Is it possible to predict gas yields of any biomass after

rapid pyrolysis at high temperature from its

composition in cellulose, hemicellulose and lignin ?
Carole Couhert, Jean-Michel Commandre, Sylvain Salvador

To cite this version:

Carole Couhert, Jean-Michel Commandre, Sylvain Salvador. Is it possible to predict gas yields of any
biomass after rapid pyrolysis at high temperature from its composition in cellulose, hemicellulose and
lignin ?. Fuel, Elsevier, 2009, 88 (3), p.408-417. �10.1016/j.fuel.2008.09.019�. �hal-01846917�

HAL Id: hal-01846917

https://hal.archives-ouvertes.fr/hal-01846917
Submitted on 6 Nov 2018

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
Is it possible to predict gas yields of any biomass after rapid pyrolysis at
high temperature from its composition in cellulose, hemicellulose and lignin?
Carole Couhert *, Jean-Michel Commandre, Sylvain Salvador
Université de Toulouse, Ecole des Mines d’Albi – Carmaux, RAPSODEE UMR CNRS 2392, Campus Jarlard, Albi F 81013, France

a b s t r a c t

Ligno-cellulosic biomass from different sources presents very variable compositions. Consequently, there
is a wide variation in the nature and quantities of gaseous products obtained after thermal treatment of
biomasses.
The objective of this work is to establish a link between the composition of a biomass and its pyrolysis
gas yields and composition. Experimental flash pyrolysis of several biomasses at a temperature of 950 !C
and a gas residence time of about 2 s was carried out. An attempt was then made to predict gas yields of
any biomass according to its composition. We show that an additivity law does not allow the gas yields of
a biomass to be correlated with its fractions of cellulose, hemicellulose and lignin. Several potential
explanations are then offered and quantitatively demonstrated: it is shown that interactions occur
between compounds and that mineral matter influences the pyrolysis process.

