You are on page 1of 6

Bioresource Technology 107 (2012) 476–481

Contents lists available at SciVerse ScienceDirect

Bioresource Technology
journal homepage: www.elsevier.com/locate/biortech

Application of the distributed activation energy model to the kinetic study


of pyrolysis of the fresh water algae Chlorococcum humicola
Kawnish Kirtania, Sankar Bhattacharya ⇑
Department of Chemical Engineering, Monash University, Clayton, VIC 3800, Australia

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

Article history: Apart from capturing carbon dioxide, fresh water algae can be used to produce biofuel. To assess the
Received 9 September 2011 energy potential of Chlorococcum humicola, the alga’s pyrolytic behavior was studied at heating rates of
Received in revised form 12 December 2011 5–20 K/min in a thermobalance. To model the weight loss characteristics, an algorithm was developed
Accepted 16 December 2011
based on the distributed activation energy model and applied to experimental data to extract the kinetics
Available online 24 December 2011
of the decomposition process. When the kinetic parameters estimated by this method were applied to
another set of experimental data which were not used to estimate the parameters, the model was capable
Keywords:
of predicting the pyrolysis behavior, in the new set of data with a R2 value of 0.999479. The slow weight
Fresh water algae
Chlorococcum humicola
loss, that took place at the end of the pyrolysis process, was also accounted for by the proposed algorithm
Pyrolysis which is capable of predicting the pyrolysis kinetics of C. humicola at different heating rates.
Kinetics Crown Copyright Ó 2011 Published by Elsevier Ltd. All rights reserved.
DAEM

1. Introduction Wjtowicz et al. (2003) applied the DAEM to biomass pyrolysis suc-
cessfully. Logistic distribution of activation energy was assumed
Since some algae have a high oil content, they can be used to and validated for cellulose pyrolysis by Cai et al. (2011) with nth
produce biodiesel (Demirbas, 2011); however, the separation of order DAEM; however, the complex pyrolysis behavior of fresh
lipids from the algae is difficult and energy intensive (Pimentel, water algae has not yet been modeled with the distributed activa-
2008). Instead of oil extraction, algae can be gasified to produce tion energy model. For the present study, the blue–green algae,
oil and gas (Tang et al., 2011; Bruhn et al., 2011). Pyrolysis is Chlorococcum humicola, was selected for pyrolytic and kinetic
the first step in the combustion or gasification process. The pyro- examination because it produces a high amount of biomass and lip-
lytic behavior and kinetics of three kinds of red algae were inves- ids (Chaichalerm et al., 2011). Thermogravimetric experiments
tigated by Li et al. (2011), and it became evident that their were performed at different heating rates to understand the pyro-
pyrolysis behavior was more complicated than that of fossil fuels. lytic behavior of the algae. Thereafter, a new algorithm for deter-
The authors estimated the activation energy for those three algae mining the kinetic parameters of DAEM was developed and the
by the Flynn–Wall–Ozawa (FWO) (Flynn and Wall, 1966; Ozawa, new algorithm for DAEM was validated with experimental data.
1965), KAS (Kissinger, 1957) and Popescu (Popescu, 1996) meth-
ods. Brown algae pyrolysis at different conditions was investi- 2. Methods
gated by Ross et al. (2009) and Anastasakis et al. (2011). For
extracting kinetic parameters, Anastasakis et al. (2011) used the 2.1. Pyrolysis experiments
methods of Weber (2008) and Saddawi et al. (2010). Pyrolysis
characteristics of blue–green algae with coal were studied by C. humicola biomass was grown in the Bio Engineering
Yuan et al. (2011) without any kinetic parameter estimation. Laboratory, Department of Chemical Engineering, Monash Univer-
Due to the availability of high performance computing, the dis- sity as described by Harun and Danquah (2011). The strain was
tributed activated energy model (DAEM) is widely used for differ- cultured in 100-L bag photobioreactors outdoors and kept under
ent kinds of solid and liquid fuels. Miura and Maki (1998) and semi-continuous conditions by 20% (v/v) dilution with fresh med-
Scott et al. (2006) modified the DAEM so that no assumptions ium after harvesting an equal volume of culture on a daily basis.
for the form of the activation energy distribution are required. The medium consists of potassium nitrate, KNO3 (150 g/L, MW
Navarro et al. (2008), Sonobe and Worasuwannarak (2008) and 101.11, 99.5%), sodium phosphate monobasic, NaH2PO4 (11.3 g/L,
MW 119.98, 99.5%), manganese chloride, MnCl2 (360.0 mg/L, MW
⇑ Corresponding author. Tel.: +61 3 9905 9623; fax: +61 3 99055686. 125.84, 98.0%), zinc sulfate, ZnSO4 (44.0 mg/L, MW 161.47, 99.5%),
E-mail address: Sankar.Bhattacharya@monash.edu (S. Bhattacharya). cobalt nitrate, Co(NO3)2 (22.0 mg/L, MW 182.94, 97.5%), copper

