J. Anal. Appl.

Pyrolysis 77 (2006) 3540

www.elsevier.com/locate/jaap

Torrefaction of wood
Part 2. Analysis of products
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 14 February 2005; accepted 16 January 2006
Available online 3 March 2006

Abstract
Torrefaction is a mild pyrolysis process carried out at temperatures ranging from 225 to 300 8C, in which hemicellulose, the most reactive
fraction of wood, is decomposed. Dehydration and decarboxylation reactions cause a mass loss of the wood, whereas the lower heating value of the
wood is largely conserved. Deciduous wood types (beech and willow) and straw were found to produce more volatiles than coniferous wood
(larch), especially more methanol and acetic acid. These originate from acetoxy- and methoxy-groups present as side chains in xylose units present
in the xylan-containing hemicellulose fraction. The torrefied wood product has a brown/black color, reduced volatile content and increased energy
density: 20.7 MJ/kg (after 15 min reaction time at 270 8C) versus 17.7 MJ/kg for untreated willow. It has favorable properties for application as a
fuel for gasification and/or (co-)combustion.
# 2006 Elsevier B.V. All rights reserved.
Keywords: Thermal analysis; Torrefaction; Pyrolysis; Pyrolysis products; Wood; Lignocellulosic fuels

1. Introduction
In the transition to a more sustainable energy supply, there is
a clear need for better biofuels. Untreated woody biomass has a
relatively low energy density, high moisture content and is
difficult to comminute into small particles. These properties
make transport of wood relatively expensive. After drying, it
can regain moisture and may rot during storage. Furthermore,
enhancement of the energy density is advisable because a large
amount of wood is required to replace an equivalent amount of
coal in applications such as combustion and gasification. A
technology that could help to overcome these problems is a
thermal processing step known as torrefaction. In this process,
wood is roasted in the temperature range of 225300 8C under
an inert atmosphere. The weight loss kinetics of the process
have been treated in the first part of the present study. This part
focuses on the products obtained, which consist of three phases:
a solid product of a dark color, an acidic aqueous phase of
yellowish color and permanent gases such as carbon monoxide
and carbon dioxide. The solid product, torrefied wood, has
DOI of original article: 10.1016/j.jaap.2006.01.002.
* Corresponding author. Tel.: +31 40 2473734; fax: +31 40 2446653.
E-mail address: [email protected] (M.J. Prins).
0165-2370/$ see front matter # 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jaap.2006.01.001

attractive fuel properties, such as an improved heating value,

low moisture content and ease of size reduction [1].
Research focused on the torrefaction process was already
performed in France in the 1930s but no significant publications
remain. In the 1980s, the process was pioneered by Bourgois and
Doat [2], who published the results of torrefaction experiments
with two temperate and two tropical wood samples. For the
deciduous wood types, eucalyptus and a mixture of chestnut and
oak, torrefaction at 270275 8C increased the lower heating
value of the wood from 18.6 to 22.7 MJ/kg and 17.9 to 21.5 MJ/
kg, respectively, while retaining over 90% of the energy. The
coniferous wood types, maritime and Caribbean pine, required a
higher temperature of 280285 8C, while the increase in energy
density was slightly less. This research resulted in the building of
a continuous wood torrefaction plant in 1987. Torrefied wood
from this plant was used as a reducer in the production of silicon
metal. Two batch plants were also built for the production of
barbeque fuel and firelighters. Torrefaction of wood has recently
attracted more interest. Pentanunt et al. [3] carried out
torrefaction experiments at the Asian Institute of Technology
in Bangkok. They compared combustion characteristics of
torrefied wood produced after 23 h of torrefaction and
concluded that torrefied wood produces less smoke compared
with untreated wood. Felfli et al. [4] used a bench unit to