Keywords:
Flash pyrolysis
Biomass
Components
Ash

1. Introduction ent constituents (cellulose, hemicellulose and lignin), which have

very different thermal behaviours. There is thus a need to develop
Biomass is today considered as a renewable fuel, whose use laboratory tools that are able to predict the thermal and kinetic
would help to reduce greenhouse gases. Over the past two decades, behaviour of biomasses from their composition.
worldwide interest has been growing in the thermochemical con- Some authors have tried to predict thermal and kinetic behav-
version of renewable resources through pyrolysis or gasification iour of biomasses from their composition by using an additivity
[1]. law. Rajeswara Rao and Sharma proposed a method in which the
There has been a great interest in pyrolysis, especially fast pyro- pyrolysis reaction is modelled by an nth-order rate equation, for
lysis of biomass to maximise gas production. Pyrolysis can be a components of the biomass and also for the main biomass materi-
process in itself, and is also the first step of any gasification pro- als. They found that experimental data on different biomass mate-
cess. Several studies in this field have looked at the effect of exper- rials can be satisfactorily predicted from the behaviour of their
imental parameters on product distribution. The three principal components and from their compositions assuming that there is
parameters studied are temperature, heating rate and the nature no interaction between them [8].
of the biomass. Yang et al. [9] pyrolysed cellulose, hemicellulose and lignin, and
When the temperature of pyrolysis increases, the yields of liq- synthesized biomass samples containing two or three of the bio-
uids and char decrease and the gas yields increase. Cracking of li- mass components. Samples were heated to 900 !C at 10 !C/min.
quid hydrocarbons favours the formation of gases, and cracking The pyrolysis of the synthesized biomass samples indicated negli-
of gaseous hydrocarbons favours the formation of H2 [2–6]. The gible interaction among the three components. A computational
higher the pyrolysis temperature, the higher the CO and H2 yields approach was made firstly to predict the weight loss of a synthe-
and the lower the CH4 and CO2 yields [3,4]. sized biomass from its composition in cellulose, hemicellulose
When the heating rate increases, at low temperature (up to and lignin, and secondly to predict the proportions of the three
500 !C), gas and char yields decrease whereas liquid yield in- components of a biomass. The results calculated for the weight loss
creases; at high temperature, char and liquids yields decrease of the synthesized biomass are quite consistent with the experi-
while gas yield increases [5,7]. mental results. However, results concerning the prediction of the
But the nature of the biomass can also influence the quantities composition of a biomass in cellulose, hemicellulose and lignin
and composition of products. Biomass is made up of many differ- were not very satisfactory.
Biagini et al. studied TG-FTIR analysis of ligno-cellulosic bio-
* Corresponding author. masses. A constant heating rate of 20 !C/min was used in all
E-mail address: couhert@enstimac.fr (C. Couhert). the experimental runs, from 105 to 1000 !C. The aim of their
work was to apply a weighted sum law (validated in the litera- – the higher the char yield;
ture) for the TG results, and to validate and extend this law for – the higher the total gas yield. H2 and CO2 yields increase and
the profiles of volatile species released and measured by FTIR. CH4 and CO yields decrease.
Assuming no interactions between the components, composition
of biomass in cellulose, hemicellulose and lignin was deduced The objective of this paper is to establish if it is possible, from
from the experiments. Once the chemical composition of the an additivity law, to correlate the gas yields after flash pyrolysis at
materials was obtained, a summative law was applied to the re- high temperature (950 !C) of any biomass from its composition in
lease of each gaseous species. The profiles of volatile species re- cellulose, hemicellulose and lignin.
leased were obtained with good agreement between the The approach consisted in first pyrolysing real biomasses and
experimental results and the calculated values. However, experi- searching for the gas yields of ‘‘theoretical components” (cellu-
mental results are not in good agreement with the calculated lose, hemicellulose and lignin) able to predict the gas yields of
values for paper sludge. This discrepancy can be ascribed to these biomasses after flash pyrolysis at high temperature. The
the predominant ash content of the material, which is expected first conclusion was that such a simple approach was not success-
to produce significant effects on the devolatilization reactions ful. The influence of interactions between the components and
[10]. the effect of mineral matter on pyrolysis gaseous yields were then
Caballero et al. studied the flash pyrolysis of almond shells and demonstrated and shown to be a possible cause of this
