J. Anal. Appl. Pyrolysis 72 (2004) 9–15

Kinetics of the pyrolysis and combustion of olive oil solid waste

Jaakko Jauhiainen, Juan A. Conesa, Rafael Font, Ignacio Martın-Gullón∗

Chemical Engineering Department, Universidad de Alicante, P.O. Box 99, 03080 Alicante, Spain

Accepted 23 January 2004

Available online 9 April 2004

Abstract

The pyrolysis and combustion of solid residues from olive oil processing were studied by dynamic TG–DTG at heating rates between 5 and20◦C min−1 at atmospheric pressure. Two different atmospheres were used: on the one hand, an inert atmosphere (He) in order to study thepyrolysis of the material, and on the other hand an oxidative atmosphere (He:O2 in different ratios) to study its combustion. Pyrolysis followsa two parallel and independent fractions model, with kinetic parameters typical of holocelulose (12%) and lignin (48%). Combustion adds athird reaction to the model, which is cocurrent, due to the combustion does not start until the main devolatilization is finished, attaining kineticparameters and reaction orders typical of a gasification reaction.© 2004 Elsevier B.V. All rights reserved.

Keywords: Olive oil; Lignocellulosic; Pyrolysis; Kinetics

1. Introduction

The world production of olive oil was estimated to beabout 1.7 billion l (statistics from year 2000) yearly, produc-ing also a similar amount of a solid waste. Spain is the worldleader producer country with a share of approximately 30%[1,2].The composition of the olives depends on the olivevariety, the soil and the climate, but on an average, olivesare composed by pulp (70–85%), stone (9–27%), and seed(2–3%), and from the biochemical point of view, the overallcontent is divided as follows: 18–25% oil and grease, 50%water, 20% carbohydrates, 6% cellulose, and 1.5% proteins.The same qualities affect also on the final properties of theoil.

Olives are first processed in virgin olive oil plants, wherethe virgin olive oil is separated from the rest (solid cake, in-cluding stones, and non-organic particles) by centrifugation.The virgin olive oil is then sent to either consumption or torefine treatments to reduce acidity. Solid cake still containsolive oil, so it is sent to extraction plants, where the cake isfirst dried in rotary kilns, and then extracted with hexane,where the oil passes to the organic phase. Meanwhile thehexane is recovered and recycled for the process, there is asolid residue, formed mainly by exhausted pulp and stones,

∗ Corresponding author. Tel.:+34-96-5903867; fax:+34-96-5903826.E-mail address: gullon@ua.es (I. Martın-Gullon).

which forms the waste pomace ororujillo. This waste, whichis commonly used as a self fuel for solid cake extractionplants, is usually accumulated at open air in plants until itis further used, frequently provoking autofiring.

Recently, the Spanish leading electricity company Endesahas built two power plants where this solid residue is usedas the main fuel, for a total power of 16 MW each plant.Since the liquor waste from olive has high toxicity level forthe fauna aquatic environment, and its disposal is not per-mitted, the combustion of an olive oil solid waste producedconcerns. This work is inside a bigger project to study thecombustion of this waste.

Thermogravimetric analysis (TGA) is one of the mostcommonly used techniques to study the primary reactions ofdecomposition of solids. In the case of carbonaceous mate-rials (lignocellulosics, plastics, municipal solid wastes. . . ),thermogravimetric techniques have been used for the iden-tification of the different fractions of polymers present inthe material, and their proportions[3–8], to determine thedecomposition kinetic constants[9] or as a previous step inrefuse incineration[10,11]. In the same way, TGA is a use-ful tool to study devolatilized charcoal gasification since itprovides accurate data about charcoal reactivities under ac-tive atmospheres (CO2, oxygenated mixtures or steam) (e.g.[12–15]). However, thermogravimetric studies of the rawmaterials under oxidative atmosphere are not so frequent inspite of the information about incineration processes that can

0165-2370/$ – see front matter © 2004 Elsevier B.V. All rights reserved.doi:10.1016/j.jaap.2004.01.003

10 J. Jauhiainen et al. / J. Anal. Appl. Pyrolysis 72 (2004) 9–15

be obtained from this technique, and the number of studieshas increased[8,16–22].

The aim of the present work is to present and discuss thethermogravimetric behavior of the pomace both in inert andoxidative atmospheres. A kinetic scheme able to correlatesimultaneously (with no variation of the kinetic constants)runs performed at different heating rates and different atmo-spheres of reaction is presented.

