Professional Documents
Culture Documents
Torrefaction of wood
Part 1. Weight loss kinetics
Mark J. Prins *, Krzysztof J. Ptasinski, Frans J.J.G. Janssen
Eindhoven University of Technology, Environmental Technology Group, Department of Chemical Engineering and Chemistry,
P.O. Box 513, 5600 MB Eindhoven, The Netherlands
Received 15 February 2005; accepted 16 January 2006
Available online 3 March 2006
Abstract
Torrefaction is a thermal treatment step in the relatively low temperature range of 225–300 8C, which aims to produce a fuel with increased
energy density by decomposing the reactive hemicellulose fraction. The weight loss kinetics for torrefaction of willow, a deciduous wood type, was
studied by isothermal thermogravimetry. A two-step reaction in series model was found to give an accurate description. For the two steps,
activation energies of 76.0 and 151.7 kJ/mol, respectively, and pre-exponential factors of 2.48 104 and 1.10 1010 kg kg1 s1, respectively,
were found. The first reaction step has a high solid yield (70–88 wt%, decreasing with temperature), whereas less mass is conserved in the second
step (41 wt%). The fast initial step may be representative for hemicellulose decomposition, whereas the slower subsequent reaction represents
cellulose decomposition and secondary charring of hemicellulose fragments. The kinetic model is applied to give recommendations for industrial
torrefaction process conditions, notably operating temperature, residence time and particle size.
# 2006 Elsevier B.V. All rights reserved.
Keywords: Thermogravimetry; Torrefaction; Pyrolysis; Kinetics; Wood; Hemicellulose; Cellulose; Lignocellulosic fuels
0165-2370/$ – see front matter # 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jaap.2006.01.002
M.J. Prins et al. / J. Anal. Appl. Pyrolysis 77 (2006) 28–34 29
because the sample has to be heated from environmental weights varied from 2 to 10 mg. Analysis was carried out with
temperature to the reaction temperature at which the weight a nitrogen flow rate of 20 ml (STP)/min. Ceramic crucibles
loss kinetics is studied. If the heating rate is too high, the results loaded with biomass particles were placed inside the TG
may be affected by heat transfer limitations within the sample. apparatus where the weight was constantly measured. The
If a too slow heating rate is applied, the weight loss that takes temperature was varied from 230 to 300 8C and the heating
places during the warm-up phase is not negligible, which rate from 10 to 100 8C min1. A temperature program
complicates the interpretation and deduction of kinetic data. consisting of a dynamic heating period followed by an
However, for the low temperatures applied in torrefaction (225– isothermal heating period was applied. During the dynamic
300 8C), the warm-up phase is relatively short, even for low heating period the biomass was thoroughly dried. The weight
heating rates of 10–100 K/min common in TGA equipment and loss graphs of the dry biomass were used for determination of
the error introduced by the weight loss during this phase is reaction kinetics.
acceptable. Due to variations in ash content of the various samples, the
An important research parameter is the type of biomass, the measured weights were corrected in the following way:
composition of which determines its behaviour in the torrefaction
process. Deciduous wood types as well as coniferous wood types Mt WTGA Wash
¼ (1)
are considered. These types are often referred to as hard wood M0 exp Winitial Wash
and soft wood, respectively. There is not much difference in the
distribution of the three wood fractions between these types. where Winitial is the solid weight of the dry sample, Wash the
Coniferous wood may contain slightly more lignin than weight of the ash in the dry sample and WTGA is the solid weight
deciduous wood (25–35 wt% versus 18–25 wt%), slightly less that is measured as a function of time.