of their components in a Pyroprobe 1000 between 700 and 900 !C. conclusion.
They found that the almond shell yields obtained with the Pyrop-
robe 1000 are close to those calculated from the yields from lignin,
hemicellulose and cellulose for CO and CO2, but not always for light 2. Experiments
hydrocarbons. They studied the decomposition kinetics in a ther-
2.1. Experimental device
mobalance too. The thermal global decomposition of almond shells
cannot be reproduced by adding up the kinetics of decomposition
The entrained flow reactor (EFR) is described in Fig. 1. It is a
of their fractions. The authors suggest that there are interactions
75 mm i.d. and 2 m long tube into which a 1 m long isothermal
between the components [11].
reaction zone is fed with an electrically preheated laminar gas
Miller and Bellan worked on the pyrolysis of biomass and of cel-
flow. The powdered solid is injected through a water-cooled
lulose, hemicellulose and lignin. They developed a kinetic model to
feeding probe and dispersed over the cross-section of the reactor
predict the pyrolysis of arbitrary biomass feedstocks – both in low
using a dispersion dome. It is pyrolysed for a controlled resi-
temperature and in high temperature conditions – via a superposi-
dence time before being sampled by a water-cooled probe.
tion of cellulose, hemicellulose and lignin kinetics. This model,
The solid injection consists of a low velocity conveyor belt that
which is based on the models of Bradbury et al. [12] and Di Blasi
ensures an accurate solid mass flow rate. The belt is fed from a V-
and Russo [13], is intended for typical feedstock specimens and
shaped rail along which a precisely weighted quantity of solid is
atmospheric pyrolysis pressure. Further comparisons with the
regularly spread using a calibrated wedge. The gas and solid phases
experiments that are not used to fit the kinetic parameters show
can be collected at different heights (z) using a water-cooled sam-
a good agreement with the previous experiments of bagasse,
pling probe, allowing precise control of the residence time. After
cellulose, cherry wood, oak and pine for a large variety of both
separating gas from particles, the gases are forwarded to the ana-
TGA and isothermal pyrolysis conditions. The kinetic schemes have
lysers, via a heated line
even been incorporated into the previous porous particle model
of Miller and Bellan in order to model macro-particle pyrolysis.
– a Fourier transform infrared (FTIR) analyser to quantify con-
However, this macro-particle model can be improved, and the
tinuously CO, CO2, CH4, C2H2, C2H4, C2H6, C3H8, NO, NO2, N2O,
authors quoted several possible explanations among which the
NH3, HCN and SO2;
mineral matter in the wood is one [14].
– a non-dispersive infrared (NDIR) analyser for CO, CO2, SO2,
Finally, conclusions are different depending on the authors and
NO and NOx concentrations coupled with a paramagnetic
on the experimental conditions. Some authors affirm that it is pos-
analyser for O2 concentration;
sible to use linear correlation to predict gas yields of a biomass
– a thermal conductivity detector (TCD) for H2 concentration;
from its composition, but others affirm that this is not possible
– a methane and total hydrocarbons analyser, using two flame
and refer to possible interactions between the components and
ionisation detectors (FID);
the likely effect of mineral matter.
Few studies have been done on the interactions between
During the experiments, CO, CO2, H2, CH4, C2H2 and C2H4 were
components, but we can find several works on the influence of
detected.
mineral matter. Usually, two techniques are used to characterize
Gas sampling was operated at residence times up to 2.5 s.
the effect of mineral matter on pyrolysis reactions: biomass
The atmosphere gas (nitrogen) and the reactor walls were
washing and biomass impregnation. Washing is generally done
heated to the controlled temperature of 950 !C. The atmosphere
with water or with acid solutions. Depending on the technique,
gas plus the solid transport gas flow rate was 14 L/min (12 L/min
minerals are removed from the biomass to a greater or lesser de-
for atmosphere gas and 2 L/min for solid transport gas) at STP;
gree. Extraction of minerals from biomass is always selective
the solid mass flow rate was 1 g/min.
[15]. Moreover, washing biomass leads to a change in its compo-
sition in cellulose, hemicellulose, lignin and extractives. In fact,
washing removes not only mineral matter but other components 2.2. Properties of the solids used
like extractives and a part of the hemicellulose too [17].
In the literature, results obtained after thermal treatment of Because commercial hemicellulose cannot easily be purchased,
washed biomass and impregnated biomass are in agreement xylan has been widely used as a representative of the hemicellu-
[15–20]. The higher the concentration of minerals in biomass lose component in pyrolysis processes.
For all products, the fraction between 50 and 125 lm was se-
– the higher the degradation rate; lected by pneumatic sieving prior to any analysis and
– the lower the liquid yield; experiment.
 W M
2
M
C.
A
N2
3
0