0960-8524/$ - see front matter Crown Copyright Ó 2011 Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.biortech.2011.12.094
K. Kirtania, S. Bhattacharya / Bioresource Technology 107 (2012) 476–481 477

sulfate, CuSO4 (19.6 mg/L, MW 159.61, 98.5%), sodium molybdite, ( "  #)ð1n
1
Þ
k0 RT 2 E
Na2MoO42H2O (12.6 mg/L, MW 241.95, 98.0%), sodium ZðE; TÞ ¼ 1 þ ð1  nÞ  exp  ð6Þ
aE RT
metasilicate, Na2SiO3 (22.7 g/L, MW 122.06, 99.7%), iron(III) citrate,
C6H5O7Fe3H2O (9.0 g/L, MW 298.99, 98%), and citric acid monohydrate,
This asymptotic approximation provides similar results to the ac-
COOHCH2C(OH)COOHCH2COOHH2O (9.0 g/L, MW 210.14, 99.5%).
tual one. Now, assuming k0 is independent of temperature, Z(E,T)
The ultimate analysis was performed with a CHNS/O analyzer
can be written as:
(Model 2400, Perkin-Elmer, USA). To determine carbon, hydrogen,
 
nitrogen and sulfur, the samples were combusted at 975 °C to obtain RT 2 E
ZðE; TÞ ¼ f1 þ ð1  nÞk0 nðE; TÞgð1nÞ where nðE;TÞ ¼ 
1

CO2, H2O, N2 and SO2 and analyzed through a chromatographic col- exp 
aE RT
umn. The system detects the gases by thermal conductivity and di-
ð7Þ
rectly gives the concentration of C, H, N, S in percent by comparing
with the standard values. Ash content was determined by combust- As n(E,T) is defined now, it is important to understand that the tem-
ing the alga sample at 800 °C. As the concentration of other materials perature integral is evaluated at each value of decomposition. So for
are in ppm range, the oxygen concentration was calculated by differ- each value of activation energy, the n(Ei,T) can be evaluated to form
ence. The pyrolysis behavior was studied in a thermobalance (Model a matrix. This matrix is formulated as:
STA 449 F3 JupiterÒ, NETZSCH-Gerätebau GmbH, Germany). The 2 3
nðE1 ; T 1 Þ nðE2 ; T 1 Þ    nðEL ; T F Þ
alga was dried in an oven for at least 12 h at 105 °C, ground to a size 6 7
of 106–150 lm using a hand mortar. The pyrolysis was done in non- 6 nðE1 ; T 2 Þ nðE2 ; T 2 Þ    nðEL ; T F Þ 7
6 7
isothermal condition and the temperature was ramped to 1100 °C at n ¼ 6 nðE1 ; T 3 Þ nðE2 ; T 3 Þ    nðEL ; T F Þ 7
6
7 ð8Þ
heating of 5, 10, and 20 K/min with one repeat for estimating the 6 7
4     5
kinetic parameters and model validation. nðE1 ; T F Þ nðE2 ; T F Þ    nðEL ; T F Þ
2.2. Kinetic modeling This matrix implies that Eq. (2) can be expressed in an equivalent
matrix which is possible to evaluate using any mathematical pack-
A Gaussian distribution based distributed activation energy mod- age. So the integrated matrix form of the mass loss equation
el (DAEM) was developed to analyze the pyrolysis kinetics and the becomes:
‘end loss’ of the algae. Usually the distributed activation energy 2 3 8 2 39ð1n
1 Þ2 3
model has a general form for several parallel first order reactions: wðT 1 Þ > > nðE1 ;T 1 Þ nðE2 ;T 1 Þ nðEL ;T F Þ > > f ðE1 Þ
Z  Z t  Z 6 7 >
> 6 7>
> 6 7
w 1 1 6wðT 2 Þ7 > > 6nðE1 ;T 2 Þ nðE2 ;T 2 Þ nðEL ;T F Þ7> > 6f ðE2 Þ7
¼ exp k0
E
eRT dt f ðEÞdE where f ðEÞdE ¼ 1 ð1Þ 16 7 < 6 7= 6 7
6wðT 3 Þ7 ¼ 1þð1nÞk0 6nðE1 ;T 3 Þ nðE2 ;T 3 Þ nðEL ;T F Þ7 6f ðE3 Þ7 ð9Þ
w 0 0 0 w 6
6
7 >
7 >
6
6
7>
7>
6
6
7
7
4 >
5 > 4     5> > 4 5
>
> >
>
where w is the weight of volatile content remaining and w⁄ is the : ;
wðT F Þ nðE1 ;T F Þ nðE2 ;T F Þ  nðEL ;T F Þ f ðEL Þ
total volatile content. k0 is the pre-exponential factor and f(E)dE is
the distribution that characterizes the activation energy. To model w
¼ f1 þ ð1  nÞk0 ngð1nÞ f
1

