis of EL, EB and ESD were performed according !
to ASTM standards [22]. And the low heating values (LHV) of the 1 ðE  E0 Þ2
f ðEÞ ¼ pffiffiffiffiffiffiffi exp  ð7Þ
samples (dried basis) were measured by using a bomb calorimeter. r 2p 2r2
Thermogravimetric analysis (TGA) tests were conducted on a
Pyris1 TGA instrument (Perkin Elmer Co., Ltd) with platinum cru- where E0 and r of Eq. (7) is the mean value and standard deviation
cibles. Approximately 5.0 mg of each sample was used for all of Gaussian PDF, respectively. Then unknown kinetic parameters
TGA runs. Argon was used as sweeping gas and maintained at a (i.e. E0, r and A) of Gaussian DAEM for Eucalyptus residues pyrolysis
constant flow rate of 100 ml/min. Heating rates of 10 °C/min and can be estimated by using optimization algorithms to fit the exper-
40 °C/min were used for EL, EB and ESD samples from room tem- imental data. In this study, the minimizing of the following objec-
perature to 800 °C. Each TGA experiment was repeated three times tive function (Eq. (8)) is used to perform the optimization.
to eliminate the errors. rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X2 XNd .
2
Fitð%Þ ¼ 100 j¼1 i¼1
ðx c  xe Þ Np ð8Þ
2.3. Kinetics
where j (j = 1, 2) denotes j-th heating rate TGA set; i denotes the
2.3.1. Theory of distributed activation energy model data points used; Nd denotes the number of experimental data; Np
Reaction rate of solid-state pyrolysis in linear non-isothermal denotes the number of unknown parameters; xc represents the cal-
TGA could be interpreted as the homogeneous kinetic combined culated x from Eq. (6); xe represents the experimental x; Fit(%) is
 Z. Chen et al. / Energy Conversion and Management 105 (2015) 251–259 253

fitting quality parameter between calculated and experimental x. The activation energy Ek can be calculated by solving Eq. (14).
And a lower Fit value indicates a better quality of fit. According to Eq. (5), the pre-exponential factor Ak can be calculated
by solving following equations (Eqs. (15) and (16)) with the known
2.3.3. Discrete DAEM Ek:
According to Scott et al. [19], the infinite reactions can be dis-
xjb1 ¼ xjb1 ð15Þ
cretized as n parallel first-order reactions. Each of these reactions
has its characteristic activation energy and pre-exponential factor. i.e.
For the pyrolysis process of Eucalyptus residues under "  1:92  #

non-isothermal condition, the reaction rate of one such Ak Ek T k R Ek

exp  exp 1:0008  0:312

first-order reaction can be given as b1 R Ek RT k

b1
  "  1:92  #

dmk Ak Ek Ak E k T k R Ek
 ¼ exp mk ð9Þ

dT bj RT  exp  exp 1:0008  0:312

ð16Þ
b1 R Ek RT k

b2
where mk, Ak and Ek is the instantaneous mass, pre-exponential fac-
tor and activation energy of k-th reaction, respectively. The integral Once the Ek and Ak are known, Eq. (12) can be solved by using
form of Eq. (9) is given as: Nonnegative Linear Least Square Method (NLLSM) to minimize
" # an objective function (Eq. (17)) and the initial fraction fk,0 can be
Ak estimated.
M k ¼ M k;0 exp  wðEk ; TÞ ð10Þ
bj X
n
O:F ¼ jxc  xe j2 ð17Þ
where Mk is the mass of k-th reaction and Mk,0 is the initial value of k¼1
Mk. The mass of sample is the sum of mass of n reactions. And then
The calculation of NLLSM must guarantee the constraints of:
the conversion function during pyrolysis process can be expressed
8
as: > f 2 ½1
< k;0
Pn " # Xn
Xn Xn ð18Þ
k¼1 M k Ak >
: f i;0 ¼ 1
x ¼ Pn  wchar;k ¼ f k;0 exp  wðEk ; T k Þ ð11Þ
k¼1 M k;0 i¼1 k¼1
bj k¼1