5
1
6

C. N2
A

11

1650
10

8
z (mm)
15
PARAMAGNETIC
NDIR
FTIR W 9
14 M
FID
12
TCD M N2
13

1- Conveyor belt / vibrating corridor / ejector 9- Cyclone collector M- Mass flow meters and controllers
2- Pneumatic transport 10- Exhaust fan N2- Nitrogen
3- Electrical preheater 11- Isokinetic water cooled C.A- Compressed Air
4- Flow meters sampling probe W- Water (probes cooling)
5- Water cooled feeding probe 12- Particle collector (filter)
6- Dispersion dome 13- Sampling pump
7- Three zones electrical furnace 14- Gas conditioner
8- 75mm i.d quartz tube reactor 15- Gas analysers

Fig. 1. Schematic view of the entrained flow reactor of Ecole des Mines d’Albi.

The components used in this work were – Three xylans supplied by Sigma–Aldrich
! xylan from birchwood
– Two microcrystalline celluloses supplied by Sigma–Aldrich ! xylan from beechwood
! ‘‘microcrystalline cellulose” ! xylan from oat spelts
! ‘‘Sigmacell microcrystalline cellulose” – Three lignins
 ! alkali lignin supplied by Sigma–Aldrich 2.3. Preparation of the mixes
! lignin supplied by Meadwestvaco
! lignin supplied by Borregaard As illustrated in Fig. 3, three-component mixes were performed
and two ways
The extraction protocol that was used for each component is
not known. (i) Simple mix: The products were mixed in equal mass propor-
The biomasses used in this work were chosen because they tion with a spatula in a container. In this case, the particles,
have very different contents of cellulose, hemicellulose and lignin once injected in the EFR, may be in contact but will be essen-
tially dispersed in the gas phase. In these conditions, interac-
– Beech wood I and beech wood II (supplied by Lignex) tions in gas phase are favoured.
– ‘‘Spruce and fir” mix (supplied by Lignex) (ii) Intimate mix: The components were mixed and then co-
– rice husk (from Camargue, south-east of France) ground to thin elements (around 10 lm) using a laboratory
– grass (from Albi, France) ball mill. They were then agglomerated with a press, and then
– wood bark (from Les Landes, France) dispersed with a pestle and mortar to obtain particles around
100 lm. In this case, the thin elements of the components are
As it is difficult to determine accurately the composition in cel- in contact inside a given particle. Interactions between the
lulose, hemicellulose and lignin of a biomass because of the multi- components can now also occur inside the particles.
tude of current extraction processes, average compositions found
in the literature were adopted [21,22]. Grass contains large quan-
tities of cellulose and hemicellulose and a very small amount of lig- 2.4. Preparation of washed wood and impregnated wood
nin. On contrary, wood bark contains a large quantity of lignin and
very little cellulose. Wood (spruce and fir) contains the smallest 2.4.1. Washed wood
quantity of hemicellulose. Fifteen grams of beech wood II (50–125 lm) was placed in
Proximate analysis gave the amounts of ash, volatile matter 400 ml of distilled water for 24 h to remove part of the mineral
(VM) and fixed carbon (complement to 100%) as shown in Table matter from the wood particles. After vacuum filtration, the
1. The amounts of carbon (C), hydrogen (H), nitrogen (N) and sul- ‘‘washed wood” was placed in a drying oven at 105 !C until evap-
phur (S) were obtained by ultimate analysis. The amount of oxygen oration of the water.
(O) was obtained as the complement to 100% of C, H, N, S and ash.
Celluloses contain a large quantity of O and H in comparison 2.4.2. Impregnated wood
with lignins. More unexpected are the differences observed inside Firstly, straw ash was produced by burning straw in a muffle
a given family of components. For example, alkali lignin contains furnace at 815 !C for 2 h. Twenty-five grams of this ash was placed
37.1% of O, while lignin from Meadwestvaco contains 26.1% of O. and shaken in 400 ml of distilled water for 24 h to concentrate the
Alkali lignin contains 56.6% of C and lignin from Meadwestvaco water in the minerals. Secondly, 15 g of beech wood II was placed
contains 66.7% of C. As far as the biomasses are concerned, wood in the mineral concentrated water (after filtration) for 24 h. Finally,
bark contains a large quantity of C but low quantities of H and O after filtration, this ‘‘impregnated wood” was dried in an oven at
in comparison with the other biomasses. 105 !C until evaporation of the water.
Proximate analysis also indicates large differences. Celluloses Straw ash was chosen because straw contains a large quantity
contain ashes in undetectable quantities and very little fixed car- of minerals, especially K which is known to have a catalytic effect
bon; they form a large quantity of volatile matter during pyrolysis. during thermal treatment of biomass [18].
Lignins contain large quantities of fixed carbon and produce little Washing and impregnation of wood are known to affect the
amounts of volatile matter. composition of wood by removing some extractives and hemicellu-
Woody biomasses (beechwood and mix ‘‘spruce and fir”) con- lose. Consequently, ‘‘washed wood” and ‘‘impregnated wood” were
tain a small amount of ash in comparison with the other bio- prepared with the same mass of wood and of water. The aim was
masses. They can form a large quantity of volatile matter during thus to affect the woods’ compositions in cellulose, hemicellulose,
pyrolysis. Rice husk contains a large quantity of ash. lignin and extractives in the same manner. Mass loss caused by
Mass fractions of cellulose, hemicellulose and lignin in each bio- washing – calculated from weighing the dry wood before and after
mass are shown in Fig. 2. washing – represents 8.2% of the initial dry wood mass. Among this

Table 1
Ultimate analysis and proximate analysis of components and of biomasses

Ultimate analysis (mass% daf) Proximate analysis (dry mass%)