2. Experimental

The pomace, solid residues from olive oil factories, is ba-sically a biomass formed by lignin, cellulose, and hemicellu-lose. The pomace used in the present study was supplied by“Espuny Castellar”, a solid cake extraction company. Thispomace was extracted for the oil traces, the form in whichit usually is used for co-firing or combustion alone. It hasrelatively low humidity, 7.93 wt.%. The residue is not veryhomogenous; the olive pits are easy to detect amongst thedried pulp. Consequently, the pomace was milled to attainhomogeinity. The elemental analysis of the material is shownin Table 1. This shows that the pomace has a high content ofoxygen which can be estimated by difference to over 40%,and nearly no sulfur content. The net calorific value (NCV)was obtained from a AC-350 calorimetric bomb from LecoCorp. and resulted to be 17 799 kJ kg−1. The chemical anal-ysis of the pomace was carried out according standard pro-cedures[23], attaining a lignin content of 45-wt.% and aholocellulose content of 44-wt.%. This results of a higherlignin content than holocellulose means a biomass naturedifferent to what it is usually common. With the aim of re-serving the original properties, the sample was kept in aclosed container at temperatures below 6◦C.

The thermobalance used in the study was a SETARAM92-16.18 Model TGA. The temperature control was madeby a Pt/Pt–Rh thermocouple located directly below the sam-ple basket. The carrier gas flow used in the experiments wasintroduced in the thermobalance from the upper side of theequipment causing this way a small buoyancy force down-wards. This effect brings a small error to the weight mea-surements, which is corrected with blank runs. Three runs instrict pyrolysis conditions (helium atmosphere) at the sameabove mentioned heating rates were carried out with Avi-cell Cellulose. Grønli et al.[24] suggested that all the TGAequipments used for kinetic studies, should be tested with

Table 1Elemental analysis of pomace

Compound Share (%)

N 1.14C 47.06H 5.68S 0.05Ash 7.78

this reference material, in order to compare the kinetic pa-rameters obtained when fitting the experimental data to asingle reaction decomposition model. This test would assurethat the TGA is valid, and the kinetic parameters have nosistematic errors. Varhegyi and Antal[25] reported valuesof preexponential factor, activation energy and reaction or-der of 2.2 × 1019 s−1, 234 kJ mol−1, and 1.2, respectively,whereas similar values of 1.32×1019 s−1, 248 kJ mol−1, and1.09, respectively, were obtained in the present work.

Different gas-mix compositions were used in this study.Pyrolysis runs were carried out in helium atmosphere, andtwo different oxygen ratios were used in combustion runs:poor oxygen atmosphere was accomplished in a He:O2 =9:1 atmosphere; in other runs a rich oxygen atmosphere wasused with He:O2 = 4:1 (similar to the amount of oxygen inthe air). Experiments at three different heating rates (5, 10,and 20 K min−1) were carried out for each different atmo-sphere.

3. Results and discussion

3.1. Proposed model for pyrolysis

Fig. 1 shows the curves obtained for the three differentatmospheres studied at one of the heating rates (5 K min−1).According to this Figure, the pyrolysis takes place in a broadtemperature range, from 200 up to 500◦C, yielding 40-wt.%of solid residue. On the other hand, in either high or pooroxygen atmosphere, the process is divided in two clear andvisible steps: the first one goes joint with the pyrolysis curve,which means that the first step in the combustion is a thermaldegradation of the initial pomace, as occurs in many cases.The second step corresponds to the oxygen gasification ofthe formed remaining char. The change in weight takingplace in the oxygen gasification part indicates the amountof fixed carbon in the pomace, which can be estimated to be25% of the dry mass.