where fk,0 = Mk,0 and fk = Mk; n is number of first-order reactions; bj

denotes the j-th (j = 1,2) heating rate; Tk is the corresponding tem- 2.3.4. Calculation details
perature; and wchar,k is char mass of k-th reaction. If the value of Ek All the calculation of DAEM kinetic parameters were performed
and Ak were known, Eq. (11) would be solved linearly by using Eq. on MATLABÒ software (version R2012a). The correlation coeffi-
(4), as: cients (R2) between calculated date and experimental data were
2 3 obtained using corrcoef command of MATLABÒ.
 Ab0 w1 ðE0 ; T 0 Þ  Ab0 w2 ðE0 ; T 0 Þ . . .  Ab0 wn ðE0 ; T 0 Þ 1
6 A1 7
6  w1 ðE1 ; T 1 Þ  Ab1 w2 ðE1 ; T 1 Þ . . .  Ab1 wn ðE1 ; T 1 Þ 1 7 3. Results and discussion
6 b 7
6 7
6: : ... : : 7
x ¼ exp 6 7 3.1. Characterization of samples
6: : ... : : 7
6 7
6 7
4: : ... : : 5 The proximate and ultimate analysis results of Eucalyptus leaves
An An An
 b w1 ðEn ; T n Þ  b w2 ðEn ; T n Þ . . .  b wn ðEn ; T n Þ 1 (EL), Eucalyptus bark (EB) and Eucalyptus sawdust (ESD) are listed
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} in Table 1. The three samples are rich in carbon and oxygen con-
WðA;E;TÞ
2 3 tents (41.82–52.76%). Except of the EB, the carbon contents of EL
f 1;0
6f 7 Table 1
6 2;0 7 Ultimate, proximate analyses results of Eucalyptus waste.
6 7
6 f 3;0 7
6
6 7 ð12Þ Ultimate analysesa (wt. %)
: 7
6 7
6 7 C H N S P Oc
4: 5
Pn EL 47.71 4.85 3.49 0.45 0.13 43.37
i¼0 wchar;k EB 41.82 5.01 0.28 0.07 0.06 52.76
|fflfflfflfflfflfflfflfflfflfflffl
ffl{zfflfflfflfflfflfflfflfflfflfflfflffl}
ESD 49.75 5.77 0.17 0.03 0.02 44.26
f–
Proximate analysesb (wt. %)
i.e. M(T) = exp[W(A,E,T)]  f–, where M(T)–, W(A,E,T) and f– are in
–

matrix forms. MC VMC FC AC LHV(MJ/kg)

The same x is thought to have the same Ek and Ak under two dif- EL 4.26 77.73 12.12 5.89 14.98
ferent heating rates (such b1 and b2) TGA runs. And one character- EB 10.89 69.89 14.85 4.37 15.04
istic x represents one first-order reaction. Then, from w(E,T) ESD 6.12 74.38 18.79 0.81 17.64

expression in Eq. (4) one can get: Chemical componentsa (wt. %)

wk ðEk ; T 1 Þjb1 ¼ wk ðEk ; T 2 Þjb2 ð13Þ Hemicellulose Cellulose Lignin

EL 11.28 17.93 9.25
i.e. EB 12.90 27.48 32.09
 1:92  

ESD 24.74 33.89 20.77

Ek T k R Ek

 exp 1:0008  0:312

MC, moisture content (ASTM D 3173-87); VMC, volatile matter content (ASTM D
R Ek RT k
3175-89); AC, ash content (ASTM D 3174-89); FC, fixed carbon content (100-MC-
b1
 1:92  

VMC-AC); LHV, low heating value.

Ek T k R Ek
a
Dried basis (DB).
¼  exp 1:0008  0:312
ð14Þ b
R Ek RT k
Air-dried basis (ADB).
b2 c
Calculated by difference.
254 Z. Chen et al. / Energy Conversion and Management 105 (2015) 251–259

and ESD are higher than oxygen contents. The nitrogen (0.17– 0.7
3.49%) and sulfur (0.03–0.45%) contents of EL, EB and ESD samples 100 Solid line:
1-TG of EL 0.6
are lower than some microalgae biomasses which are usually used 90
3
(A) 6 2-TG of EB
via pyrolysis: Nannochloropis oculata (N: 5.02% and S: 0.64%) [24], 2
3- TG of ESD 0.5
Chlorella pyrenoidosa (N: 8.39% and S: 1.76%) [17], bloom-forming 80

DTG (−%⋅oC-1)
Dash line
Cyanobacteria (N: 8.41% and S: 0.84%) [17]. It can be even more 4- DTG of EL 0.4

TG (%)
70
confirmed that the nitrogen and sulfur percentages of EL are quite 5 5- DTG of EP
Stage 1 6- DTG of ESD
higher than that of EB and ESD. However, the N content of EL 60 0.3
(3.49%) is higher than some lignocellulosic biomass: alfalfa
50
(2.39%) [25], wheat straw (0.38%) [25], giant reedgrass (2.24%) Stage 3 0.2
1
[26], cotton cocoon shell (1.3%) [27], cotton stalk (1.12%) [28], sug- 40
arcane bagasse (0.15%) [28]. And the sulfur content of EL (0.45%) is 0.1
30 4 Stage 2
also higher than some other lignocellulosic biomass, such as palm
0.0
kernel shell (0.38%) [29], sugarcane bagasse (0.24%) [30], cotton 0 100 200 300 400 500 600 700 800
stalk (0.17%) [30]. In addition, the phosphorus contents of these Temperatue (oC)
three feedstocks are lower than N and S contents. Biomass with
low nitrogen and sulfur contents are attractive to 4.0 0.06
thermo-chemical conversion processes because high N and S con-
3.5 (B) 0.04
tents will lead to more toxic NOx and SO2 emission [31].
DDTG DDTG=0
As depicted in Table 1, the three feedstocks are in high volatile 3.0
0.02

