Accepted Manuscript

Pyrolysis, kinetics analysis, thermodynamics parameters and reaction mecha-

nism of Typha latifolia to evaluate its bioenergy potential

Muhammad Sajjad Ahmad, Muhammad Aamer Mehmood, Syed Taha Haider

Taqvi, Ali Elkamel, Chen-Guang Liu, Xujian Ren, Sawsan Abdulaziz
Rahimuddin, Munazza Gull

PII: S0960-8524(17)31476-1
DOI: http://dx.doi.org/10.1016/j.biortech.2017.08.162
Reference: BITE 18771

To appear in: Bioresource Technology

Received Date: 6 July 2017

Revised Date: 3 August 2017
Accepted Date: 27 August 2017

Please cite this article as: Ahmad, M.S., Mehmood, M.A., Taqvi, S.T.H., Elkamel, A., Liu, C-G., Ren, X.,
Rahimuddin, S.A., Gull, M., Pyrolysis, kinetics analysis, thermodynamics parameters and reaction mechanism of
Typha latifolia to evaluate its bioenergy potential, Bioresource Technology (2017), doi: http://dx.doi.org/10.1016/
j.biortech.2017.08.162

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
 Pyrolysis, kinetics analysis, thermodynamics parameters and reaction mechanism
of Typha latifolia to evaluate its bioenergy potential

Muhammad Sajjad Ahmad b,c, Muhammad Aamer Mehmood a,b*, Syed Taha Haider Taqvi c, Ali
Elkamel c, Chen-Guang Liu d*, Xujian Ren d, Sawsan Abdulaziz Rahimuddin e, Munazza Gull e

a
College of Bioengineering, Sichuan University of Science and Engineering, Zigong-643000,
People’s Republic of China
b
Bioenergy Research Center, Department of Bioinformatics and Biotechnology, Government

College University Faisalabad, Faisalabad-38000, Pakistan

c
Chemical Engineering Department, University of Waterloo, Ontario, Canada

d
State Key Laboratory of Microbial Metabolism, School of Life Sciences and Biotechnology,

Shanghai Jiao Tong University, Shanghai, China

e
Biochemistry Department, Faculty of Science, King Abdulaziz University, Jeddah, 21551,
Saudi Arabia

Corresponding authors: MA Mehmood ([email protected]); CG Liu ([email protected])

1
Abstract

This work was focused on understanding the pyrolysis of Typha latifolia. Kinetics,
thermodynamics parameters and pyrolysis reaction mechanism were studied using
thermogravimetric data. Based on activation energies and conversion points, two regions of
pyrolysis were established. Region-I occurred between the conversion rate 0.1 to 0.4 with peak
temperatures 538K, 555K, 556K at the heating rates of 10 Kmin-1, 30 Kmin-1, and 50 Kmin-1,
respectively. Similarly, the Region-II occurred between 0.4 to 0.8 with peak temperatures of
606K, 621K, 623K at same heating rates. The best model was diffusion mechanism in Region-I.
In Region-II, the reaction order was shown to be 2nd and 3rd. The values of activation energy
calculated using FWO and KAS methods (134-204 kJ mol-1) remained same in both regions
reflecting that the best reaction mechanism was predicted. Kinetics and thermodynamic
parameters including E, ∆H, ∆S, ∆G shown that T. latifolia biomass is a remarkable feedstock
for bioenergy.

Keywords: Typha latifolia; thermogravimetric study; reaction mechanism; bioenergy; low-cost

biomass

2
1. Introduction
At present, energy requirements of the world are met through fossil fuels. Owing to their
non-renewable nature, fossil fuel deposits may be completely exhausted from Earth in 70 years
from now. Alternatively, biomass is believed to be a reliable future energy source along with
solar, wind and hydrothermal. Among all the energy resources worldwide, biomass provides up
to 10 %, with an annual increase of 2.5% (Edrisi and Abhilash, 2016). In the USA and Brazil,
almost 80 percent of the fuel comes from renewables, specifically maize and sugarcane, which
are not only expensive but have also created a food versus fuel dilemma. Alternatively, non-
edible plants produced on non-arable lands offer a low-cost alternative without any direct or
indirect competition with food or land (Ahmad et al., 2017a). Here, the problem is not the
availability of the biomass, but rather the cost-effective and efficient retrieval of the energy
stored in the biomass. Several processes have been developed to retrieve the biomass energy
including direct combustion, thermochemical and biological conversion, where pyrolysis and
biological fermentation are the cleanest methods to convert the biomass into valuable products.
However, the latter is a tedious, expensive and time-consuming process, mainly due to the
recalcitrant nature of the biomass. While, the thermal transformation of biomass into various
products including solid, liquid and gasses often under an inert environment is called pyrolysis.
Moreover, pyrolysis process leaves almost no waste, and all converted components can be used
for one or another purpose ranging from energy (heat, bio-oil) through agricultural (char) and
industrial chemicals (gases). However, the pyrolysis process depends upon various factors
including nature of the biomass, particle size and temperature parameters. Hence, for an efficient
conversion of any biomass, it is essential to understand its pyrolysis behavior to design an
optimized pyrolytic process.

Thermogravimetric analyses performed under controlled conditions are practically

feasible to understand the pyrolytic behavior and to determine the optimum pyrolytic conditions
of any biomass (Di Blasi, 2009; Kow et al., 2014). The thermochemical conversion has
dominance over biological processes because it has higher efficiency (Zhang et al., 2006).
However, a clear understanding of the pyrolytic conditions is required for any biomass prior to
feeding this biomass to any commercial thermal plant. Artificial Neural Networks (ANN) are
3
models which are developed based on the functionality of a human brain in which neurons
transfer data from synapses on the way to axons through chemicals called neurotransmitters.
Similarly, ANNs are composed of multiple neuron layers including input layers, hidden layers,
and output layer to learn the relationship between them for further problem-solving approaches
between experimental and predicted data (Kalogirou, 2003). Consequently, ANNs have been
proven to be a useful tool to predict the reaction chemistry (Conesa et al., 2004; Yıldız et al.,
2016) and has been employed on thermal data for estimation from several standpoints through
training, validation and testing of experimental data (Chen et al., 2017; Uzun et al., 2017).