Maschio et al.[26] studied the degradation of poplar woodcoming to a conclusion that a biomass can be analyzed as asum of its main components. The composition of lignocel-lulosic materials such as pomace can be expressed as

[Biomass]= a[Cellulose]+ b[Lignin] + c[Hemicellulose]

Normally a three fraction model would be the simplestsolution to simulate the decomposition, with a correspond-ing step for all three components. Nevertheless the celluloseand hemicellulose are sometimes combined to holocellulosewhich permits to simplify the model to two different simul-taneous fractions as the pyrolysis curve inFig. 1 implies.In this work, it is assumed that the original pomace is com-posed by two different fractions, the holocellulose-type andthe lignin-type fractions, where each one behaves as thoughthey were pure holocellulose and pure lignin, respectively.This fractions will be used to correlate the kinetic model,but it does not imply that the content of each fraction were

J. Jauhiainen et al. / J. Anal. Appl. Pyrolysis 72 (2004) 9–15 11

0.0

0.2

0.4

0.6

0.8

1.0

100 200 300 400 500 600 700 800T (ºC)

Weight fraction

Pyrolysis

Poor oxygenatmosphereCombustion

Fig. 1. Experimental data of the weight fraction ‘W’ vs. temperature at the three different gas-mix compositions: pyrolysis (He), poor oxygen combustion(He:O2 9:1), and combustion (He:O2 4:1).

pure holocellulose or pure lignin, although the real contentof this type fractions is only related to its parent materials(i.e. mainly holocellulose and mainly lignin). These frac-tions decompose individually according to the next scheme:∣∣∣∣∣∣∣

SHC → VHC + RHC

SLIG → VLIG + RLIG

I → I

∣∣∣∣∣∣∣ (S.1)

In the reaction modelSi represents the initial solid pomacefraction whileVi andRi are used for the volatile and residualsolid parts of each fraction, respectively, and the subscriptsHC and LIG denote holocellulose-type and lignin-type frac-tions. I stands for the inert part that does not react (ash).Since the thermobalance detects the total weight (both unre-acted pomace and the char formed), the difference betweenthe initial weight and the solid fraction is the volatile frac-tion. According to Font et al.[27] the following differentialequations can be written:

−dW

dt= dV

dt= dVHC

dt+ dVLIG

dt(1)

dVHC

dt= kHCV∞HC

(1 − VHC

V∞HC

)nHC

(2)

dVLIG

dt= kLIGV∞LIG

(1 − VLIG

V∞LIG

)nLIG

(3)

whereki is the kinetic constant;ni, the reaction order; andV∞i is the final volatile fraction at time infinity of the com-ponent. The kinetic constant of each fraction can be substi-tuted according to the Arrhenius law:

ki = koi exp

(−Ei

RT

)(4)

wherekoi andEi are the pre-exponential factor and the ap-parent activation energy, respectively. As stated by Martin-

Gullon et al.[28], there is an interrelation between the re-action order and the kinetic constant, which can be takeninto account for the optimization numerical process. Conse-quently, instead of calculating directly the pre-exponentialfactor of each fraction, the following equation was used:

ki = K∗i

0.64niexp

[−Ei

R

(1

T− 1

Tmax

)](5)

whereK∗i is the comparable kinetic constant at a given ref-

erence temperatureTmax, which is the temperature wherethe decomposition rate of material is at its highest.

The integration of the kinetic equations was carried outusing the fourth-order Runge-Kutta method, and the opti-mization used the Generalized Reduced Gradient method,v.2 (GRG2), implemented in the Microsoft Excel spread-sheet. For each of the reactions a total of four differentparameters are used: the comparable kinetic constantK∗

i ,activation energyEi, the reaction orderni and the maximumamount of formed volatilesV∞i. The objective function canbe chosen based on either DTG or TG curves. In this studyboth the TG and DTG data was used in the objective func-tion with the last refining done with the DTG data to obtainthe best possible match. The final objective function was:

OF =∑runs

∑data

[(dW

dt

)exp

−(

dW

dt

)cal

]2

(6)

where ‘exp’ and ‘cal’ stand for the experimental data and thecalculated mathematical model, respectively. It is necessaryto fit at least three TG runs at different heating rates in orderto consider both the kinetic model and calculated parameterspotentially correct[5,6]. Fig. 2arepresents the curves for themodel and the data for the pyrolysis with the three differentheating rates studied.Fig. 2balso presents the DTG curves,used to see clearly some aspects of the TG curve.Table 2

12 J. Jauhiainen et al. / J. Anal. Appl. Pyrolysis 72 (2004) 9–15

0

0.2

0.4

0.6

0.8

1

100 200 300 400 500 600 700 800T (ºC)

Weight fraction

PYROLYSIS

dots --> experimentallines --> calculated

5 ºC/min

10 ºC/min

20 ºC/min

-0.002

-0.0016

-0.0012

-0.0008

-0.0004

0

100 200 300 400 500 600 700 800

T (ºC)

dW/dt

(a)

(b)

Fig. 2. Experimental and calculated data for the pyrolysis at the threedifferent heating rates studied. TG (a) and DTG (b).