36

M.J. Prins et al. / J. Anal. Appl. Pyrolysis 77 (2006) 3540

determine the effect of raw material, temperature and residence

time on the properties of torrefied wood. Similar research has
been carried out at by Nimlos et al. [5] and Pach et al. [6].
In torrefaction, the most reactive fraction of biomass, i.e. the
hemicellulose fraction, is decomposed so that torrefied wood
and volatiles are formed. Gaur and Reed [7] summarize the
hemicellulose degradation as a two-step reaction where light
volatiles (mono- and polysaccharide fractions and dehydrosugars) are formed in the first step followed by their catalytic
degradation (by mineral matter) resulting in the formation of
CO and CO2. Thermal cleavage of carbon bonds of the
carboxylic groups might result in the formation of acids. Acid
catalyzed dehydration and thermal cleavage reactions form
CO2 and carbonyls like hydroxy acetone, methanol, propanal,
etc., most of them consisting of one to three carbon atoms and
one to two oxygen atoms. They differ a lot due to
stereochemistry and internal rearrangements like retro aldol
reactions. When the biomass particles are relatively small the
main part of the acids diffuses out of the particle.
The research described in this paper aims to give a good
insight into the overall mass balance of the process. This
requires analysis of all reaction products, i.e. both the torrefied
wood and the volatile product (the latter is seldom analyzed in
the torrefaction literature). An important research parameter is
the type of wood: both deciduous and coniferous wood types
are considered since the composition of their hemicellulose
fraction differs. The torrefaction behavior of straw is also
tested because it is an abundant and cheap biomass feedstock.
The effect of process conditions such as temperature and
residence time on the yield and composition of solid product
and volatile products is studied. Special emphasis is given
to the fuel properties and the energy conserved in torrefied
wood.

Fig. 1. Schematic set-up of the bench-scale torrefaction unit.

controlled with a Eurotherm master slave controller. Tracing

the inlet and outlet of the reactor prevents condensation of
products in the reactor and tubing. The argon inlet lines are
also traced in order to preheat the gas for a more stable
temperature control. A batch of 10 g of biomass is loaded in
the reactor resulting in a bed height of approximately 5 cm.
The biomass is heated with a heating rate of 1020 8C/min to
the torrefaction temperature. After the isothermal heating
period the reactor is cooled in approximately 5 min with
pressurized air until the temperature is below 200 8C. In order
to prevent condensation of reaction products the tracing is
continued for another 15 min.
2.2. Experimental plan

2. Experimental
The biomass used was beech, willow, larch and straw, the
properties of which are shown in Table 1. Particle sizes were in
the range of 0.72.0 mm in all cases, except for straw where it
was < 5 mm. Experiments were carried out with beech and
willow at temperatures in the range of 220300 8C and
residence times in the range of 1060 min. For comparison, the
coniferous larch wood was torrefied at 230 8C (50 min), 250 8C
(30 min), 270 8C (15 min), 290 8C (10 min) and straw at 250 8C
(30 min). Note that the residence time excludes the heating
time, e.g. to heat the reactor bed from 200 8C (at which the
torrefaction reactions begin to occur) to a torrefaction
temperature of 250 8C takes approximately 8 min.

2.1. Equipment
Fig. 1 presents a schematic view of the experimental setup. The bench-scale torrefaction unit consists of a quartz
fixed-bed reactor placed inside an oven. A continuous flow of
argon is controlled with a Brooks mass flow controller. The
argon flow is used to keep the system inert and remove
volatile products from the reactor. The volatile products are
split into a liquid and gas phase in a cold trap at 5 8C. Gases
are collected in a gasbag. The sample temperature is measured
by a thermocouple in a glass well inside the reactor bed and
Table 1
Properties of biomass types
Moisture (wt%)

Beech
Willow
Larch
Straw

8.1
8.6
9.8
10.7

Dry material (wt%)

Ash

Volatiles

1.2
1.6
0.1
7.1

82.7
81.4
82.8
79.0

47.2
47.2
48.8
44.3

6.0
6.1
6.1
5.8

0.40
0.34
0.10
0.40

45.2
44.8
44.9
42.4

O/C ratio

HHV (dry) (MJ/kg)

LHV (dry) (MJ/kg)

0.72
0.71
0.69
0.72

18.3
19.0
19.5
17.4

17.0
17.7
18.2
16.1

M.J. Prins et al. / J. Anal. Appl. Pyrolysis 77 (2006) 3540

37

2.3. Analyses
All the reaction products were collected and weighed in
order to make an overall mass balance. The solid product, the
produced liquid and the produced gases were analyzed with
different techniques. The composition of the solid product was
determined by proximate and ultimate analysis and its heating
value was measured in a bomb calorimeter. Liquid products
collected in the cold trap were diluted with 2-butanol because
not all the products dissolve in water. The diluted liquid was
analyzed with HPLC using a Chrompack Organic Acids
column with detection based on refraction index. The
composition of produced gas was analyzed with a Varian
Micro GC with a Poraplot Q and a Molsieve column.
3. Results and discussion

Fig. 3. Yield of torrefied wood as a function of temperature and residence time,

for different biomass types; solid lines from kinetic model for torrefaction of
willow at 15 min (upper line) and 30 min (lower line) residence time.