Previously, several grasses including Camel grass (Mehmood et al., 2017), Corn cobs,
Miscanthus, wheat straw (Álvarez et al., 2016) and Sorghum weeds (Rezende and Richardson,
2017), along with microalgae (Maurya et al., 2016), red-peppers waste (Maia and de Morais,
2016) and rice-husk (Zhang et al., 2016) were previously studied for their bioenergy potential
using thermogravimetric analyses. The Napier Grass is already being used to produce fuel on the
commercial thermal plant (He et al., 2017). The present study was focused on understanding the
pyrolytic behavior, reaction chemistry and bioenergy potential of Typha latifolia via
thermogravimetric analyses for the very first time. T. latifolia occurs across the globe including
Asia, Africa, Americas and Europe. It is a perennial grass produced on marginal lands and is a
famous wetland species and can be grown in brackish or polluted water, hence is a low-cost
biomass resource. While cultivating on polluted or brackish water, its bioremediatory action can
remove pollutants developing a bitter taste making it unsuitable for food or feed. It has higher
energy potential and can produce 25 units more energy when compared to fossil fuel (Ussiri and
Lal, 2017). To date, there is no study available on pyrolytic behavior using Artificial Neural
Network Approach of T. latifolia, studying the reaction mechanism of the thermal degradation.
Generally, the reaction order model is considered as a suitable reaction mechanism of the
biomass as it confirms its viability throughout the pyrolysis process. High precision
thermogravimetric and kinetic analyses of devolatilization process (Trninić et al., 2012) are also
applied in this study.

4
2. Materials and Methods

2.1 Elemental composition and proximate analyses

The biomass of T. latifolia was collected from soil affected by salinity which was being
irrigated with underground brackish water. Collected sample was washed under a tap and left to
dry in air for several days. Air dried biomass was crushed using a manual crusher and put into an
oven for 48 h, and was grounded to get particles of size ranging from 150-200 µm. Sun dried
crushed biomass was subjected to proximate analyses to determine volatile matter (VM %), ash
(%) moisture content (%) using the standard methods as described in ASTM (E872-&82 2006,
E871-82 2006, E1755-01 2007). The fixed carbon (FC %) was calculated using the equation:

 % = 100 −  ℎ  +  +  . To determine the VM and moisture
content, known mass was put in oven-dried at 380K in triplicate for 16-24 h to get a constant
mass. The loss in mass reflected the moisture content. Similarly, known mass from the oven
dried sample was put into pre-weighed ceramic crucibles in triplicate and left at 775K in a
Muffle furnace for 3-4 h to get a constant mass. Where, loss in mass reflected the volatile matter
(VM) and residual mass reflected the ash content. The composition of organic elements
including Carbon (C), Hydrogen (H), Sulphur (S), Nitrogen (N) and Oxygen (O) in the sample
was estimated by an elemental analyzer (Vario EL Cube, Germany). During analyses, the Argon
(Ar) was used as a carrier gas.

The High Heating Value (HHV) indicates the amount of energy to be evolved from a
biomass. However, the experimental procedures to determine the HHV (MJ kg-1) are expensive
and may give undesired experimental faults (Nhuchhen and Salam, 2012). Hence, previously
several correlation models were developed to calculate HHV. Here, the most appropriate model
established to date was used to calculate the HHV as described (Nhuchhen and Salam, 2012).

5
2.2 TGA-DSC Experiment
TGA-DSC analyses were performed using an STA-409, NETZSCH-Gerätebau GmbH,
Germany. After calibration (as described in the instruction manual), almost ten mg (10) of milled
biomass (150-200 µm particle size) was put into the alumina crucibles and constantly heated
from ambient temperature to 1275 K. Where, constant heating at the rate of 10, 30 and 50 Kmin-1
was used. An inert environment was maintained using the nitrogen gas flow (100 mL min-1) into
the reaction chamber.

2.3 ANN Model Development

An Artificial Neural Network (ANN) model was established to govern weight loss as
output data, using heating rate and temperature as input variables. The feed-forward Levenberg-
Marquardt back-propagation algorithm was selected in MATLAB® R2014b for data prediction.
The model depicts the input, hidden and output layers of the multi-layer network. Two neurons
were included in the input layer; heating rate and temperature. In contrast, the output layer had a
single neuron, the temperature dependent weight loss. 1,021 data points were used in this
analysis which were divided among the training (70%), validation (15%) and testing (15%)
phases. The epoch number was set to 52 and 6 validation checks were applied. The Mean Square
Error (MSE) function, as seen in Equation 1, was used as an error function to evaluate the
performance of each phase. The network model was optimized based on the target (t) and output
values (o), as expressed in Equation 2.

 = ∑% $( (1)
!&' !" #! 

where λi: experimental values; βi: predicted values; n: number of data points

∑!,! -.! $
)* = 1 − + ∑!.! $
/ (2)

2.4 Mathematical model development for thermogravimetric analyses

A mathematical model was derived from analyzing the data obtained from the TGA-DSC
experiments. In isoconversional methods, the disintegration rate of the sample is depicted as:

6
01
= 234 (3)
0,

Where;

4 =  . − ,  /  . − 6  (4)

Here
. is the initial mass, , is change in the mass, and 1 is the residual mass

Using 2 , equation (3) was re-written as follows;

01 ;
= 7 89 :− <= > 3 4  (5)
0,

Where
A is pre-exponential factor (s-1)
?@
E is activation energy :A.B >
T is temperature in Kelvin (K)
@
R is Universal gas constant 8.314 ?.A.B 

t is the time in sec

0=
Later, the heating rate G = and the conversion function, 34  = 1 − 4 were intrudced by
0,

which following equation was obtained;

01 H ;
= # 89 :− <= > 1 − 4  (6)
0=

Then, equation (6) was integrated for the initial conditions, α = 0, at I = IJ , and after some
mathematical manpulation, the following equation was obtained;

1 ;
K4  = LJ M4⁄1 − 4  = 7)I * ⁄G O1 − 2)I⁄ Q89 − <=  (7)

K is the integral conversion function.