cellulose (35–50 wt% versus 40–50 wt%) and, on average,
comparable amounts of the hemicelluloses (20–32 wt% versus 3. Results and discussion
15–35 wt%) [6]. However, the composition of the polysugars that
form the hemicelluloses fraction is very different. In the case of 3.1. Thermal degradation of different biomass types
deciduous wood, the hemicelluloses contain 80–90 wt% of 4-O-
methyl glucoronoxylan (referred to as xylan) whereas they Fig. 1 shows the weight loss curves of the various biomass
contain 60–70 wt% of glucomannan and 15–30 wt% of types at 248 and 267 8C obtained from isothermal TGA
arabinogalactan for coniferous wood [6]. The thermal behaviour experiments. The weight loss observed during heating of the
of these components may be different. Therefore, it is an sample from 200 8C, the temperature at which thermal
important question whether the thermal behaviour of deciduous decomposition begins to occur, to the required temperature
and coniferous wood types is the same, i.e. do they react with is relatively small, except for xylan. From Fig. 1 it can be
comparable rates in the torrefaction temperature range. concluded that xylan, the main hemicellulose component of
Furthermore, weight loss kinetics of wood is compared to wood deciduous wood, is the most reactive component. It starts
components such as xylan. decomposing around 200 8C and has a high weight loss. The
Finally, the kinetic model obtained in this work is applied to cellulose decomposition rate is very low in the temperature
make recommendations for industrial torrefaction process range used. This is in agreement with results of other studies,
conditions, such as temperature, residence time and particle size. e.g. Gaur and Reed [7] mention that cellulose may
depolymerise at low temperatures, but volatilization of the
2. Experimental cellulose is very slow below 250 8C. At 267 8C, limited weight
loss of cellulose is found.
2.1. Materials The hardwoods beech and willow have comparable
reactivity. The weight loss observed lies between that of
Several biomass types have been tested in the experiments: xylan and cellulose, which was expected as wood contains
deciduous wood (beech and willow), coniferous wood (larch) these fractions (and other fractions, such as lignin). High xylan
and straw. The deciduous wood types were obtained from content also explains the relatively high weight loss of wheat
Rettenmaier, Germany, the coniferous wood from Praxis, straw, although catalytic effects due to the presence of mineral
Netherlands and wheat straw was taken from a field in North- matter could also play a role. Finally, the coniferous larch
Holland. Particle sizes were in the range of 0.7–2.0 mm in all reacts a lot slower than the hardwoods. Although some
cases, except for straw where it was <5 mm. researchers have attributed the lower weight loss for
Model compounds of wood, such as cellulose and 4-O-methyl coniferous wood to higher lignin content, Ramiah [8] has
glucuronoxylan, extracted from oak spelt, were also used. These experimentally shown that differences in composition of the
were purchased from Sigma–Aldrich in powdered form. hemicellulose fractions in deciduous and coniferous wood
(xylan- and mannan-based, respectively) may be the main
2.2. Equipment and procedures explanation.
As a preliminary conclusion, hardwood looses considerably
TGA measurements were carried out with a Perkin-Elmer more weight than softwood and therefore a higher increase in
Pyris 6 TG apparatus with autosampler. The analysed sample energy density (J/g) may be expected, a topic that is studied in
30 M.J. Prins et al. / J. Anal. Appl. Pyrolysis 77 (2006) 28–34
k1
A!y1 B þ ð1 y1 ÞV1 (2a)
k2
B!y2 C þ ð1 y2 ÞV2 (2b)
Fig. 1. TGA of various biomass compounds, (a) at 248 8C and (b) at 267 8C.
Heating rate 10 8C min1, particle size 0.5–2 mm; dotted line is the heating
curve. (3)
the second paper. This makes hardwood an attractive feedstock The solid yields are given by:
for torrefaction processes. In the rest of this section, a kinetic
k1
model is therefore developed for torrefaction of willow, a y1 ¼ (4a)
classical hardwood. k1 þ kV1
k2
3.2. Kinetic model for willow torrefaction y2 ¼ (4b)
k2 þ kV2
Because the pyrolysis of wood is a very complex set of When the activation energy of kV1 (respectively, kV2 ) is higher
reactions, simplified kinetic models are generally applied. For than that of k1 (respectively, k2), the solid yield drops with
pyrolysis of cellulose and biomass at temperatures higher than increasing temperature. So the introduction of separate forma-
300 8C, the Broido–Shafizadeh model [9,10] is the mostly used, tion of different product classes, which may be questioned from
consisting of an activation step followed by two subsequent the viewpoint of analytical chemistry [17], is a mathematical
reactions. Other models consisting of three first-order approximation to describe the fact that the ratio between
consecutive reactions or of a simple first-order reaction (global formed solids and volatiles depends on temperature.