3.1. Overall mass balance

Fig. 2 displays the overall mass balance of several
torrefaction experiments using willow, larch and straw. The
overall mass balance of the torrefaction experiments closed
quite well and was generally in the range of 95100%. This
high accuracy is possible because the solid product, which can
be recovered and weighed accurately, constitutes the largest
proportion of the product yield. The amount of lost material
increases with temperature, which means that experimental
error in recovery and weighing of volatiles is larger than for the
solid.
3.2. Yield of solid product
Fig. 3 shows the amount of solid product recovered, for
beech, willow, straw and larch at various conditions. The yield
of solid product decreases with temperature and, to a lesser
extent, residence time, whereas the yield of volatiles increases
accordingly. Clearly, the extent of devolatilization increases
with temperature and reaction time. The two lines shown in
Fig. 3 have been calculated using the kinetic model for
torrefaction of willow obtained in the first part of this study. The
upper line is for a residence time of 15 min and the lower line
for 30 min. For willow and beech wood, the measured data
points are in the range described by the kinetic model.

Fig. 2. Overall mass balance of several torrefaction experiments.

Therefore, there is good agreement between thermogravimetric

weight loss data and the data from the fixed-bed reactor set-up.
Fig. 3 also shows a number of data points for larch
(coniferous wood) and straw. The observed weight loss for
straw is comparable to willow and beech wood (both deciduous
woods). However, the total weight loss for larch is much
smaller, which is also in agreement with the results in the
previous paper. This is most probably caused because the
hemicellulose fraction in coniferous wood has a different
chemical composition, i.e. the amount of the most reactive
hemicellulose (xylan) is less than for the deciduous wood types
[8]. The hemicellulose fraction of coniferous wood contains a
high proportion of glucomannan, which is less reactive and has
lower weight loss in this temperature range than xylan [9].
3.3. Yield of volatile products
The volatiles consist of a condensable fraction and a noncondensable fraction. For both fractions, the amount formed
increases with torrefaction temperature. Fig. 4a and b shows the
yields of condensable products (weight formed divided by the
dry- and ash-free weight of wood) for willow, respectively,
larch and straw at varying experimental conditions.
Acetic acid and water are the main liquid torrefaction
products of willow, while smaller quantities of methanol,
formic acid, lactic acid, furfural, hydroxyl acetone and traces of
phenol are found. Simkovic et al. [10], who used TG/MS to
study thermal decomposition of xylan have also observed these
compounds. Thus, the products found stem mainly from the
decomposition of the hemicellulose fraction. Straw yields
similar condensable products in comparable quantities as for
willow, but these differ from larch: the weight loss of larch is
mainly due to dehydration and the main acid formed is formic
acid. The difference in product distribution between willow
(and straw) as compared to larch may be explained from
differences in hemicellulose composition. The hemicellulose of
deciduous woods has acetoxy- and methoxy-groups attached to
the polysugars (in particular to xylose units), which form acetic
acid and methanol when the wood is heated to temperatures

38

M.J. Prins et al. / J. Anal. Appl. Pyrolysis 77 (2006) 3540

Fig. 4. Product yields of condensable volatiles formed in torrefaction at

different conditions, for (a) willow and (b) larch and straw.

around 200 8C [11]. The amount of acetic acid formed from

willow increases with temperature: the amount of acetic acid
formed is as much as 13% at 290 8C. For larch wood, the yield
of condensable products is much smaller, in agreement with the
observed lower weight loss, and also increases with temperature.
The composition of the non-condensable volatile product
obtained from torrefaction is depicted in Fig. 5a for willow and
Fig. 5b for larch and straw at various conditions. The amount of
non-condensable volatiles is again higher for willow and straw
than that for larch, which further exemplifies the higher degree
of devolatilization for deciduous, xylan-containing wood. The
permanent gases formed are mainly carbon dioxide and carbon
monoxide. Traces of hydrogen and methane are detected as
well. The formation of carbon dioxide may be explained by
decarboxylation of acid groups in the wood. However, the
formation of carbon monoxide cannot be explained by
dehydration or decarboxylation reactions. The increased
carbon monoxide formation is reported in literature [11] as
the reaction of carbon dioxide and steam with porous char to
carbon monoxide with increasing temperature. Mineral matter
may catalyze such reactions, especially since the gases formed
in the torrefaction experiment using straw contained a
relatively higher amount of carbon monoxide than the
experiments using wood (which contains much less mineral
matter).