DDTG (- %⋅min-1⋅oC-1)
contents (69.89–77.73%) that could be considered suitable for

DTG (- %⋅min-1)
0.00
pyrolysis, gasification or combustion processes [4,31]. The ash con- 2.5 H P S
tents of EL, EB and ESD are low, about 0.81–5.89%. Low ash content -0.02
2.0
features of EL, EB and ESD suggest that they can be regarded as -0.04
suitable feedstocks for thermo-chemical conversion processes, 1.5
-0.06
since lower ash content of feedstock takes the advantages to lower
1.0
fouling or aggregation on reactors. Furthermore, processing costs, DTG -0.08
Tf
poor combustion, disposal problems and low energy conversion 0.5 Ti L -0.10
are resulted from high ash content [32]. Biomass conversion tech- J Tp
nologies are also influenced by the moisture content of the feed- 0.0 -0.12
0 100 200 300 400 500 600 700 800
stock. The moisture content of EL (4.26%) and ESD (6.12%) are
higher than that of EB (10.89%) and some reported biomass feed- Temperature (oC)
stocks, such as cotton stalk (9.13%) [30], wheat husk (13.90%)
Fig. 1. (A) TG/DTG curves of Eucalyptus leaves (EL), Eucalyptus bark (EB) and
[33], microalgae bloom-forming Cyanobacteria (9.59%) [17], Eucalyptus sawdust (ESD) at heating rate of 10 °C/min. (B) DTG/DDTG curves of
microalgae Chlorella vulgaris (9.10%) [34]. Low moisture content Eucalyptus leaves pyrolysis at of 10 °C /min heating rate.
is favorable to thermal conversion technique [32].

To obtain the characteristic temperatures of stage 2, the curve

3.2. Thermal behaviors of samples
of differential DTG (DDTG) is used in this study. The function of dif-
ferential DTG, i.e. second derivative of TG, is given as:
The TG/DTG curves obtained at a heating rate of 10 °C/min of EL,
EB and ESD are compared in Fig. 1(A). Similar curves at heating rate 2
d mu
of 40 °C/min were obtained for all samples. The DTG curves show f ðmu ; TÞDDTG ¼  2
ð19Þ
dT
that the pyrolysis process of EL, EB and ESD be further divided into
three stages. As shown in Fig. 1(A), stage 1 takes place in low tem- where mu is the mass of samples, %. The peaks or valleys of DTG
perature region and results in low mass loss (TG curves). It is curves represent the fastest or slowest mass change rate. And then
believed to be attributed to the evaporation of physically adsorbed the variation rate of DTG v.s. temperature (i.e. DDTG curve), will be
water and light volatiles [17]. Stage 2, which covers a wide temper- obtained as f (mu, T)DDTG = 0 at the peaks and valleys. DTG/DDTG
ature range and manifests high mass loss percentage, is caused by curves of EL pyrolysis at heating rate of 10 °C/min are shown in
the devolatilization. The pyrolysis behavior in stage 3 is character- Fig. 1(B). As can be seen, point H of DDTG curve corresponds to
ized as little mass is probably attributed to the decomposition of point J of DTG curve. And the corresponding temperature Ti is
carbonaceous materials retained in char residues. However, the obtained and denotes the initial temperature of stage 2. Point S of
thermal behaviors of EL, EB and ESD of stage 2 are complex due DDTG curve corresponds to point L of DTG curve corresponds the
to the overlapping peaks and corresponding shoulders. final of stage 2 its final temperature, Tf, is obtained. Similar

Table 2
Initial, peaks and final temperatures and weigh loss percent of pyrolysis Stage 2 for EL, EB and ESD.

Biomass HR (°C/min) Ti (°C) Tp (°C) Tf (°C) WLP (%)

Peak 1 Peak 2 Peak 3
EL 10 143.8 277.8 373.6 / 546.3 86.93
40 161.4 297.4 384.8 / 578.5 87.35
EB 10 174.3 295.1 338.6 490.6 582.7 88.96
40 203.3 323.5 364.7 522.5 601.9 89.48
ESD 10 141.5 / 344.5 / 573.5 95.20
40 173.1 / 355.5 / 639.2 97.84

HR is the heating rate; Ti is the initial temperature; Tp is the peak temperature; Tf is the final temperature; WLP is the weight loss percentage.
 Z. Chen et al. / Energy Conversion and Management 105 (2015) 251–259 255

100 0.7 3.3. Effect of heating rate

(A) 1: TG-10oC/min 0.6 As shown in Table 2 and Fig. 2, the EL, EB and ESD pyrolysis
80 2: TG-40oC/min characteristic temperatures of stage 2 increases with the heating
3: DTG-10oC/min 0.5
2 rate. It is obvious that the increasing of heating rate shifts the char-

DTG (−% ⋅ oC-1)