one step) have been proposed as well [11,12]. For hemi- For this research, it was decided to verify whether a two-step
cellulose decomposition, a global one-step reaction gives a very mechanism (as given in Eq. (3)) is also suitable to describe the
rough impression of kinetics of the overall reaction but it is weight loss kinetics of wood in the torrefaction temperature
difficult to compare parameters consistently with other range. This study does consider solid formation and volatile
researchers [13]. Órfão et al. [14] concluded that hemicellulose formation as parallel reactions and aims to determine k1, k2, kV1
decomposition could not be modeled by simple kinetics. and kV2 . Our aim was to develop a simple model that describes
Research done on multi-step reaction mechanisms gives better the weight loss of willow wood in the temperature range of
results [11,12,15,16]. Várhegyi et al. [11,12] have studied the torrefaction—without exact knowledge of the wood composi-
reaction kinetics of thermal decomposition of xylan by tion. Other wood types or biomass types with different
dynamic TGA–MS. They found a model of competitive composition will have different char yields, and also different
reactions could not describe their data, and proposed a model ‘kinetic’ parameters. Therefore, the char yields and reaction
M.J. Prins et al. / J. Anal. Appl. Pyrolysis 77 (2006) 28–34 31
rates of the first and second step for willow wood are compared
with literature data for xylan.
The differential rate equations are given for the solids by
Eqs. (5a)–(5c).
d½A
rA ¼ ¼ ðk1 þ kV1 Þ½AnA (5a)
dt
d½B
rB ¼ ¼ k1 ½AnA ðk2 þ kV2 Þ½BnB (5b)
dt
d½C
rC ¼ ¼ k2 ½BnB (5c)
dt Fig. 2. Experimental weight loss curve for willow decomposition at various
final temperatures. Starting weight defined at 200 8C; time includes a heating
Similar equations may be derived for the volatile products. period of 3–10 min (heating rate = 10 8C min1).
If all reactions are assumed to be of first order, the system of
equations can be solved analytically. Integration of the reaction, as is the case for xylan decomposition, a demarcation
differential equations, with the initial condition that only A time exists after which the first reaction is completed and the
is present at the beginning of the reactions, and addition gives second one starts. For decomposition of willow wood, it was
an expression for the relative solid weight: possible to establish a demarcation time at temperatures in
excess of 250 8C. From this temperature, it seems that
M ¼ ½A þ ½B þ ½C (6) secondary charring reactions begin to occur.
To determine the reaction orders, a plot of the logarithm of
Mt k1 K1 k1 k2 K1 t k1 K2 þ k1 k2 K2 t the non-dimensional solid weight versus time is shown in Fig. 3
¼ 1þ e þ e (for clarity, experimental data points are shown with 5–10 min
M0 K1 ðK2 K1 Þ K2 ðK2 K1 Þ
intervals). Straight lines are observed before and after the
k1 k2
þ demarcation time, so that a model with two first-order reactions
K1 K2 is valid. This is advantageous as it gives an analytical
(7) expression for the weight loss, whereas heat transfer limitation
where
calculations become a lot more complicated for higher order
K1 ¼ k1 þ kV1 (8a) models.