Equation 7 was rearranged, as the quantity 2)I⁄ was negligible when compared with unity,
hence it was ignored (Coats and Redfern, 1964), which resulted in following equation;

7
K4  = 7)I * ⁄G 89− ⁄)I (8)

2.5 Calculation of kinetic and thermodynamic parameters

Kinetic parameters of the pyrolysis reaction are vital to understanding the thermal
degradation behavior of the sample under study. Here, these parameters were calculated using
the isoconversional models as described by FWO (Flynn-Wall-Ozawa) and KAS (Kissenger-
Akahira-Sunose) (Akahira and Sunose, 1969; Flynn and Wall, 1966; Ozawa, 1965). A model-
fitting method named Coats-Redfern (Coats and Redfern, 1964) was used to describe the reaction
mechanism which includes an order of reaction, diffusional, and the contracting geometry (White
et al., 2011). Accordingly, activation energy could be determined by using 34  OR K4  that
was further compared with the results found from KAS and FWO methods to predict the most
accurate reation mechanism.

The equation 8 was rearranged after taking logarithm on both sides to get following equation;

#
ln := $> = T7)⁄K4  −  ⁄)I (9) KAS method

Moreover, equation 6 was subjected to integration using the initial conditions, α = 0, at I = IJ .

Later, Doyle’s approximation was introduced followed by a few mathematical modifications
(Doyle, 1961), which gave equation (10);
ln G = T 7/)K4 – /)I (10) FWO method

*<=
The Coats and Redfern (CR) Method relies on asymptotic approximation → 0, giving
;
following equation;

W1
ln   = T 7)/G – /)I (11) CR Method
=$

The left side of each equation (9, 10 and 11) was plotted (y-axis) against the inverse of pyrolysis
temperature (x-axis), for selected conversion point (α) to calculate kinetic parameters. The
-
conversion point (α) was used to calculate the pre-exponential factors (7  using the
# W1
conversion points plotted between ln = $, ln G and ln : > against 1/T which gave a straight
=$

line. The activation energy values were calculated from the slopes (E). Moreover, the

8
thermodynamic parameters including ∆Y (enthalpy), ∆Z (Gibbs free energy) and ∆ (entropy)
were calculated as described previously (Kim et al., 2010; Xu and Chen, 2013).

7 = OG. 89 ⁄)IA Q⁄[)IA* \ (12)

∆Y =  − )I (13)

∆Z =  + )IA ln ]^ IA ⁄ℎ7 (14)

∆ = ∆Y − ∆Z ⁄IA (15)

Where:

]^ Boltzmann Constant (1.381 ∗ 10-*` a/]

ℎ Plank Constant 6.626 ∗ 10-`c a 

IA DTG peak temperature, K

3 Results and Discussion

3.1 Physicochemical parameters

The T. latifolia biomass contained C, H, N, S, and O as 44.00%, 6.09, 2.45%, 0.34%,

and 32.34%, respectively. The sample was shown to contain 71% volatile matter, 19.5% fixed
carbon and 8.8% alkali. The lower nitrogen (<2.45%) and sulfur content (<0.34%) in the sample
indicated that there is a lower risk of emission of toxic gases (NOx, SOx) from its pyrolysis. The
range of volatile content (%) in the was shown to be within the range exhibited by the
established bioenergy crops including Miscanthus gigantus and Arduno donax (Jeguirim et al.,
2010).

The HHV indicates the amount of energy available from the biomass upon combustion.
The estimated HHV of the sample was shown be 18.32 MJ kg-1 which is reasonably higher than
the HHVs of several well-known energy crops including A. donax (Giant reed), M. gigantus,
Phalaris arundinacea (Reed canarygrass), Salix spp. (Willow), Para grass and Camel grass,
which had shown the HHVs as 17.2, 17.80, 16.30, 15.03 (Howaniec and Smolinski, 2011;

9
Jeguirim et al., 2010; Paulrud and Nilsson, 2001), 15.10 (Ahmad et al., 2017b), and 15.00 MJ kg-
1
(Mehmood et al., 2017), respectively. The estimated HHV indicate the remarkable energy
potential of the T. latifolia when compared to recognized bioenergy crops. However, the HHV of
T. latifolia biomass was shown to be lower when compared to HHV of sweet sorghum, i.e. 20-25
MJ kg-1 (Monti et al., 2008; Yan et al., 2016).

3.2 Analyses of TG-DTG curves

Thermogravimetric analyses exhibit the loss in biomass in response to increasing reaction

temperature. Where lost mass is converted into various products. The curves obtained during this
analysis (TG-DTG curves) indicate the thermochemical conversion of the subjected biomass into
solids, liquids and gaseous products (Maia and de Morais, 2016). For the sample under study, the
curves showed typical trend of thermal degradation of lignocelluloses when compared to the TG-
DTG curves (Fig. 1) found for Switchgrass, Cardoon leaves, Elephant grass, Camel grass, rice
husk, and red pepper waste (Biney et al., 2015; Braga et al., 2014; Maia and de Morais, 2016;
Mehmood et al., 2017; Xu and Chen, 2013).
The characteristic temperatures associated with the mass loss during thermal degradation,
are shown in Tables 1 and 2. The rate of thermal conversion of the sample was shown to be
increased with the increased heating rate (Tables 1 and 2). The thermal conversion of the sample
was shown to comprise of three stages with two zones during stage-II. The first stage started
from ambient temperature to 485-495 K for all heating rates, with the loss of 7.42-8.32 % in
mass, which indicates the release of retained moisture content within intercellular spaces or
intracellular compartments. The second stage ranged from 485-660 K taking all heating rates into
account, where most of the mass loss occurred (i.e. 51%). This stage was shown to contain two
zones for all heating rates, where zone-I occurred between 485-591 K while zone-II appeared
between 568-660 K. The third stage occurs from 660-1275 K, where almost 17% of the total
mass was lost. The biomass containing lesser that 10 % retained moisture, is considered feasible
for combustion, which makes this sample suitable for pyrolysis and combustion (Braga et al.,
2014). However, thermal transformation showed the typical pattern of lignocellulosic biomass.