the slow weight loss that happens after the rapid loss in case of ð10Þ
w
the microalgae, the nth order model is selected over the exponential
Here, T1 and TF represent the values of minimum temperature and
model stated above. The nth order model with distributed activa-
maximum temperature, respectively in an experiment. Though only
tion energy can be expressed as follows:
k0 and n can be seen from Eq. (10), mE and r are the other two
Z 1 Z   ð1nÞ
1
t
w E parameters implicit in activation energy distribution vector, f.
¼ 1 þ ð1  nÞ k0 exp  dt f ðEÞdE where n – 1
w 0 0 RT Now from Eq. (10), it is possible to form an objective function which
ð2Þ can be optimized to estimate those four parameters.
The objective function is defined by the difference between the
As the distribution is assumed to be Gaussian for this model, it can
experimental and the model predicted data:
be expressed as:
( !) j¼F  2
Z 1 Z 
 ð1n Þ X
1
w t
E 1 ðE  mEÞ2 wj ðexpÞ wj ðestÞ
¼ 1 þ ð1  nÞ k0 exp  dt pffiffiffiffiffiffiffiffiffiffiffiffi exp  dE O:F: ¼  ð11Þ
w 0 0 RT 2pr2 2r 2
j¼1
w w
ð3Þ
The wj(exp) and wj(est) denotes the experimental and estimated
where mE is the mean activation energy and r is the standard deviation weight, respectively and i denotes the corresponding temperature.
of distribution. In this case it is assumed that the distribution remains To minimize the objective function, the algorithm checks every
the same for the whole pyrolysis process. To develop the algorithm, the possible values of mass fraction within the given activation energy
equation has been represented by multiplication of two functions: and temperature range. As there are four unknown parameters to
Z 1
w be evaluated, a new minimization algorithm has been used. The
¼ ZðE; tÞf ðEÞdE ð4Þ
w 0 algorithm is known as Multistart. Unlike other algorithms such as
where Z(E,t) contains the pre-exponential factor, k0 and the order n. direct search, pattern search or genetic algorithm, it tries multiple
The Z(E,t) varies with the time and activation energy. If the heating starting points for global optimization. The Global Optimization
rate is a K/min, the Z(E,T) can be equated to: Toolbox of Matlab with parallel computing facility was used to
 Z T   ð1n
1
Þ accelerate computing speed. The curve fitting method for the
k0 E Multistart object was selected to be lsqcurvefit based on the least
ZðE; TÞ ¼ 1 þ ð1  nÞ exp  dT where T ¼ T 0 þ at ð5Þ
a RT
0 square method. The code was generated in Matlab and a custom-
This limit for the integration is valid as the value of T0 can be ized number of starting points for parameter estimation was used.
selected low enough where no reaction occurs (Miura, 1995). To Since activation energies for algae were reported in the literature
simplify the calculation, the exponential integral in Z(E,T) is evalu- as 60 kJ/mol and 322 kJ/mol (Li et al., 2011; Tang et al., 2011). The
ated by asymptotic approximation when E/R ? 1 (Burnham and expected range of activation energy was selected from 10 kJ/mol
Braun, 1999): to 500 kJ/mol) to form the matrix n and the energy distribution
vector f. The temperature range was selected as required for
478 K. Kirtania, S. Bhattacharya / Bioresource Technology 107 (2012) 476–481