Table 2Kinetic constants obtained for the models representing the thermal de-composition of pomace in the three different gas-mix compositions

He 9:1 He:O2 4:1 He:O2

First fraction (mainly holocellulose)k0HC (s−1) 1.402× 1014 7.531× 1011 9.319× 1011

EHC (kJ mol−1) 181.8 153.7 153.7nHC 0.96 1.57 1.57V∞HC 0.118 0.145 0.143

Second fraction (mainly lignin)k0LIG (s−1) 1.236× 104 1.146× 104 1.371× 104

ELIG (kJ mol−1) 69.4 66.4 66.4nLIG 3.79 2.94 2.94V∞LIG 0.479 0.398 0.369

Third fraction (char)k0O2 (s−1) 1.077× 107 1.323× 107

EO2 (kJ mol−1) 133.3 133.3nO2 1.00 0.81V∞O2 0.347 0.348

shows the kinetic parameters optimized for the pyrolysisprocess.

It can be seen that the highest decomposition rate isreached at the surroundings of 330◦C. The higher heatingrate tends to push the peak slightly towards the right. Asit can be seen from the figure, the match obtained with themodel for two simultaneous processes might be consideredcorrect. The values at the most critical stages, that is withthe highest decomposition velocity, are met precisely, whilethe edges of the curves, where the decomposition is weaker,show only slight differences between the experimental dataand those calculated from the model.

The kinetic parameters calculated are compared with datafrom other lignocellulosic materials in the literature. Kalous-tian et al.[29] studied the decomposition of holocellulosederived from several different types of Mediterranean plants.Their results show the activation energy to be between 150and 200 kJ mol−1 in air flow while noticing somewhat highervalues for experiments made with nitrogen. That is in totalaccordance with the values shown inTable 2. The varia-tions in the values of the reaction order can be explained bythe difference in the chemical structure of hemicelluloses,mainly pentosan and hexosan, as the researchers propose.

The values for lignin oscillate more than for the otherbiopolymers. That is due to the influence of pre-treatmentssuch as hexane extraction in the case of pomace as well asinteractions with other components of the biomasses. Ca-ballero et al.[30] showed in their study of pyrolysis oflignin and holocellulose that the lignin decomposes at a rel-atively wide temperature range which starts already beforethe cellulose decomposition and forms a long tail in theTG curve. That is found also in the case of pomace. How-ever, in the case of lignin, the TG curves can vary slightlyas pre-treatments as well as interactions with other presentcompounds interfere with the process as these researchersshow. Caballero et al.[30] pointed out that it was not pos-sible to reproduce the thermal decomposition curves whenlignin and hemicellulose were separated from each otherdespite containing equal amounts as a biomass. Like hemi-cellulose, the lignin also has different chemical structures,which results in notable variations in the kinetic data as wellas TG–DTG curves. However, the values tend to be under100 kJ mol−1 for the apparent activation energy with reac-tion orders reaching values as high as 8.

3.2. Proposed model for combustion

When the inert He gas used in the pyrolysis simulationwas changed to a mix of He and O2 with the objective tostudy the combustion, an expected result of a further oxy-gen gasification step in the TG curves was obtained, as seenin Fig. 1. A third reaction was added to the kinetic modelfor pyrolysis to obtain the one for combustion. The firstassumption of an independent third reaction did not givean expected result and, for that reason, a model where thesolid product from reaction (2) interacts with oxygen was

J. Jauhiainen et al. / J. Anal. Appl. Pyrolysis 72 (2004) 9–15 13

proposed. In this way, new volatile and solid product areformed. In many cases, the oxygen gasification does notstart before the main pyrolysis (i.e. main volatile evolution)is finished[13,14]. Consequently it is reasonable to assumea co-current reaction after pyrolysis. Since cellulose pyrol-ysis individually yielding practically no residue[23], it isassumed in this model that the oxygen gasification only oc-curs with the resulting lignin-type fraction derived char.∣∣∣∣∣∣∣∣

SHC → VHCSLIG → VLIG + RLIG

RLIG → VO2

I → I

∣∣∣∣∣∣∣∣(S.2)

The resulting differential equations in accordance to themathematical model developed by Font et al.[27] are