10
Where, most of the thermal conversion happened during stage-II, indicating the degradation of
hemicellulose, cellulose, and pectin where the typical temperature for their degradation ranges
from 485K through 660 K (Xu and Chen, 2013). The temperature range associated with the third
stage followed by the long tail reflects lignin degradation and char formation (Braga et al., 2014).
Most of the thermal transformation happened up to 700 K where 58%-62% loss of the mass
happened. Hence, the thermal conversion of the T. latifolia biomass into various products may be
optimized within this temperature range, using lower heating rate in an energy efficient manner.
These values indicated the advantage of using T. latifolia for pyrolysis and combustion, when
compared to the previously studied biomass samples including rice husk, water hyacinth, and
elephant grass (Biney et al., 2015; Braga et al., 2014; Huang et al., 2016). Biochar yields of
24.59, 25.45 and 23.67% were observed up to 660 K at three heating rates, which were
comparable to the biochar yields obtained from the pyrolysis of straw (23.68%) and bran
(25.17%) of rice plant (Xu & Chen, 2013), and lower than Para grass (31.5%) (Ahmad et al.,
2017b), and Camel grass (30.46%) (Mehmood et al., 2017). These values indicated the
appropriateness of the sample for biochar production.
3.3 Heat flow during pyrolysis reaction

The DSC curves showed a direct connection between the heating temperature and the
flow of heat (mWmg-1) because most of the pyrolysis reaction indicated an active reaction
mechanism (Fig. 2). However, ending stages showed a decreasing heat flow. The reaction rate
was steadily enhanced from ambient temperatures to 589, 754 and 799 K at 10, 30 and 50 Kmin-
1
, respectively, which reflects exothermic reactions. These differences in heat flow may be due to
the poor thermal conductivity of the biomass at various heating rates. The DSC curves started
shifting towards x-axes at higher temperatures, which indicates the reaction to cease due to
depletion of reactant (the biomass) or the change of reaction mechanism due to changing
composition of the residual biomass. Because most of the volatiles were lost above 700 K,
leading towards a different heat flow into the changing composition of the residual biomass.
These curves exhibited parallel trend when compared to the DSC curves observed for bamboo
leaves (Kow et al., 2016), the Potamogeton crispus and Sargassum thunbergii, where former is a
freshwater plant and later is a marine macroalga (Li et al., 2012).

11
3.4 Kinetics and thermodynamic parameters

Figure 3 shows the linear fit plots to determine the activation energy values using KAS
and FWO methods. These slopes and derived equations were used to calculate the conforming
values of E and A as shown in Table 3. The average E-values were 184 kJ mol-1 and 182 kJ mol-1
as evaluated by both (KAS and FWO) methods. Moreover, a plot between E-values and
conversion points (α) indicated that both methods have nearly same values at each conversion
point. The relationship of activation energies, temperature and conversion points is described in
Table 4. The observed range of E is lower than E-values (118-257 kJ mol-1) of tobacco plant
waste (Wu et al., 2015), rice husk (221-229 kJ mol−1), cellulose (191 kJ mol−1) and elephant grass
(218-227 kJ mol−1) (Braga et al., 2014; Sanchez-Jimenez et al., 2013) and this range was
approximately same as Para grass (Ahmad et al., 2017b) and shown to be higher than
switchgrass (Biney et al., 2015). This correspondence of E-values of T. latifolia makes it suitable
for co-pyrolysis with several other biomass feedstocks.

The difference between the activation energy values and enthalpies reflects the likelihood
of the pyrolysis reaction to occur. Where lower difference indicates that product formation would
be favorable. A difference of ~5 kJ mol-1 was observed between the E and ∆H values that
indicated that there is little potential energy barrier to achieve the product formation, reflecting
that product formation would be easier to achieve (Vlaev et al., 2007). Moreover, pre-
exponential factors (A-values) explain the reaction chemistry, which is critically important to
know while optimizing the pyrolysis of biomass. While lower A-values (<109 s-1) show largely a
surface reaction. However, if the reaction is not surface dependent, then lower A-values also
designate a closed complex. Alternatively, higher A-values (≥ 109 s-1) show a simpler complex
(Turmanova et al., 2008). For the sample under study, the A-values ranged from 5.53x1010 ̶
3.02x1015 s-1 and 7.61x1009 ̶ 7.93x1015 s-1 as obtained from KAS and FWO methods, respectively
(Table 3) that indicated the complexity of biomass. Moreover, A-values of sample were higher
when compared to A-values of red-peppers waste (3.80 x 100 to 2.80 x1012 s-1), rice straw (1.70
x1007 to 9.35 x1012 s-1), rice bran (1.00 x1007 and 1.58 x1010 s-1) and were lower when compared
to switchgrass (3.70 x1003-1.65 x1021 s-1) (Maia and de Morais, 2016). Gibbs free energy (∆G)
reflects the amount of energy which become available from that biomass upon pyrolysis. Here,
12
the ∆G values were shown to be ranging from 173-175 that are higher when compared with the
∆G values of the rice bran (167.17 kJ mol-1), rice straw (164.59 kJ mol-1) and waste from red-
peppers (139.4 kJ mol-1) (Maia and de Morais, 2016; Xu and Chen, 2013). It indicated that the
pyrolysis of the T. latifolia will provide more energy when compared to the rice bran, rice straw
and red-peppers waste.