4: DTG-40oC/min
60 acteristic temperature to higher value but do not change DTG pro-
TG (%)

peak 1 0.4
file (according to Fig. 2). The effect of heating rate on the
EL 1
peak 2
4 0.3 characteristic temperature mainly attributes to the limited rate
40 of heat conduction into the sample due to the thermal resistance
3 0.2 of biomass. The increasing of heating rate leads to a simultaneous
20 decrease of the effect temperature and an increase of heat effect
0.1 [35]. Increasing of heating rate signify higher temperature is
required to set off the decomposition process. One the other hand,
0 0.0 the pyrolysis reaction can affect the heat transfer in particle. In
100 200 300 400 500 600
fact, the release of volatiles leads to the formation of char which
Temperature (oC) is in poor heat transfer performance and prevent the heat transfer
to particle core to some extent. Moreover, the thermal decomposi-
100 0.8 tion reaction of biomass and the secondary reaction between the
1: TG-10 oC/min 0.7 volatiles and char are endothermic processes. Such endothermic
80 (B) 2: TG-40 oC/min reactions absorb the heat that transfers along the radial direction
o
3: DTG-10 C/min 0.6 DTG ( − % ⋅oC-1) before transfer to the core.
4: DTG-40oC/min
60 0.5
2 peak 1
TG (%)

peak 2 3.4. Gaussian DAEM kinetics

0.4
1 peak 3
40 EB 4
3 0.3 To perform the calculation of Gaussian DAEM, the Lobatto
Quadrature method (i.e. quadl solver that built-in MATLABÒ) was
0.2
20 employed to deal with the improper integral (dE integral from 0
0.1 to infinity) in Eq. (6). The effect limit of dE integral referenced to
literature [15] with a range of 0 – E0 + 30r. For the fitting program,
0 0.0
100 points of experimental data (xe = 0.999–0.001) was obtained
200 300 400 500 600
using linspace command and the corresponding temperature Tx
Temperature (oC)
was obtained by using cubic spline interpolation method (interp1
and spline commands). The kinetics of EL, EB and ESD for
100
Gaussian DAEM were investigated by minimizing objective func-
2 1: TG-10 oC/min 1.0
(C) tion of Eq. (8) through patternsearch algorithm (pattrensearch
2: TG-40 oC/min
80 Global optimization tool built-in MATLABÒ). E0 = 125 kJ/mol,
3: DTG-10 oC/min
4: DTG- 40 oC/min
0.8 A = 1E+3 s1 and r = 10 kJ/mol were used as the initial guess values
DTG ( −% ⋅ oC-1)

for kinetic parameters referenced to the literature [23]. The trust

60 peak 2
TG (%)

0.6 region for the parameters are E0 2 (0, 350 kJ/mol), r 2 (0,
50 kJ/mol) and A 2 (0, 10E+20 s1), which were similar to literature
3 1
40 ESD [13].
0.4
The optimized kinetic parameters and final objective function
4
values of Gaussian DAEM for EL, EB and ESD pyrolysis are listed
20 0.2 Table 3. Gaussian DAEM fits to the experimental data of the three
biomass feedstocks are shown in Fig. 3. The fitting qualities are
0 0.0 good with low Fit (Table 3) and high correlation coefficients (R2).
100 200 300 400 500 600 According to Gaussian DAEM, parameter E0 represents the dis-
Temperature (oC) tributed center of activation energy and parameter r represents
the distributed width. It is obvious that activation energy of ESD
Fig. 2. TG/DTG curves of Eucalyptus leaves (EL), Eucalyptus bark (EB) and Eucalyptus pyrolysis has a narrower (r = 14.41 kJ/mol) distribution than that
sawdust (ESD) pyrolysis of Stage 2 at 10 and 40 °C/min heating rates: (A) for EL, (B)
for EB and (C) ESD.
of EL (r = 18.35 kJ/mol) and EB (r = 18.37 kJ/mol). The distribu-
tions of EL and EB are similar to each other. However, the distribu-
tion center (E0) of ESD is higher (E0 = 175.79 kJ/mol) than that of EL
(E0 = 141.15 kJ/mol) and EB (E0 = 149.21 kJ/mol). The mean activa-
tion energy of woody ESD is similar to some kinds of lignocellulosic
processes have been done for the EB and ESD. And the characteristic
biomass such as rice straw (170 kJ/mol), rice husk (174 kJ/mol)
temperatures together with the weight loss percentage (WLP) of EL,
[36]. The activation energy denotes the energy barrier which
EB and ESD are listed in Table 2. It is obvious that stage 2 is the main
pyrolysis stage with high WLP of 86.93–97.84%. The TG/DTG curves
Table 3
of stage 2 for all samples at heating rates of 10 °C/min and Gaussian parameters and fitting qualities of continuous DAEM simulation for
40 °C/min are exhibited in Fig. 2. Fig. 2(A) shows that there are Eucalyptus residues pyrolysis.
two peaks locates on the DTG curve of EL. However, three peaks
E0 (kJ/mol) A (s1) r (kJ/mol) Fit (%)
and one peak are observed for EB and ESD, respectively. Peak 2
which appears at temperature range of 338.6–384.8 °C is the main EL 141.15 1.15E+10 18.35 4.61
EB 149.21 4.34 E+10 18.37 3.38
peak of the DTG curves and it is believed to be mainly contributed ESD 175.79 7.44E+12 14.41 3.29
to the decomposition of cellulose.
256 Z. Chen et al. / Energy Conversion and Management 105 (2015) 251–259