The kinetic parameters could be determined by a graphical
K2 ¼ k2 þ kV2 (8b)
method or by a numerical method. In the graphical method, a
The final char yield, which can be determined experimen- demarcation time t* should be determined at each experimental
tally, is equivalent to the last term in Eq. (7): temperature. The sample weight at this time is assumed to
Ct ! 1 k1 k2 consist only of intermediate pseudo-component B (as A has
¼ ¼ y1 y2 (9) reacted away and C has not formed yet), i.e. B*. However, for
M0 K1 K2
For higher reaction orders, e.g. a second-order reaction fol-
lowed by a first-order sequential reaction, the system of equa-
tions is non-linear and of stiff nature. This system can be solved
numerically using a suitable algorithm, e.g. in Matlab Release
12, a one-step solver based on a modified Rosenbrock formula
of order two may be applied.
Fig. 5. Experimental and modeled relative weight of willow vs. time, with varying heating rates and final temperature of 260 8C. Bottom lines indicate sample
temperatures.
the required heating time is less than a minute and the The maximum yield of B, which is also shown in Fig. 6,
associated weight loss less than 1%. decreases with temperature since more volatiles are formed at
higher temperature. This may be favourable if these volatiles
3.6. Recommendations for industrial torrefaction process consist of species that carry little energy, such as water vapour
conditions or carbon dioxide. However, at temperatures above 300–
320 8C, fast thermal cracking of cellulose may cause tar
The kinetic data may be used to make recommendations for formation, so that operation below 300 8C is recommended.
industrial torrefaction process conditions and for process scale- This also ensures a good selectivity to the intermediate product;
up. In torrefaction, the fast initial reaction, in which the highly the selectivity to B decreases with temperature because the
reactive hemicellulose is decomposed, is important. This activation energy of the secondary reactions is higher than that
reaction stage should go more or less to completion. The second of the primary reactions.
reaction stage, which could represent cellulose decomposition Apart from temperature and residence time, another
and secondary charring reactions of hemicellulose reaction important process parameter for torrefaction is the particle
products, takes much more time to complete, which may lead to size. To ensure the minimum residence time, it is important that
uneconomically large equipment. Fig. 6 shows the time the rate of heat transfer to and within the particle is fast, as
required to obtain the maximum yield of intermediate product compared with the reaction rate. This means that the solid
B. This time reduces from over 3 h at 230 8C to 10 min at temperature will be essentially the same as the surrounding
300 8C. The latter reaction time appears short enough for temperature, and the overall controlling factor is the intrinsic
commercially viable processes, since the estimated reactor kinetics. The ratio between heat convection and conduction
volume may be fitted into a single reactor even at large rates is expressed by the Biot number:
capacities up to 10 kg s1. arp
Bi ¼ (12a)
l
while the ratio of heat convection rate and reaction rate is
expressed by the (external) Pyrolysis number [21]:
a
Py0 ¼ (12b)
K1 rcp rp
Acknowledgements
References
Fig. 7. Biot number and Pyrolysis number (for reaction at 250 and 300 8C) as a
function of biomass particle diameter. Biomass properties: l = 0.1 W m1 K1, [1] F. Shafizadeh, Pyrolytic reactions and products of biomass, in: R.P.
r = 1500 kg m3 and cp = 2000 J kg1 K1. External heat transfer coefficient Overend, T.A. Milne, L.K. Mudge (Eds.), Fundamentals of Biomass
a = 100 W m2 K1. Thermochemical Conversion, Elsevier, London, 1985, pp. 183–217.
[2] J.P. Bourgeois, J. Doat, H. Egnéus, A. Ellegård, Bioenergy 84, Biomass
Conversion, vol. III, Elsevier Applied Sciences, London, 1984, pp. 153–
and (b) external Pyrolysis number much larger than unity, 159.
which indicates that heat convection to the particle is much [3] R. Pentanunt, A.N.M. Mizanur Rahman, S.C. Bhattacharya, Energy 15
faster than the chemical reactions taking place. Fig. 7 shows (1990) 1175–1179.