3.5 Reaction mechanism of the pyrolysis

The CR method was devised on thermogravimetric data at three different heating rates to
W1
obtained E-values based on various reaction mechanisms (Table 5). Plots between ln : >
=$

against 1/T were produced on various reaction mechanism functions as shown in Fig. 4. There
were two regions formed based on conversion of the sample on three different heating rates.
Region-I was defined when α ranged from 0. 1 ≤ 4 ≤ 0.4 and Region-II was defined when α
ranged from 0. 4 ≤ 4 ≤ 0.8, where the major part of decomposition occurred. If the average
activation energy values obtained from these mechanism functions at different heating rates are
nearly equal to the energy values obtained from KAS and FWO methods, it shows this
mechanism function should be the best-fit reaction of that region.
It was shown that in region-I the average E values are different in different reaction
mechanism functions under three different heating rates. For reaction order model, the average E
values of three heating rates ranged from 5.75 to 95.79 kJ mol-1, whereas in diffusional stage
these values ranged from 177.14 to 192.25 kJ mol-1 which were closest to the values obtained
from KAS and FWO methods. It indicates that during the reactions displayed in Region-I, the
diffusion played a key role (A. Khawam, 2006). In contraction geometry, the average E-values
were too small and ranged from 88 to 91 kJ/mol. Similarly, in Region-II the average E-values
depended upon 2nd and 3rd order reaction model and ranged from 116 to 200 kJ mol-1 which
shows that E-values obtained from KAS and FWO methods are in between these values.
Therefore, reaction order mechanism is classified as diffusion type followed by the 2nd and 3rd
order reaction which is proportional to the concentration, total or residual amount of reactant(s)
in a certain reaction (A. Khawam, 2006).
3.6 Prediction of pyrolytic behavior by ANN model

13
 ANN model was applied to predict and validate the pyrolysis experiment at three
different heating rates to simulate the predicted results with experimental data to further
understand the pyrolytic behavior of T. latifolia. First, one hidden layer was created to have a
simpler ANN model but three different breakdown stages resulted in sharp changes in data which
complicated behavior of the sample. The best network performance was achieved with two
hidden layers. Moreover, error distribution was analyzed at each step to ensure the accuracy of
the network. The regression of each step was carried out during the optimization of the network
that showed good correlation between targets and the output values, as shown in the Fig. 5.
Moreover, the histogram error distribution diagram appears to be normally distributed for the
major part of the obtained dataset. The R2 value for the model fit at all stages appeared to be very
close to 1, signifying a very good fit of the model to the experimental data (Fig. 5). Additionally,
all these errors fall within a relatively narrow range; thus, indicating a good model fit. Here, 52
iterations were carried out by MATLAB and the best performance was observed at the 46th
iteration with the minimum MSE (i.e. 0.53478). The obtained best performance results, as
generated by MATLAB, are shown in Fig. 6. This finding indicates that ANN may be frequently
applied to understand and envisage the pyrolysis of the biomass.

Conclusion

Major pyrolysis products of T. latifolia can be achieved at 485-660K. The E and ∆G values
ranged from 182-184 kJmol-1 and 171-175 kJmol-1 , respectively. The HHV value (18.32 MJ kg-1)
was shown to be higher than several established energy crops. Moreover, best-fit plots were
obtained by comparing experimental data with the predicted data points obtained from ANN
simulation. The reaction mechanism showed the pyrolysis to contain two regions. Region-I
(0.1 ≤ 4 ≤ 0.4) and Region-II (0.4 ≤ 4 ≤ 0.8). The best reaction model in Region-I was
diffusion while in Region -II the best model was built on the 2nd to 3rd order reaction.

Acknowledgements

The financial support of Higher Education Commission Pakistan is highly appreciated.

Appendix-A: Supplementary data

14
This article contains supplementary information in its online version.

References
1. A. Khawam, D.R.F., 2006. Solid-state kinetic models: basics and mathematical
fundamentals. J. Phys. Chem. B. 110, 17315–17328.
2. Ahmad, M.S., Mehmood, M.A., Ye, G., Al Ayed, O.S., Ibrahim, M., Rashid, U., Luo, H.,
Qadir, G., Nehdi, I.A. 2017a. Thermogravimetric analyses revealed the bioenergy
potential of Eulaliopsis binate. J Therm. Anal. Calorim. DOI: 10.1007/s10973-017-6398-
x.
3. Ahmad, M.S., Mehmood, M.A., Al Ayed, O.S., Ye, G., Luo, H., Ibrahim, M., Rashid, U.,
Nehdi, I.A., Qadir, G., 2017b. Kinetic analyses and pyrolytic behavior of Para grass
(Urochloa mutica) for its bioenergy potential. Bioresour. Technol. 224, 708-713.
4. Akahira, T., Sunose, T., 1969. Transactions of Joint Convention of Four Electrical
Institutes. 246.
5. Álvarez, A., Pizarro, C., García, R., Bueno, J.L., Lavín, A.G., 2016. Determination of
kinetic parameters for biomass combustion. Bioresour. Technol. 216, 36-43.
6. Biney, P.O., Gyamerah, M., Shen, J., Menezes, B., 2015. Kinetics of the pyrolysis of
arundo, sawdust, corn stover and switch grass biomass by thermogravimetric analysis
using a multi-stage model. Bioresour. Technol. 179, 113-122.
7. Braga, R.M., Melo, D.M., Aquino, F.M., Freitas, J.C., Melo, M.A., Barros, J.M., Fontes,
M.S., 2014. Characterization and comparative study of pyrolysis kinetics of the rice husk
and the elephant grass. J. Therm. Anal. Calorim. 115, 1915-1920.
8. Chen, J., Liu, J., He, Y., Huang, L., Sun, S., Sun, J., Chang, K., Kuo, J., Huang, S., Ning,
X., 2017. Investigation of co-combustion characteristics of sewage sludge and coffee
grounds mixtures using thermogravimetric analysis coupled to artificial neural networks
modeling. Bioresour. Technol. 225, 234-245.
9. Coats, A., Redfern, J., 1964. Kinetic parameters from thermogravimetric data.
Nature 201, 68 – 69.
10. Conesa, J.A., Caballero, J.A., Reyes-Labarta, J.A., 2004. Artificial neural network for
modelling thermal decompositions. J. Anal. Appl. Pyrol. 71, 343-352.