1.0 biomass. Thus, the Gaussian DAEM kinetic parameters of EL, EB

0.025
and ESD pyrolysis should not be compared.
0.020
(A)

f (E) (mol/kJ)
0.8 0.015 3.5. Discrete DAEM kinetics
0.010

0.005 3.5.1. Kinetics parameters

0.6
0.000
According to discrete DAEM, each characteristic remaining con-
80 100 120 140 160 180 200 220 240
x=1-α

Experiment: version (x) represents an irreversible first-order reaction, which

E (kJ/mol) has a characteristic activation energy (Ex) and pre-exponential fac-
0.4 10 oC/min
40 oC/min tor (Ax). Furthermore, such reaction represents a pseudo compo-
Simulation: nent which contains in biomass. These components react through
EL
0.2 10 oC/min parallel reaction regime [18,19]. At the same time, a characteristic
40 oC/min remaining conversion at different heating rate has the same activa-
tion energy. By choosing a batch of x in spite of at different heating
0.0 rate, the Ex and Ax can be estimated from Eq. (14) and Eq. (16)
200 300 400 500 600 respectively. And then the initial contributed fraction (fk,0) of each
Temperature (oC) hypothetical pseudo component can be estimated from Eq. (12) by
minimizing the objective function of Eq. (17).
1.0 0.025 100 characteristic remaining conversion points of experimental
data (xe = 0.999–0.001) have been chosen the same with Gaussian
(B) 0.020
DAEM. The fzero solver built-in MATLABÒ has been used to solve
f (E) (mol/kJ)

0.015
0.8 Eq. (14) and Eq. (16) to obtain the Ex and Ax for EL, EB and ESD pyrol-
0.010 ysis. fk,0 have been calculated from Eq. (12) and Eq. (17) using lsqnon-
0.005 neg optimization tool built-in MATLABÒ. The variations of activation
0.6 energies, pre-exponential factors (in ln(Aa) form) and fk,0 along with
0.000
80 120 160 200 240 conversion for EL, EB and ESD pyrolysis are shown in Fig. 4.
x=1-α

Experiment: E (kJ/mol)
10 oC/min The variations of activation energy versus conversion (a) are
0.4
40 oC/min complex. The activation energy of EB almost increases with the
Simulation:
EB increasing of conversion. But a downtrend can be seen for EL and
0.2 10 oC/min ESD at conversion range of 0.87–0.99 and 0.79–0.99, respectively.
40 oC/min Platforms with very little activation energy changes also can be
seen in dependencies of Ea v.s. a for EL, EB and ESD pyrolysis.
0.0 The platform of EL is narrow at about 0.45–0.60 conversion range.
200 300 400 500 600 However, the platforms for the EB and ESD are wider than EL,
Temperature (oC) about range of 0.06–0.78 for EB and 0.06–0.61 for ESD. Because
the discrete DAEM used in this study assumes the Ex is the same
at two different heating rates. It is similar to the iso-conversional
1.0 0.030
methods [12,34]. As indicated by [12,37] and [12], the first-order
(C) 0.025 reaction model can be seen as an alternative for different mecha-
nism model thought it is simple. The Ea v.s. a dependencies can
f (E) (mol/kJ)

0.8 0.020
0.015 used to identify the change of decomposition mechanisms. In fact,
the nuclei, diffusion and catalysis effects should be account for the
0.010
0.6 activation energy variation during solid-state thermal decomposi-
0.005
Experiment: tion [12]. The platform and downtrend of Ea v.s. a dependencies in
x=1-α