−dW

dt= dV

dt= dVHC

dt+ dVLIG

dt+ dVO2

dt(7)

dVHC

dt= k0HC exp

(−EHC

RT

)V∞HC

(1 − VHC

V∞HC

)nHC

(8)

dVLIG

dt= k0LIG exp

(−ELIG

RT

)V∞HC

(1 − VLIG

V∞LIG

)nLIG

(9)

dVO2

dt= k0O2 exp

(−EO2

RT

)V∞O2

(VLIG

V∞LIG− VO2

V∞O2

)nO2

(10)

where for each fraction, the kinetic pre-exponential factor,the reaction order, the apparent activation energy and themaximum amount of volatiles are the parameters to opti-mize. It is important to note that in the case of the char gasi-fication, the amount of volatilesVO2 andV∞O2 refer to the

0

0.2

0.4

0.6

0.8

1

100 200 300 400 500 600 700 800

T (ºC)

Weight fraction

He:O2=9:1

Fig. 3. Experimental (dots) and calculated (lines) curves for the decomposition of pomace at 5, 10, and 20◦C min−1 in poor oxygen atmosphere.

mass fraction of carbon that reacts with oxygen, and con-sequently, corresponds to the weight loss detected by thethermobalance. With the help of these equations accordingto the procedure explained before (includingEq. (5)), it waspossible to obtain the calculated values for the combustionin mixture of helium and oxygen, both with fractions 4:1and 9:1 simulating air and impoverished air, respectively.Table 2shows the kinetic parameters obtained.

Fig. 3 shows the obtained curves for the 9:1 He:O2 com-bustion (both experimental and calculated by the model),the solid weight fraction as a function of temperature for allthree heating velocities.Fig. 4corresponds to the case of the4:1 He:O2 combustion. It can be seen that curves are fittedsatisfactorily with the proposed models. Again the objec-tive function used was for the DTG. The first decompositionsteps can be traced to the surroundings of 300◦C, while thegasification step ends at the temperatures close to 500◦C.

One striking data can be picked out from the data inTable 2: the small percentage of holocellulose-type mate-rial present in the pomace, which can be approximated to15 wt.%. In the same way the lignin-type fraction can bethought of to be around 75 wt.%, taking into account thatis was supposed that only the lignin-type fraction yieldedchar. The missing 10% are a combination of inorganic ashforming matter combined with dirt. The relatively low frac-tion of present holocellulose in pomace could be result ofthe pre-treatments such as the overall processes carried outin the extraction plants (such as a high temperature dryingstep). This result is confirmed with the chemical analysisperformed to the pomace, where, contrarily to it is commonfor biomass, the lignin content is higher that the holocellu-lose one.

Other comments on the data are that the activation energyas well as reaction order for holocellulose fraction changesslightly from pyrolysis to combustion while the different

14 J. Jauhiainen et al. / J. Anal. Appl. Pyrolysis 72 (2004) 9–15

0

0.2

0.4

0.6

0.8

1

100 200 300 400 500 600 700 800

T (ºC)

Weightfraction

He:O2=4:1

Fig. 4. Experimental (dots) and calculated (lines) curves for the decomposition of pomace at 5, 10, and 20◦C min−1 in rich oxygen atmosphere.

oxygen amounts seem to have no impulse on them. Thesame effect can be seen for the reaction order at lignin data.

The kinetic constant of the combustionEq. (10)can beexpressed as a function of partial pressure of oxygen presentin the atmosphere.

k3 = k′pm3O2

(11)

wherepO2 is the oxygen partial pressure. Using the valuesobtained for the two conditions gives a value ofm3 = 0.42.

Respect to the char gasification parameters, Haji-Sulaimanand Aroua[31] studied the combustion of mineral carboncoming to a result of 131–137 kJ mol−1 for the activa-tion energy, while Henrich et al.[32] obtained a value of145 kJ mol−1 with soot combustion. The values obtained inthis study fit perfectly in the small range the mineral carbonstudy proposes. Gasification reaction orders close to 1 areindicative that the gasification is taking place under kineticcontrol.

Acknowledgements

The study has received economical help from the SpanishMinistry of Science and Technology (MCYT) project PPQ2002-00567. Authors thank ‘Espuny Castellar’ Co. for sup-plying the raw material. The authors also thank Ms. AraceliGálvez for her valued help in the chemical analysis deter-mination.