15
11. Di Blasi, C., 2009. Combustion and gasification rates of lignocellulosic chars. Prog.
Energy Combust. Sci. 35, 121-140.
12. Doyle, C., 1961. Kinetic analysis of thermogravimetric data. J. Appl. Poly. Sci. 5, 285-
292.
13. Edrisi, S.A., Abhilash, P., 2016. Exploring marginal and degraded lands for biomass and
bioenergy production: An Indian scenario. Renew. Sustain. Energy Rev. 54, 1537-1551.
14. Flynn, J.H., Wall, L.A., 1966. A quick, direct method for the determination of activation
energy from thermogravimetric data. J. Poly. Sci. B. 4, 323-328.
15. He, C.-R., Kuo, Y.-Y., Li, S.-Y., 2017. Lignocellulosic butanol production from Napier
grass using semi-simultaneous saccharification fermentation. Bioresour. Technol. 231,
101-108.
16. Howaniec, N., Smolinski, A., 2011. Steam gasification of energy crops of high
cultivation potential in Poland to hydrogen-rich gas. Int. J. Hydro. Energy 36, 2038-2043.
17. Huang, L., Liu, J., He, Y., Sun, S., Chen, J., Sun, J., Chang, K., Kuo, J., 2016.
Thermodynamics and kinetics parameters of co-combustion between sewage sludge and
water hyacinth in CO2/O2 atmosphere as biomass to solid biofuel. Bioresour. Technol.
218, 631-642.
18. Jeguirim, M., Dorge, S., Trouve, G., 2010. Thermogravimetric analysis and emission
characteristics of two energy crops in air atmosphere: Arundo donax and Miscanthus
giganthus. Bioresour. Technol. 101, 788-793.
19. Kalogirou, S.A., 2003. Artificial intelligence for the modeling and control of combustion
processes: a review. Prog. Energy Combust. Sci. 29, 515-566.
20. Kim, Y.S., Kim, Y.S., Kim, S.H., 2010. Investigation of thermodynamic parameters in
the thermal decomposition of plastic waste− waste lube oil compounds. Environ. Sci.
Technol. 44, 5313-5317.
21. Kow, K-W., Yusoff, R., Abul Aziz, A.R., Abdullah, E.C. 2014. Characterisation of bio-
silica synthesised from cogon grass (Imperata cylindrica). Powder Technol. 254, 206-
213.

16
22. Kow, K-W., Yusoff, R., Abul Aziz, A.R., Abdullah, E.C. 2016. Determination of kinetic
parameters for thermal decomposition of bamboo leaf to extract bio-silica. Energ. Source
Part A. 38, 3249-3254.
23. Li, D., Chen, L., Chen, S., Zhang, X., Chen, F., Ye, N., 2012. Comparative evaluation of
the pyrolytic and kinetic characteristics of a macroalga (Sargassum thunbergii) and a
freshwater plant (Potamogeton crispus). Fuel. 96, 185-191.
24. Maia, A.A.D., de Morais, L.C., 2016. Kinetic parameters of red pepper waste as biomass
to solid biofuel. Bioresour. Technol. 204, 157-163.
25. Maurya, R., Ghosh, T., Saravaia, H., Paliwal, C., Ghosh, A., Mishra, S., 2016. Non-
isothermal pyrolysis of de-oiled microalgal biomass: Kinetics and evolved gas analysis.
Bioresour. Technol. 221, 251-261.
26. Mehmood, M.A., Ye, G., Luo, H., Liu, C., Malik, S., Afzal, I., Xu, J., Ahmad, M.S.,
2017. Pyrolysis and kinetic analyses of Camel grass (Cymbopogon schoenanthus) for
bioenergy. Bioresour. Technol. 228, 18-24.
27. Monti, A., Di Virgilio, N., Venturi, G., 2008. Mineral composition and ash content of six
major energy crops. Biomass Bioener. 32, 216-223.
28. Nhuchhen, D.R., Salam, P.A., 2012. Estimation of higher heating value of biomass from
proximate analysis: A new approach. Fuel. 99, 55-63.
29. Ozawa, T., 1965. A new method of analyzing thermogravimetric data. Bull. Chem. Soc.
Jpn. 38, 1881-1886.
30. Paulrud, S., Nilsson, C., 2001. Briquetting and combustion of spring-harvested reed
canary-grass: effect of fuel composition. Biomass Bioener. 20, 25-35.
31. Rezende, M.L., Richardson, J.W., 2017. Risk analysis of using sweet sorghum for
ethanol production in southeastern Brazil. Biomass Bioener. 97, 100-107.
32. Sanchez-Jimenez, P.E., Perez-Maqueda, L.A., Pereon, A., Criado, J.M., 2013.
Generalized master plots as a straightforward approach for determining the kinetic
model: the case of cellulose pyrolysis. Thermochimica Acta 552, 54-59.
33. Trninić, M., Wang, L., Várhegyi, G., Grønli, M., Skreiberg, Ø., 2012. Kinetics of corncob
pyrolysis. Energy Fuels 26, 2005-2013.

17
34. Turmanova, S.C., Genieva, S., Dimitrova, A., Vlaev, L., 2008. Non-isothermal
degradation kinetics of filled with rise husk ash polypropene composites. Exp. Poly. Lett.
2, 133-146.
35. Ussiri, D.A.N., Lal, R., 2017. The Role of Bioenergy in Mitigating Climate Change
Carbon Sequestration for Climate Change Mitigation and Adaptation. Springer
International Publishing, Cham, pp. 433-495.
36. Uzun, H., Yıldız, Z., Goldfarb, J.L., Ceylan, S., 2017. Improved prediction of higher
heating value of biomass using an artificial neural network model based on proximate
analysis. Bioresour. Technol. 234, 122-130.
37. Vlaev, L., Georgieva, V., Genieva, S., 2007. Products and kinetics of non-isothermal
decomposition of vanadium (IV) oxide compounds. J. Therm. Anal. Calorim. 88, 805-
812.
38. White, J.E., Catallo, W.J., Legendre, B.L., 2011. Biomass pyrolysis kinetics: A
comparative critical review with relevant agricultural residue case studies. J. Anal. Appl.
Pyrol. 91, 1-33.
39. Wu, W., Mei, Y., Zhang, L., Liu, R., Cai, J., 2015. Kinetics and reaction chemistry of
pyrolysis and combustion of tobacco waste. Fuel 156, 71-80.
40. Xu, Y., Chen, B., 2013. Investigation of thermodynamic parameters in the pyrolysis
conversion of biomass and manure to biochars using thermogravimetric analysis.
Bioresour. Technol. 146, 485-493.
41. Yan, H.-L., Zong, Z.-M., Li, Z.-K., Kong, J., Zheng, Q.-X., Li, Y., Wei, X.-Y., 2016.
Sweet sorghum stalk liquefaction in supercritical methanol: Effects of operating
conditions on product yields and molecular composition of soluble fraction. Fuel Process.
Technol. http://dx.doi.org/10.1016/j.fuproc.2016.02.011.
42. Yıldız, Z., Uzun, H., Ceylan, S., Topcu, Y., 2016. Application of artificial neural
networks to co-combustion of hazelnut husk–lignite coal blends. Bioresour. Technol.
200, 42-47.
43. Zhang, S., Dong, Q., Zhang, L., Xiong, Y., 2016. Effects of water washing and
torrefaction on the pyrolysis behavior and kinetics of rice husk through TGA and Py-
GC/MS. Bioresour. Technol. 199, 352-361.
18
44. Zhang, X., Xu, M., Sun, R., Sun, L., 2006. Study on biomass pyrolysis kinetics. J. Eng.
Gas Turbines Power 128, 493-496.