Table 1 5 and 10 K/min were analyzed and used to generate parameters for
Ultimate analysis of the fresh water alga, C. humicola (%db). the model.
C H N S O Ash The model parameters estimated from the algorithm were used
33.16% 5.58% 4.8% 2.42% 27.24% 26.8%
to generate a weight loss curve which is compared with the
experimental one by the fitness parameter. The fit between the
experimental and estimated data can be found out by the fitness
completion of pyrolysis. By defining the ranges of activation energy parameter stated below:
and temperature, the two most important vectors were defined for Pj¼F wjðexpÞ 
wjðestÞ 2
the model. To avoid multiplicity of estimated parameters (Lakshma- j¼1 w
 w
R2 ¼ 1  ð12Þ
nan et al., 1991) decomposition data at two different heating rates, Pj¼F wjðexpÞ 2
j¼1 w

Fig. 1. Pyrolysis behavior of C. humicola at different heating rates.

Fig. 2. Rate of weight loss of C. humicola at different heating rates.


K. Kirtania, S. Bhattacharya / Bioresource Technology 107 (2012) 476–481 479

(a)

(b)

Fig. 3. Comparison of experimental and estimated data from the proposed algorithm at different heating rates.

Table 2 3. Results and discussion


Estimated model parameters at different heating rates.

Rate (K/min) mE (kJ/mol) r k0 (s1) n R2 3.1. Ultimate analysis


5 189.15 14.73 1.1291  1016 7.88 0.999354
10 190.02 14.73 2.6021  1016 6.63 0.999569 The ultimate analysis was performed on the algae to observe
the amount of its elemental constituents. The results are shown
in Table 1. In the ultimate analysis, the alga was found to have a
carbon content of 33% by weight on dry basis. Also the ash content
The closer the value of R2 to 1, the better is the fit. The next step
was very high for the algae.
of the validation continued with testing the estimated model
parameters over a new pyrolysis experiment performed at 20 K/ 3.2. Pyrolysis
min. This data was not used for any model parameter generation.
The raw data from the experiment was compared with the pyroly- The pyrolysis characteristics of the alga at different rates are
sis curve generated from the generalized model parameters from shown in Fig. 1. It was found from the experiments that the
previous data. pyrolysis of fresh water alga was complete by 550 °C. As it is a kind
480 K. Kirtania, S. Bhattacharya / Bioresource Technology 107 (2012) 476–481

Fig. 4. Distribution of activation energy at various heating rates during pyrolysis of C. humicola.

Fig. 5. Comparison of experimental and predicted weight loss data by the proposed algorithm.