Fig. 5. Product yields of non-condensable volatiles formed in torrefaction at

different conditions, for (a) willow and (b) larch and straw.

Fig. 6 shows the gas composition of the non-condensable

product with time (the totals do not add exactly to 100% because
the main components, carbon dioxide and carbon monoxide, are
shown). These results were obtained by torrefaction of larch and
willow at 250 8C and analysis of non-condensable gases after 5,
15 and 30 min. It was found that the ratio of carbon dioxide to
carbon monoxide decreases with time, in line with the theory that
carbon monoxide is formed in a secondary reaction.

Fig. 6. Change in gas composition with time for larch and willow, 250 8C,
30 min.

M.J. Prins et al. / J. Anal. Appl. Pyrolysis 77 (2006) 3540

39

3.4. Properties of the solid product

The torrefied wood samples obtained at temperatures above
250 8C contained less than 0.1% moisture. The volatile content of
deciduous wood decreases during torrefaction from 8183% to
7075%. Torrefaction decreases the atomic O/C ratio of the solid
product from approximately 0.70 to 0.520.60; this decrease
becomes larger at higher reaction temperature and prolonged
reaction times. There was a marked difference between willow,
beech and straw compared to larch. For the latter, a smaller
decrease was observed, with O/C ratios in the range of 0.600.65.
This is explained because in torrefaction of larch, mainly water is
formed, and relatively little carbon dioxide, carbon monoxide,
acetic acid and other carbon containing species, as in torrefaction
of beech and willow. The chemical composition of torrefied
wood is comparable to that of peat. Torrefied wood is an
intermediate between wood and wood charcoal, which requires
much higher temperatures to produce.
A very important property is the increase in energy density
of the torrefied wood compared to untreated wood. Due to the
loss of water vapor and carbon dioxide, for which the lower
heating value is zero, relatively more mass is lost than energy.
Fig. 7a and b shows the lower heating value of the solid product
on dry basis for willow and larch, respectively, beech and straw

Fig. 8. Lower heating value retained in torrefied wood on dry basis as a function
of temperature and residence time, for different biomass types. Bars indicate
error margin of 240 J/g in the ASTM method for bomb calorimetry.

for various process conditions. The errors bars represent a

standard error of 240 J/g for determination of heating values in
a bomb calorimeter. Clearly, the energy density increases with
temperature: for willow a 17% increase was observed at 270 8C
(from 17.7 to 20.7 MJ/kg) and for beech an increase of almost
20% at 280 8C (from 17.0 to 20.3 MJ/kg). It is also expected to
increase with reaction time, but in many cases such an increase
could not be observed due to the error margin in determination
of lower heating values. For softwood such as larch, a smaller
but statistically relevant increase in energy density of up to 7%
is observed. As has been argued earlier, the smaller weight loss
due to the lower reactivity of softwood explains this. It must be
remarked that the torrefied wood samples produced at
temperatures above 250 8C contained less than 0.1% moisture,
whereas the untreated wood did contain moisture, e.g. for beech
approximately 8.6%. If the lower heating value of untreated
wood is corrected for the enthalpy of evaporation of present
moisture, the actual increase in energy density brought about by
torrefaction is up to 30% for deciduous wood types.
The energy retained in the solid product is shown in Fig. 8.
Clearly, in torrefaction of the deciduous wood types and straw,
more energy is transferred with the volatiles in the form of
combustibles such as methanol, acetic acid and carbon
monoxide. The energy contained in volatiles increases with
temperature and reaction time. For torrefaction of larch, energy
conservation is almost complete below 250 8C as volatile
formation is very low.
4. Conclusions and recommendations

Fig. 7. Lower heating value of torrefied wood on dry basis as a function of

temperature and residence time, for (a) willow and larch and (b) beech and
straw. Bars indicate error margin of 240 J/g in the ASTM method for bomb
calorimetry.