10 oC/min 0.000 EL, EB and ESD pyrolysis illustrate the continuous change of
120 150 180 210 240
0.4 40 oC/min
E (kJ/mol) decomposition mechanisms during their pyrolysis process. [38]
have reported that under low linear non-isothermal heating rate
Simulation: conditions (10 °C/min and 40 °C/min in the present study), the
0.2 10 oC/min ESD decomposition of solid-state was limited by nuclei growth which
40 oC/min
leads to a rising trend in activation energies. However, the diffu-
0.0 sion effect leads to a downtrend for activation energies. It is
believed that more porous char will be formed with progresses
200 300 400 500 600
of the biomass pyrolysis [39]. This porous char enhance the diffu-
Temperature (oC) sion of the volatiles and then reduce the activation energy. In case
of non-isothermal decomposition, the two opposite effects (nuclei
Fig. 3. Gaussian distributions (sub-graphs) and the comparisons between simula-
tions and experiments: (A) Eucalyptus leaves (EL), (B) Eucalyptus bark (EB) and (C) growth and gas diffusion) may finally make the activation energy
Eucalyptus sawdust (ESD). as a weak function of conversion, which also leads to downtrend
of activation energy. Thus, the platform of Ea v.s. a dependencies
may attribute to the weak balances between nuclei growth and
pyrolysis reaction has to be overcome. And it seems that the pyrol- gas diffusion effects. Additionally, the in-situ catalysis of alkalis
ysis activities of EL, EB and ESD follow an order of EL > EB > ESD. metals also leads to the decrease of activation energy especially
However, the pyrolysis activities between EL, EB and ESD cannot in high temperature or high conversion [40,41].
be compared directly because the activation energy also depends The values of ln (Aa) were not constant for all reactions but have
on the temperature. As shown in Table 2, pyrolysis temperature ranges of 31.51–52.50 (EB), 28.28–59.07 (EB) and 4.15–26.80
ranges for EL, EB and ESD are different from to each other. (ESD), i.e. the pre-exponential factors have wide ranges of
Furthermore, the activation energy mainly depends on species of 4.84E+13–6.12E+22 s1 (EL), 1.91E+12–4.51E+25 s1 (EB) and
 Z. Chen et al. / Energy Conversion and Management 105 (2015) 251–259 257

55 0.14

300
(A)
0.12
50
ln (Aα)
0.10
250
Eα
45

Eα (kJ/mol)
0.08

ln (Aα )

fk , 0
200 40
0.06

fk ,0 35 0.04
150
0.02
30

100 0.00
0.0 0.2 0.4 0.6 0.8 1.0
α

450
(B) 60
0.20
400
55
350
Eα
fk,0 50 0.15
ln (Aα )
Eα (kJ/mol)

300

ln (Aα)
45

f k, 0
250 0.10
40
200
35
150 0.05
30
100
25 0.00
0.0 0.2 0.4 0.6 0.8 1.0
α

200 40 0.6
(C)
35
0.5
180
30
0.4
25
Eα (kJ/mol)

160
ln(Aα )

Eα f k,0
f k,0

20 0.3
ln(Aα )
140
15
0.2
10
120
0.1
5

100 0 0.0
0.0 0.2 0.4 0.6 0.8 1.0
α

Fig. 4. Discrete DAEM kinetic parameters (activation energy, pre-exponential factor (lnAa) and initial mass fraction of Eucalyptus waste pyrolysis: (A) Eucalyptus leaves; (B)
Eucalyptus bark and (C) Eucalyptus sawdust.

63.43–4.36E+11 s1 (ESD). The ln(Aa) v.s. a dependencies for EL, EB 3.5.2. Discrete DAEM simulation
and ESD are very similar to that of Ea v.s. a. It may be attributed to The comparsion of DAEM simulation and experiments of
the so-called energy compensation effect [12,42]. EL, EB and ESD pyrolysis are shows in Fig. 5. As can be seen,
Although 100 reactions are assumed for the discrete DAEM, excellent fits have been obtained by using discrete DAEM
actually not all the reactions show effective contribution to the with even higher correlation coefficient (R2) than Gaussian
pyrolysis process. As can be seen from Fig. 4, some fk,0 value is zero DAEM (as listed in Table 4). It indicates that the discrete
but some are not zero. The corresponding reactions of fk,0 – 0 can DAEM is better than single Gaussian DAEM for pyrolysis
be seen as the effective contribution to the pyrolysis [17]. studies.
258 Z. Chen et al. / Energy Conversion and Management 105 (2015) 251–259

1.0 4. Conclusion
Experiment:
10 oC/min The pyrolysis behaviors of Eucalyptus leaves (EL), Eucalyptus
0.8 40 oC/min bark (EB) and Eucalyptus sawdust (ESD) are different from each
Simulation: other. The pyrolysis processes can be divided into three stages
0.6 10 oC/min where Stage 2 contributes to the main pyrolysis part with 86.93–
40 oC/min 97.84% weight loss percentages. Pyrolytic temperature ranges of
1-α

EL EL, EB and ESD is 143.8–578.5 °C, 174.3–601.9 °C and 141.5–

0.4
639.2 °C, respectively. The pyrolytic kinetics of Gaussian DAEM
showed good fits to experimental data. And the discrete DAEM
0.2 with 100 hypothetical first-order reactions is even better than
Gaussian DAEM for biomass kinetic studies. For biomass pyrolysis,
(A) the variation of activation energy versus conversion reveals the
0.0
reaction mechanism changes.
100 200 300 400 500 600
Temperature ( oC) Acknowledgements

1.0
This research was supported by the National High Technology
Experiment: Research and Development Program (863 Program) of China (No.
10 oC/min 2012AA101809), National Natural Science Foundation of China
0.8 (No. 21276100) and class General Financial Grant from the China
40 oC/min
Postdoctoral Science Foundation (Grant No.: 2014M560892). The
Calculation:
first author wants to acknowledge the Analytical and Testing
0.6 10 oC/min
Center of Huazhong University of Science and Technology for car-
40 oC/min
1-α

rying out the analysis of feedstock samples.