of biomass, early devolatilization occurs. There was a rapid loss of rate curves were also observed during the pyrolysis of three kinds
its weight which slowed down after 500 °C. The rate curves (Deriv- of red algae by Li et al. (2011). This indicates that the dominating
ative of Thermogravimetric curve) of algae pyrolysis in Fig. 2 reactions changed after the rate change. Anastasakis et al. (2011)
showed the two steps in the rate curve during pyrolysis of algae. studied the pyrolysis behavior of four main components of algae
It can be seen that pyrolysis started around 200 °C and the rapid and found that they decomposed at different temperatures.
part of pyrolysis ended at about 365 °C and then the step com- After 525 °C, the weight loss continued at a reduced rate up to
pleted at around 525 °C. The same behavior was observed for all 850 °C. This tail at the end of the pyrolysis accounted for about
the heating rates. The rate curves were consistent in the sense that 5.2% of the total weight loss which was a significant percentage
the rates increased between 200 °C and 525 °C with increasing the of total volatile content of the alga. This can be denoted as ‘end
heating rate. This rate increase was proportional to the increase in loss’ for algae pyrolysis. Including the ‘end loss’, the total weight
heating rate. Also two steps were found in each rate curve. Similar loss for C. humicola was found to be 55.6%.
K. Kirtania, S. Bhattacharya / Bioresource Technology 107 (2012) 476–481 481