It can be concluded that deciduous xylan-containing wood

(beech and willow) and straw are more reactive in the
torrefaction process than coniferous wood (larch). In the first
case, the mass conserved in the torrefied wood is lower: 73
83% (depending on residence time) versus 90% at a
temperature of 270 8C. Much more volatiles are formed, such
as acetic acid and methanol, which mainly originate from
acetoxy- and methoxy-groups present as side chains in xylose
units present in the hemicellulose fraction. These are not

40

M.J. Prins et al. / J. Anal. Appl. Pyrolysis 77 (2006) 3540

present in the hemicellulose fraction of coniferous wood. For

deciduous wood and straw, also more non-condensable gases
such as carbon dioxide and carbon monoxide are produced and
the extent of dehydration is higher. At a torrefaction
temperature of 270 8C and a reaction time of 15 min, the
increase in energy density of the solid product is much higher
for deciduous wood than for coniferous wood: 17% versus 7%.
Torrefaction of deciduous wood at these conditions still retains
90% of the energy in the solid material. Therefore, this process
is an attractive method to produce a renewable fuel with
favorable properties for gasification and/or (co-)combustion.
It is recommended to obtain more insight into the
torrefaction reactions by analyzing the chemical composition
of torrefied wood in more detail, to determine its main
components and degree of polymerization. Knowledge of the
hemicellulose content could enable more accurate determination of the required residence time at different temperatures.
The chemical properties of torrefied wood could also be
related to its mechanical properties, e.g. depending on the
extent of decomposition of the hemicellulose fraction in the
torrefaction process (and possibly depolymerization of the
cellulose fraction), the friability can be expected to increase.
Initial experiments for size reduction and fluidization of
torrefied wood have shown promising results [1].
Acknowledgements
Financial support of SDE (Dutch Foundation for Renewable
Energy) and Shell Global Solutions is gratefully acknowledged.

We would like to thank Barry van Dongen and Marijn Kappers

for their experimental work, which formed part of their M.Sc.
projects.
References
[1] P.C.A. Bergman, A.R. Boersma, J.H.A. Kiel, M.J. Prins, K.J. Ptasinski,
F.J.J.G. Janssen, in: Proceedings of the Second World Conference on
Biomass for Energy, Industry and Climate Protection, Rome, Italy, May
1014, 2004.
[2] J. Bourgois, J. Doat, Bioenergy 3 (1984) 153159.
[3] R. Pentanunt, A.N.M. Mizanur Rahman, S.C. Bhattacharya, Energy 15
(1990) 11751179.
[4] F.F. Felfli, C.A. Luengo, G. Bezzon, P. Beaton, in: H. Kopetz, T. Weber, W.
Palz, P. Chartier, G.L. Ferrero (Eds.), Proceedings of the 10th Eur. Conf.
Biomass for Energy & Industry, Wurzburg, Germany, June 811,
1998.
[5] M.N. Nimlos, E. Brooking, M.J. Looker, R.J. Evans, Prepr. Pap. Am.
Chem. Soc., Div. Fuel Chem. 48 (2003) 590591.
[6] M. Pach, R. Zanzi, E. Bjrnbom, Procedings of the Sixth Asia-Pacific
International Symposium on Combustion and Energy Utilization, Kuala
Lumpur, Malaysia, May 2022, 2002.
[7] S. Gaur, T.B. Reed, Thermal Data for Natural and Synthetic Materials,
Marcel Dekker, New York, 1998.
[8] R. Wagenfuhr, C. Scheiber, Wood Atlas, VEB Fachbuchverlag, Leipzig,
1974 (in German).
[9] M.V. Ramiah, J. Appl. Pol. Sci. 14 (1970) 13231337.
[10] I. Simkovic, G. Varhegyi, M.J. Antal, J. Appl. Pol. Sci. 36 (1988) 721
728.
[11] R.H. White, M.A. Dietenberger, in: K.H.J. Buschow, R.W. Cahn, M.C.
Flemings, B. Ilschner, E.J. Kramer, S. Mahajan (Eds.), The Encyclopedia
of Materials: Science and Technology, Elsevier, Amsterdam, 2001, pp.
97129716.