0.4 EB
References
0.2
[1] Mohan Jr D. Pittman CU, Steele PH. Pyrolysis of wood/biomass for bio-oil: a
critical review. Energy Fuel 2006;20(3):848–89.
0.0 (B) [2] Zhang Q, Zhou D, Fang X. Analysis on the policies of biomass power generation
in China. Renew Sust Energ Rev 2014;32:926–35.
100 200 300 400 500 600 [3] Zhu LW, Zhao P, Wang Q, Ni GY, Niu JF, Zhao XH, et al. Stomatal and hydraulic
conductance and water use in a eucalypt plantation in Guangxi, southern
Temperature (oC) China. Agr Forest Meteorol 2015;202:61–8.
[4] Ceylan S, Topcu Y. Pyrolysis kinetics of hazelnut husk using thermogravimetric
analysis. Bioresource Technol 2014;156:182–8.
1.0 [5] Bridgwater AV, Meier D, Radlein D. An overview of fast pyrolysis of biomass.
Simulation:
Org Geochem 1999;30(12):1479–93.
10 oC/min [6] Di Blasi C. Modeling chemical and physical processes of wood and biomass
0.8 40 oC/min pyrolysis. Prog Energy Combust 2008;34(1):47–90.
[7] White JE, Catallo WJ, Legendre BL. Biomass pyrolysis kinetics: a comparative
Experiment: critical review with relevant agricultural residue case studies. J Anal Appl Pyrol
10 oC/min) 2011;91(1):1–33.
0.6 [8] Jayaraman K, Goekalp I. Pyrolysis, combustion and gasification characteristics
40 oC/min
of miscanthus and sewage sludge. Energy Convers Manage 2015;89:83–91.
1-α

[9] Ma Z, Chen D, Gu J, Bao B, Zhang Q. Determination of pyrolysis characteristics

0.4 ESD and kinetics of palm kernel shell using TGA-FTIR and model-free integral
methods. Energy Convers Manage 2015;89:251–9.
[10] Lin Y, Ma X, Ning X, Yu Z. TGA FTIR analysis of co-combustion characteristics of
0.2 paper sludge and oil-palm solid wastes. Energy Convers Manage
2015;89:727–34.
(C) [11] Burnham AK, Braun RL. Global kinetic analysis of complex materials. Energy
Fuel 1999;13(1):1–22.
0.0
[12] Vyazovkin S, Burnham AK, Criado JM, Perez-Maqueda LA, Popescu C,
Sbirrazzuoli N. ICTAC Kinetics Committee recommendations for performing
100 200 300 400 500 600 700
kinetic computations on thermal analysis data. Thermochim Acta 2011;520(1–
Temperature ( oC) 2):1–19.
[13] Cai J, Wu W, Liu R, Huber GW. A distributed activation energy model for the
Fig. 5. Comparisons between the discrete DAEM simulated data and experimental pyrolysis of lignocellulosic biomass. Green Chem 2013;15(5):1331–40.
data for Eucalyptus waste pyrolysis: (A) Eucalyptus leaves (EL); (B) Eucalyptus bark [14] Varhegyi G, Chen H, Godoy S. Thermal decomposition of wheat, oat, barley,
(EB) and (C) Eucalyptus sawdust (ESD). and brassica carinata straws. A kinetic study. Energ Fuel 2009;23(1):646–52.
[15] Cai J, Wu W, Liu R. Sensitivity analysis of three-parallel-DAEM-reaction model
for describing rice straw pyrolysis. Bioresource Technol 2013;132:423–6.
Table 4 [16] Soria-Verdugo A, Garcia-Gutierrez LM, Blanco-Cano L, Garcia-Hernando N,
Correlation coefficients (R2) between the experimental data and calculated data. Ruiz-Rivas U. Evaluating the accuracy of the distributed activation energy
model for biomass devolatilization curves obtained at high heating rates.
Biomass HR Gaussian DAEM Discrete DAEM Energy Convers Manage 2014;86:1045–9.
[17] Hu M, Chen Z, Guo D, Liu C, Xiao B, Hu Z, et al. Thermogravimetric study on
EL 10 0.9892 0.9987 pyrolysis kinetics of Chlorella pyrenoidosa and bloom-forming cyanobacteria.
40 0.9891 0.9976 Bioresource Technol 2015;177:41–50.
EB 10 0.9907 0.9994 [18] Caballero JA, Conesa JA. New approach to thermal analysis kinetics by
40 0.9911 0.9987 considering several first order reactions. Thermochim Acta 2011;525(1–
ESD 10 0.9886 0.9991 2):40–9.
40 0.9889 0.9992 [19] Scott SA, Dennis JS, Davidson JF, Hayhurst AN. An algorithm for determining
the kinetics of devolatilisation of complex solid fuels from thermogravimetric
HR, heating rate (°C/min). experiments. Chem Eng Sci 2006;61(8):2339–48.
 Z. Chen et al. / Energy Conversion and Management 105 (2015) 251–259 259