19
 Heating Rate 10 Kmin-1
120 0
TMIN T1 T5 TMAX
100 -2

DTG (% min-1)
Mass Loss (%)

T3
80 T2 -4
TG
DTG
60
I II III -6

40 T4 -8

20 -10
250 350 450 550 650 750 850 950 1050 1150
Temperature (K)

Heating Rate 30 Kmin-1

120 0
TMIN T1 T5 TMAX
-4
100

DTG (% min-1)
Mass Loss (%)

T3 -8
80 -12
T2 TG
60 DTG -16
I II III -20
40
-24
T4
20 -28
250 350 450 550 650 750 850 950 1050 1150
Temperature (K)

Heating Rate 50 Kmin-1

120 0
TMIN T1 T5 TMAX
100 -10
DTG (% min-1)
Mass Loss (%)

T3
-20
80 T2
TG -30
DTG
60
I II III -40
40 -50

20 T4 -60
250 350 450 550 650 750 850 950 1050 1150
Temperature (K)

Fig. 1 TG-DTG curves indicating percent mass loss of Typha latifolia

20
 20

Heating Rate (K min-1)

Heat Flow DSC (mW mg-1)

10
15
30
50
10

-5
250 350 450 550 650 750 850 950 1050 1150 1250
Temperature (K)

Fig. 2 DSC curves indicating heat flow across the biomass

21
 -8
(A) 0.10
-8.5 0.15
0.20
0.25
-9 0.30
0.35
0.40
ln(β/T2)

-9.5
0.45
0.50
-10 0.55
0.60
-10.5 0.65
0.70
0.75
-11 0.80
0.85
-11.5 0.90
0.0012 0.0013 0.0014 0.0015 0.0016 0.0017 0.0018 0.0019 0.002 0.0021 0.0022
Inverse of pyrolysis temperature (-K-1)

4 0.10
(B) 0.15
3.8
0.20
3.6 0.25
0.30
3.4 0.35
0.40
3.2 0.45
ln(β)

3 0.50
0.55
2.8 0.60
0.65
2.6 0.70
0.75
2.4 0.80
2.2 0.85
0.90
2
0.0012 0.0013 0.0014 0.0015 0.0016 0.0017 0.0018 0.0019 0.002 0.0021 0.0022
Inverse of pyrolysis temperature (-K-1)

Fig.3 Linear fit plots to determine the activation energy values using KAS and FWO methods

22
 REGION-I Reaction Order
-11

-12

-13

-14
Diffusion
ln(gα/T²)

-15

-16

-17

-18 Contracting Geometry

-19

-20
0.00165 0.0017 0.00175 0.0018 0.00185
Inverse of pyrolysis temperature (-K-1)

REGION-II Reaction Order

-8

-9

-10

-11
Diffusion
ln(gα/T²)

-12

-13

-14

-15 Contracting Geometry

-16

-17
0.00155 0.0016 0.00165 0.0017
Inverse of pyrolysis temperature (-K-1)

Fig.4 CR Plots of for pyrolysis of T. latifolia. Region-I (. 1 ≤ 4 ≤ .4) and Region-II (. 4 ≤ 4 ≤

.8)

23
Fig.5 Regression of train, validate and testing steps together with histogram diagram
24
Fig.6 Best performance Mean Square Error (MSE) vs Targets

25
Table 1 Characteristic temperatures associated with the mass loss

Heating rate Temperature (K)

(Kmin-1) Tmin T1 T2 T3 T4 T5 Tmax
10 485 538 568 606 644
30 298 499 555 584 621 654 1300
50 495 556 591 623 660

Table 2

Mass loss during different stages of decomposition with increasing temperature

Heating rate (Kmin-1)

Stages Temperature
10 30 50
Stage-I, WL (%) Tmin-T1 7.45 7.46 8.32
Stage-II, Region-I T1-T3 19.65 20.58 21.62
WL (%) Region-II T3-T5 30.83 30.30 30.17
Stage-III, WL (%) T5-Tmax 17.48 16.21 16.22
Final residues at 1075-
-- 24.59 25.45 23.67
1275K (%)

26
 Table 3 Kinetic and thermodynamics parameters of T. latifolia

Ea ∆H A ∆G ∆S Ea ∆H A ∆G ∆S
α R² R²
-1 -1 -1 -1 -1 -1 -1 -1 -1
kJmol kJmol s kJmol Jmol kJmol kJmol s kJmol Jmol-1