3.3. Modeling biomass yield, methane production and combustion. Bioresource Technology
102, 2595–2604.
Burnham, A.K., Braun, R.L., 1999. Global kinetic analysis of complex materials.
Fig. 3 shows the weight loss curve of the algae along with the Energy and Fuels 13 (1), 1–22.
modeled weight loss at two different heating rates. The mean acti- Cai, J., Yang, S., Li, T., 2011. Logistic distributed activation energy model – Part 2:
vation energy, standard deviation, pre-exponential factor and the Application to cellulose pyrolysis. Bioresource Technology 102 (3), 3642–3644.
Chaichalerm, S., Pokethitiyook, P., Yuan, W., Meetam, M., Sritong, K., Pugkaew, W.,
order at different heating rates are listed in Table 2. The values Kungvansaichol, K., Kruatrachue, M., Damrongphol, P., 2011. Culture of
of the fitness factor, R2 for the curves were determined to be great- microalgal strains isolated from natural habitats in Thailand in various
er than 0.999 (Table 2) for all the heating rates. So it is a quite good enriched media. Applied Energy. doi:10.1016/j.apenergy.2011.07.028.
Demirbas, M.F., 2011. Biofuels from algae for sustainable development. Applied
fit with only one set of reactions of n order with single activation Energy 88 (10), 3473–3480 (special issue of energy from algae: current status
energy distribution. Ideally, the values of estimated parameters and future trends).
at different heating rates should be similar, but practically that Flynn, J.H., Wall, A.L., 1966. A quick direct method for determination of activation
energy from thermogravimetric data. Journal of Polymer Science Part B:
does not happen. As the pre-exponential factor is not truly inde- Polymer Letters 4 (5), 323–328.
pendent of the temperature and counting all the experimental er- Harun, R., Danquah, M.K., 2011. Enzymatic hydrolysis of microalgal biomass for
rors, the values of the parameters were quite close. To generalize bioethanol production. Chemical Engineering Journal 168, 1079–1084.
Kissinger, H.E., 1957. Reaction kinetics in differential thermal analysis. Analytical
the parameters for the species of fresh water algae, average values Chemistry 29 (11), 1702–1706.
of the parameters were calculated to be mE = 189.59 kJ/mol, Lakshmanan, C.C., Benett, M.L., White, N., 1991. Implications of multiplicity in
r = 14.73, k0 = 1.8656  1016 s1 and n = 7.2577. The distributions kinetic parameters to petroleum exploration: distributed activation energy
models. Energy and Fuels 5, 110–117.
of activation energy at two heating rates along with average one
Li, D., Chen, L., Zhang, X., Ye, N., Xing, F., 2011. Pyrolytic characteristic and kinetic
is shown in Fig. 4. The model successfully considered the last studies of three kinds of red algae. Biomass and Bioenergy 35, 1765–1772.
10% as loss of volatiles. With ±3r distribution of activation energy, Miura, K., 1995. A new and simple method to estimate f(E) and k0(E) in the
the activation energy varied from 145.4 kJ/mol to 233.78 kJ/mol for distributed activation energy model from three sets of experimental data.
Energy and Fuels 9, 302–307.
decomposition of algae. Miura, K., Maki, T., 1998. A simple method for estimating f(E) and k0(E) in the
A comparison of model predicted and experimental data at distributed activation energy model. Energy and Fuels 12, 864–869.
20 K/min is shown in Fig. 5. The estimated curve fit the experimen- Navarro, M.V., Aranda, A., Garcia, T., Murillo, R., Mastral, A.M., 2008. Application of
the distributed activation energy model to blends devolatilisation. Chemical
tal data with a R2 value of 0.999479. The experiment at 20 K/min Engineering Journal 142 (1), 87–94.
was repeated and results are shown in Fig. 5 as dotted line to check Ozawa, T., 1965. A new method of analysing thermogravimetric data. Bulletin of the
the reproducibility of the experiment. From this validation, it is Chemical Society of Japan 38, 1881–1886.
Pimentel, D. (Ed.), 2008. Biofuels, Solar and Wind as Renewable Energy Systems:
evident that the developed algorithm closely predicted the pyroly- Benefits and Risks. Springer, New York.
sis behavior of C. humicola. Popescu, C., 1996. Integral method to analyze the kinetics of heterogeneous
reactionsunder non-isothermal conditions a variant on the Ozawa–Flynn–Wall
method. Thermochimica Acta 285 (2), 309–323.
4. Conclusions Ross, A.B., Anastasakis, K., Kubacki, M., Jones, J.M., 2009. Investigation of the
pyrolysis behaviour of brown algae before and after pre-treatment using py-gc/
Pyrolysis behavior of a fresh water alga has been investigated at ms and tga. Journal of Analytical and Applied Pyrolysis 85 (1–2), 3–10.
Saddawi, A., Jones, J.M., Williams, A., Wjtowicz, M.A., 2010. Kinetics of the thermal
different heating rates by TGA. To describe the slow weight loss decomposition of biomass. Energy and Fuels 24 (2), 1274–1282.
kinetics at the end of the pyrolysis, a new nth order algorithm Scott, S.A., Dennis, J.S., Davidson, J.F., Hayhurst, A.N., 2006. An algorithm for
based on Gaussian distribution has been developed and used for ki- determining the kinetics of devolatilisation of complex solid fuels from
thermogravimetric experiments. Chemical Engineering Science 61 (8), 2339–
netic parameter estimation. The algorithm was further tested by 2348.
predicting weight loss at another heating rate which showed very Sonobe, T., Worasuwannarak, N., 2008. Kinetic analyses of biomass pyrolysis using
good agreement with the experimental data. The new algorithm is the distributed activation energy model. Fuel 87 (3), 414–421.
Tang, Y.T., Ma, X.Q., Lai, Z.Y., 2011. Thermogravimetric analysis of the combustion of
useful for estimating kinetic parameters from only two pyrolysis microalgae and microalgae blended with waste in N2/O2 and CO2/O2
experiments; these parameters can then be used for predicting atmospheres. Bioresource Technology 102, 1879–1885.
pyrolysis kinetics at various heating rates for C. humicola. Weber, R., 2008. Extracting mathematically exact kinetic parameters from
experimental data on combustion and pyrolysis of solid fuels. Journal of the
Energy Institute 81 (4), 226–233, 8.
References Wjtowicz, M.A., Bassilakis, R., Smith, W.W., Chen, Y., Carangelo, R.M., 2003.
Modeling the evolution of volatile species during tobacco pyrolysis. Journal of
Anastasakis, K., Ross, A.B., Jones, J.M., 2011. Pyrolysis behaviour of the main Analytical and Applied Pyrolysis 66 (1–2), 235–261.
carbohydrates of brown macro-algae. Fuel 90, 598–607. Yuan, S., Chen, X., Li, W., Liu, H., Wang, F., 2011. Nitrogen conversion under rapid
Bruhn, A., Dahl, J., Nielsen, H.B., Nikolaisen, L., Rasmussen, M.B., Markager, S., pyrolysis of two types of aquatic biomass and corresponding blends with coal.
Olesen, B., Arias, C., Jensen, P.D., 2011. Bioenergy potential of ulva lactuca: Bioresource Technology. doi:10.1016/j.biortech.2011.08.047.