References

[1] http://www.aceitedeoliva.com.[2] http://www.oliveoilsource.com/olivewaste.htm.

[3] J.A. Caballero, J.A. Conesa, R. Font, A. Marcilla, J. Anal. Appl.Pyrol. 42 (1997) 159.

[4] J.A. Conesa, A. Marcilla, J. Anal. Appl. Pyrol. 37 (1996) 95.[5] J.A. Conesa, J.A. Caballero, A. Marcilla, R. Font, Thermochim. Acta

254 (1995) 175.[6] J.A. Conesa, A. Marcilla, R. Font, J.A. Caballero, J. Anal. Appl.

Pyrol. 36 (1996) 1.[7] C. Devallencourt, J.M. Saiter, D. Capitaine, Polym. Degrad. Stab. 52

(1996) 327.[8] G. Varhegyi, P. Szábo, F. Till, B. Zelei, M.J. Antal Jr., X. Dai,

Energy Fuels 12 (1998) 969.[9] H. Teng, H.C. Lin, J.A. Ho, Ind. Eng. Chem. Res. 36 (1997) 3975.

[10] H. Bockhorn, A. Hornung, U. Hornung, S. Teepe, J. Weichmann,Combust. Sci. Technol. 116–117 (1996) 129.

[11] K.S. Chen, R.Z. Yeh, J. Hazard. Mater. 49 (1996) 105.[12] I. Martın-Gullón, M. Asensio, A. Marcilla, R. Font, Ind. Eng. Chem.

Res. 35 (1996) 4139.[13] S. Dutta, C.Y. Wen, R.J. Belt, Ind. Eng. Chem. Process Des. Dev.

16 (1977) 20.[14] S. Dutta, C.Y. Wen, Ind. Eng. Chem. Process Des. Dev. 16 (1977) 31.[15] P. Salatino, O. Senneca, S. Masi, Carbon 36 (1998) 443.[16] A.K. Sircar, T.G. Lamod, Rubber Chem. Technol. 48 (1975) 301.[17] T. Cordero, J.M. Rodriguez-Maroto, F. Garcıa, J.J. Rodriguez, Ther-

mochim. Acta 191 (1991) 161.[18] A.E. Ghaly, A. Ergüdenler, Appl. Biochem. Biotechnol. 28–29 (1991)

111.[19] A.E. Ghaly, A. Ergüdenler, Can. J. Chem. Eng. 72 (1994) 651.[20] J.A. Caballero, R. Font, M.M. Esperanza, J. Anal. Appl. Pyrol. 47

(1998) 165.[21] J.A. Conesa, R. Font, A. Fullana, J.A. Caballero, Fuel 77 (1998)

1469.[22] J.A. Conesa, R. Font, Polym. Eng. Sci. 41 (2001) 2137.[23] B.L. Browning, Methods of Wood Chemistry, Wiley, New York,

1965.[24] M. Grønli, M.J. Anta Jr, G. Várhegyi, Ind. Eng. Chem. Res. 38

(1999) 2238.[25] G. Varhegyi, M.J. Antal, Energy Fuels 3 (1989) 329.[26] G. Maschio, A. Lucchesi, C. Koufopanos, Adv. Thermochem.

Biomass Conv., vol. 2, Blakie Academic and Professional, Cam-bridge University Press, Cambridge, 1992, pp. 746–759.

J. Jauhiainen et al. / J. Anal. Appl. Pyrolysis 72 (2004) 9–15 15

[27] R. Font, I. Martın-Gullón, M. Esperanza, A. Fullana, J. Anal. Appl.Pyrol. 58–59 (2001) 703.

[28] I. Martin-Gullon, M.F. Gomez-Rico, A. Fullana, R. Font, J. Anal.Appl. Pyrol. 68–69 (2003) 645.

[29] J. Kaloustian, A.M. Pauli, J. Pastor, J. Therm. Anal. Calorim. 63(2001) 7.

[30] J.A. Caballero, R. Font, A. Marcilla, Thermochim. Acta 276 (1996)57.

[31] M.Z. Haji-Sulaiman, M.K. Aroua, J. Inst. Energy 70 (1997) 52.[32] E. Heinrich, S. Bürkle, Z.I. Meza-Renken, S. Rumpel, J. Anal. Appl.

Pyrol. 49 (1999) 221.