KAS method FWO method

0.1 147.20 0.99 142.04 5.53E+10 175.37 -53.67 134.78 0.99 129.62 7.61E+09 173.18 -70.16
0.15 173.56 0.98 168.40 1.08E+13 174.51 -9.85 173.31 0.99 168.15 1.71E+13 171.88 -6.01
0.2 183.84 0.98 178.68 8.34E+13 174.22 7.18 183.32 0.99 178.16 1.25E+14 171.59 10.57
0.25 190.85 0.98 185.68 3.37E+14 174.02 18.78 190.15 0.99 184.98 4.88E+14 171.41 21.87
0.3 189.93 0.99 184.76 2.80E+14 174.05 17.25 189.12 0.99 183.96 3.98E+14 171.43 20.17
0.35 201.89 0.99 196.73 3.02E+15 173.73 37.02 204.18 0.99 199.01 7.93E+15 171.04 45.05
0.4 185.30 0.99 180.14 1.12E+14 174.18 9.60 188.97 0.99 183.81 3.86E+14 171.44 19.92
0.45 171.24 0.99 166.07 6.77E+12 174.58 -13.71 172.40 0.98 167.24 1.42E+13 171.91 -7.52
0.5 201.89 0.99 196.73 3.02E+15 173.73 37.02 200.18 0.99 195.02 3.59E+15 171.14 38.46
0.55 193.13 0.99 187.96 5.30E+14 173.96 22.54 196.03 0.99 190.87 1.57E+15 171.25 31.60
0.6 197.73 0.99 192.57 1.32E+15 173.84 30.16 196.55 0.99 191.39 1.74E+15 171.23 32.45
0.65 186.65 0.99 181.49 1.46E+14 174.14 11.83 187.22 0.99 182.06 2.73E+14 171.49 17.02
0.7 198.37 0.99 193.21 1.50E+15 173.82 31.21 196.39 0.99 191.23 1.69E+15 171.24 32.19
0.75 196.05 0.98 190.89 9.48E+14 173.89 27.38 190.75 0.98 185.58 5.50E+14 171.39 22.86
0.8 153.94 0.99 148.78 2.14E+11 175.13 -42.44 153.00 0.99 147.84 2.95E+11 172.53 -39.76
0.85 188.69 0.99 183.53 2.19E+14 174.08 15.22 190.75 0.99 185.58 5.50E+14 171.39 22.86
0.9 177.61 0.99 172.45 2.41E+13 174.40 -3.13 158.36 0.99 153.20 8.62E+11 172.35 -30.84
Avg. 184.58 0.99 179.42 -- 174.22 -- 182.67 0.99 177.51 -- 171.64 --

27
Table 4 Relationship between conversion (α), pyrolysis temperature (T) and activation energies
(E)

Conversion Temperature Activation energy

Reactions
range (α) range (K) (E)
Release of retained water
moisture and degradation of Increased from starting
α ≤ 0.1 273-485
small and simple sugar point to 147 kJ mol-1
molecules
Thermal conversion of Increased from 147 to 185
0.1 ≤ α ≤ 0.4 485-591
cellulose and pectin kJ mol-1
Degradation of hemicellulose Increased from 185 to 196
0.4 ≤ α ≤ 0.8 591-660
and lignin kJ mol-1
Residual lignin decomposition Decreased from 196 to
0.1 ≤ α ≤ 1.0 660-1200
and formation of char 177 kJ mol-1

28
Table 5: Activation Energy values and reaction mechanism based on Coats-Redfern method

29
 10 Kmin-1 30 Kmin-1 50 Kmin-1 Average values
Reaction Ea Ea Ea Ea
Region g(α) R² R² R² R²
Model
kJmol-1 kJmol-1 kJmol-1 kJmol-1
I Reaction Order
Zero-order (F0) α 5.52 0.99 5.67 0.98 6.05 0.97 5.75 0.98
First-order (F1
−ln1 − 4 82.50 0.97 82.81 0.99 83.46 0.99 82.92 0.98
)
nth-order (Fn) O1 − 1 − 4¯¹/−1Q 94.97 0.99 95.21 0.95 97.18 0.96 95.79 0.97
O1 − 1 − 4¯²/−2Q 15.31 0.88 17.23 0.98 18.54 0.96 17.03 0.94
O1 − 1 − 4¯³/−3Q 25.55 0.96 27.67 0.99 29.33 0.98 27.52 0.98
Diffusion
1-D 4² 174.48 0.99 177.70 0.99 179.23 0.99 177.14 0.99
2-D 1 − 4T1 − 4 + 4 182.38 0.99 183.21 0.98 185.93 0.99 183.84 0.99
3-D (Jander) O1 − 1 − 4⅓Q² 190.82 0.98 191.83 0.99 194.11 0.98 192.25 0.98
3-D (Ginstling-
1 − 2/34 − 1 − 4⅔ 185.19 0.96 186.22 0.97 188.73 0.97 186.71 0.97
Brounshtein)
Contracting Geometry
Cont. Area 1 − 1 − 4½ 88.58 0.99 90.01 0.90 87.51 0.98 88.70 0.96
Cont. Volume 1 − 1 − 4⅓ 90.67 0.84 91.41 0.89 93.85 0.88 91.98 0.87
II Reaction Order
Zero-order (F0) 4 34.69 0.95 35.61 0.96 36.59 0.96 35.63 0.96
First-order (F1) −T1 − 4 67.27 0.97 68.34 0.97 70.37 0.98 68.66 0.97
nth-order (Fn) O1 − 1 − 4¯¹/−1Q 48.61 0.98 50.51 0.98 49.40 0.99 49.51 0.98
O1 − 1 − 4¯²/−2Q 116.59 0.99 115.43 0.99 118.49 0.99 116.84 0.99
O1 − 1 − 4¯³/−3Q 197.37 0.96 200.71 0.98 202.51 0.97 200.20 0.97
Diffusion
1-D 4² 79.64 0.99 81.45 0.99 83.83 0.98 80.97 0.99
2-D 1 − 4T1 − 4 + 4 97.13 0.94 98.91 0.96 100.40 0.93 98.81 0.94
3-D (Jander) O1 − 1 − 4⅓Q² 120.01 0.81 124.23 0.84 127.82 0.80 124.02 0.82
3-D (Ginstling-
1 − 2/34 − 1 − 4⅔ 104.66 0.99 108.55 0.99 110.17 0.99 107.79 0.99
Brounshtein)
Contracting Geometry
Cont. Area 1 − 1 − 4½ 49.26 0.99 51.12 0.89 54.32 0.92 51.57 0.93
Cont. Volume 1 − 1 − 4⅓ 55.68 0.79 58.59 0.76 61.48 0.72 58.58 0.76

30
• The Typha latifolia offers low-cost biomass from poor soils across the globe
• Its biomass was subjected to pyrolysis and thermogravimetric study
• Pyrolytic behavior and reaction mechanism is described
• The biomass has bioenergy potential comparable to well-known bioenergy crops

31