[4] P. Girard, N. Shah, Recent developments on torrefied wood, an alternative
these parameters at reaction temperatures of 250 and 300 8C for to charcoal for reducing deforestation, REUR Tech. Ser. 20 (1991) 101–
an external heat transfer coefficient of 100 W m2 K1. At a 114.
Biot number of unity, equivalent to a particle size of 2 mm, the [5] E.S. Lipinsky, J.R. Arcate, T.B. Reed, Fuel Chem. Div. Preprints 47 (2002)
Pyrolysis number is sufficiently high. For pyrolysis of wood 408–410.
[6] R. Wagenführ, C. Scheiber, Wood Atlas (in German), VEB Fachbuch-
below 300 8C, it follows that the reactions are the rate limiting
verlag, Leipzig, 1974.
step, provided that the particle size is sufficiently small and [7] S. Gaur, T.B. Reed, Thermal Data for Natural and Synthetic Materials,
internal heat transfer does not play a role. Marcel Dekker, New York, 1998.
[8] M.V. Ramiah, J. Appl. Pol. Sci. 14 (1970) 1323–1337.
4. Conclusions [9] A. Broido, Kinetics of solid phase cellulose pyrolysis, in: F. Shafizadeh,
K.V. Sarkanen, D.A. Tillman (Eds.), Thermal Uses and Properties of
Carbohydrates and Lignins, Academic Press, New York, 1976 (Chapter
The kinetics of torrefaction reactions in the temperature 19).
range of 230–300 8C can be described accurately by a two-step [10] A.G.W. Bradbury, Y. Sakai, F. Shafizadeh, J. Appl. Polym. Sci. 23 (1979)
mechanism. The first step is much faster than the second step, 3271–3280.
so that these steps can be demarcated in time. The first step is [11] G. Várhegyi, M.J. Antal, T. Szekely, P. Szabo, Energy Fuels 3 (1989) 329–
representative of hemicellulose decomposition, while the 335.
[12] G. Várhegyi, M.J. Antal, E. Jakab, P. Szabó, J. Anal. Appl. Pyrolysis 42
second step represents cellulose decomposition. The solid (1997) 73–87.
yield for the first step is higher than for the second step: 70–88% [13] C. Di Blasi, M. Lanzetta, J. Anal. Appl. Pyrolysis 40–41 (1997) 287–303.
(decreasing with temperature) versus 41%. This may be [14] J.J.M. Órfão, F.J.A. Antunes, J.L. Figueiredo, Fuel 78 (1999) 349–358.
explained because deciduous wood, such as willow, contains [15] S.M. Ward, J. Braslaw, Combust. Flame 61 (1985) 261–269.
less xylan (the reactive component in its hemicellulose fraction) [16] C.A. Koufopanos, G. Maschio, A. Lucchesi, Can. J. Chem. Eng. 67 (1989)
75–84.
than cellulose. [17] M.J. Antal, Biomass pyrolysis: a review of the literature. Part II—
Comparison of the kinetic parameters of willow with those of Lignocellulose pyrolysis, in: K.W. Boer, J.A. Duffie (Eds.), Advances
xylan shows that xylan reacts approximately one order of in Solar Energy, vol. 2, American Solar Energy Society, Boulder, CO,
magnitude faster than willow, a classical hardwood containing up 1985, pp. 175–255.
to 30% xylan. It is recommended to determine reaction kinetics [18] C. Branca, C. Di Blasi, J. Anal. Appl. Pyrolysis 67 (2003) 207–219.
[19] B.M. Wagenaar, W. Prins, W.P.M. van Swaaij, Fuel Process. Technol. 36
also for softwood, such as larch. However, initial experiments (1993) 291–298.
have shown that softwood is much less reactive and therefore [20] S.S. Alves, J.L. Figueiredo, Chem. Eng. Sci. 44 (1989) 2861–2869.
lower rate constants can be expected. Research into thermal [21] D.L. Pyle, C.A. Zaror, Chem. Eng. Sci. 39 (1984) 147–158.