[20] Cao H, Xin Y, Wang D, Yuan Q. Pyrolysis characteristics of cattle manures using [31] Damartzis T, Vamvuka D, Sfakiotakis S, Zabaniotou A. Thermal degradation
a discrete distributed activation energy model. Bioresource Technol studies and kinetic modeling of cardoon (Cynara cardunculus) pyrolysis using
2014;172:219–25. thermogravimetric analysis (TGA). Bioresource Technol 2011;102(10):6230–8.
[21] Starink MJ. The determination of activation energy from linear heating rate [32] McKendry P. Energy production from biomass (part 1): overview of biomass.
experiments: a comparison of the accuracy of isoconversion methods. Bioresource Technol 2002;83(PII S0960-85240100118-31):37–46.
Thermochim Acta 2003;404(1–2):163–76. [33] Mythili R, Venkatachalam P, Subramanian P, Uma D. Characterization of
[22] Ahmaruzzaman M. Proximate analyses and predicting HHV of chars obtained bioresidues for biooil production through pyrolysis. Bioresource Technol
from cocracking of petroleum vacuum residue with coal, plastics and biomass. 2013;138:71–8.
Bioresource Technol 2008;99(11):5043–50. [34] Agrawal A, Chakraborty S. A kinetic study of pyrolysis and combustion of
[23] Cai J, Liu R. New distributed activation energy model: Numerical solution and microalgae Chlorella vulgaris using thermo-gravimetric analysis. Bioresource
application to pyrolysis kinetics of some types of biomass. Bioresource Technol Technol 2013;128:72–80.
2008;99(8):2795–9. [35] Sait HH, Hussain A, Salema AA, Ani FN. Pyrolysis and combustion kinetics of
[24] Ceylan S, Kazan D. Pyrolysis kinetics and thermal characteristics of microalgae date palm biomass using thermogravimetric analysis. Bioresource Technol
Nannochloropsis oculata and Tetraselmis sp.. Bioresource Technol 2015;187: 2012;118:382–9.
1–5. [36] Sonobe T, Worasuwannarak N. Kinetic analyses of biomass pyrolysis using the
[25] Sarker S, Arauzo J, Kofoed Nielsen H. Semi-continuous feeding and gasification distributed activation energy model. Fuel 2008;87(3):414–21.
of alfalfa and wheat straw pellets in a lab-scale fluidized bed reactor. Energy [37] Sanchez-Jimenez PE, Perez-Maqueda LA, Perejon A, Criado JM. Nanoclay
Convers Manage 2015;99:50–61. nucleation effect in the thermal stabilization of a polymer nanocomposite: a
[26] Guan Y, Ma Y, Zhang K, Chen H, Xu G, Liu W, et al. Co-pyrolysis behaviors of kinetic mechanism change. J Phys Chem C 2012;116(21):11797–807.
energy grass and lignite. Energy Convers Manage 2015;93:132–40. [38] Vyazovkin S, Wight CA. Kinetics of thermal decomposition of cubic ammonium
[27] Caglar A, Demirbas A. Conversion of cotton cocoon shell to hydrogen rich perchlorate. Chem Mater 1999;11(11):3386–93.
gaseous products by pyrolysis. Energy Convers Manage 2002;43(4):489–97. [39] Avila C, Pang CH, Wu T, Lester E. Morphology and reactivity characteristics of
[28] Munir S, Daood SS, Nimmo W, Cunliffe AM, Gibbs BM. Thermal analysis and char biomass particles. Bioresource Technol 2011;102(8):5237–43.
devolatilization kinetics of cotton stalk, sugar cane bagasse and shea meal [40] Huber GW, Iborra S, Corma A. Synthesis of transportation fuels from biomass:
under nitrogen and air atmospheres. Bioresource Technol 2009;100(3): chemistry, catalysts, and engineering. Chem Rev 2006;106(9):4044–98.
1413–8. [41] Zabeti M, Nguyen TS, Lefferts L, Heeres HJ, Seshan K. In situ catalytic pyrolysis
[29] Asadieraghi M, Daud WMAW. Characterization of lignocellulosic biomass of lignocellulose using alkali-modified amorphous silica alumina. Bioresource
thermal degradation and physiochemical structure: effects of demineraliza- Technol 2012;118:374–81.
tion by diverse acid solutions. Energy Convers Manage 2014;82:71–82. [42] Lesnikovich AI, Levchik SV. A method of finding invariant values of kinetic
[30] El-Sayed SA, Mostafa ME. Pyrolysis characteristics and kinetic parameters parameters. J Therm Anal 1983;27(1):89–93.
determination of biomass fuel powders by differential thermal gravimetric
analysis (TGA/DTG). Energy Convers Manage 2014;85:165–72.