C H N S O Ash Volatile matters Fixed carbon
Components Microcrystalline cellulose 44.4 5.8 0.3 0.1 49.3 0 94.8 5.2
Sigmacell microcrystalline cellulose 44.5 5.6 0.3 0.1 49.5 0 94.1 5.9
Xylan from birchwood 46.7 5.7 0.2 0 47.4 4.1 73.3 22.6
Xylan from beechwood 44.3 5.4 0.3 0 49.9 4.8 75.3 19.9
Xylan from oat spelts 44.8 5.6 0.04 0 49.5 10.1 71.1 18.8
Alkali lignin 56.6 4.2 0.5 1.6 37.1 10 49.9 40.1
Lignin from Borregaard 65 4.9 0.6 0.8 28.7 13.5 45 41.5
Lignin from Meadwestvaco 66.7 5.6 1.1 0.5 26.1 2.4 58.9 38.7
Biomass Beechwood I 46.1 5 0.03 0 48.9 0.5 84.3 15.2
Beechwood II 50.3 4.9 0 0 44.8 0.2 83.2 16.6
Mix spruce and fir 49.1 5.5 0 0 45.4 0.4 82.9 16.7
Rice husk 50.7 6.4 0.9 0.06 42 15.2 61.8 23
Wood bark 55.3 4.6 0.6 0 39.6 4.9 65.3 29.8
Grass 45.5 6 3.2 0 45.4 8.5 74.5 17
 60 analysed. Gas mass yields of biomasses were calculated (as at% of
Cellulose
the initial dry-ash-free (daf) biomass) and are shown in Fig. 4. Each
Hemicellulose
Lignin experiment was repeated several times, typically 3–5 times;
50
uncertainties were calculated as follows, and are indicated in the
figure
40
Pn
mass % (daf)

i¼1 jX m # Xij
unc ¼
30 n

with Xm, mean gas yield; Xi, gas yield for experiment i; and n, num-
20 ber of experiments.
Note that the yield of CO and the total gas yield were divided by
10 10 in Fig. 4.
Gas yields are very different depending on the nature of the bio-
mass. Results show that gas yields of the two woody biomasses
0 (beech wood I and ‘‘spruce and fir” mix) are very similar. Moreover,
"spruce and fir" beechwood rice husk bark grass
they form more CO, CH4 and H2 than the other biomasses. In fact,
Fig. 2. Mass fractions of cellulose, hemicellulose and lignin in each biomass. CO yield represents more than half of the total quantity of analysed
gases. Consequently, a change in CO yield has a significant effect on
8.2%, only 0.02% represents mineral loss (calculated from proxi- the total quantities of analysed gases. This is why the total quanti-
mate analysis). Finally, ‘‘washed wood” contained 0.2% of minerals ties of analysed gases are greater for the two woody biomasses
(like the initial beech wood II) and ‘‘impregnated wood” contained than for all others biomasses.
1.2% of minerals. Grass forms less H2 and CO than other biomasses; it forms a
large quantity of CO2.
3. Results and discussion Note that after a time period of 0.5–1 s, the gas yields of the bio-
masses do not appear to change significantly with time. Pyrolysis
3.1. Pyrolysis of biomasses and correlation between composition of the can be considered as finished after these short residence times.
biomass and gas yields Moreover, it seems that gas phase cracking reactions stop forming
permanent gas after these short residence times. It is not possible
3.1.1. Pyrolysis of biomasses from the present results to determine accurately from what time
The beech wood I, ‘‘spruce and fir” mix, rice husk, grass and the pyrolysis and tar cracking are over; this is not the scope of
wood bark were pyrolysed in the EFR, and the resulting gases were the work.

Simple mix
Intimate mix

grinding compacting dispersion

50 -125 µm 10-20 µm
50 - 125 µm

Fig. 3. Simple and intimate mixes.

mix "spruce and fir" beechwood I rice husk bark grass

14

12

10
mgas / mo (% daf)

0
C2H4 C2H2 CH4 CO2 H2 CO/10 total / 10

Fig. 4. Gas yields of biomasses.

3.1.2. Search for correlation Y beech wood ¼ abeech wood Y cellulose þ bbeech wood Y hemicellulose
In this study, we aim to research correlations between compo- þ cbeech wood Y lignin
sitions of biomasses and gas yields obtained during their pyrolysis.
Y spruce and fir ¼ aspruce and fir Y cellulose þ bspruce and fir Y hemicellulose
Several approaches can be considered. Biomass can be described as
a complex mix of polymers: þ cspruce and fir Y lignin
Y rice husk ¼ arice husk Y cellulose þ brice husk Y hemicellulose þ crice husk Y lignin
– composed of carbon, hydrogen and oxygen; Y bark ¼ abark Y cellulose þ bbark Y hemicellulose þ cbark Y lignin
– composed of functional groups;
Y grass ¼ agrass Y cellulose þ bgrass Y hemicellulose þ cgrass Y lignin
– composed of cellulose, hemicellulose and lignin.
with Y, gas yield (for a given gas species); ai, mass fraction of cellu-
In this work, we will focus on the search for correlation between lose in biomasses ‘‘i”; bi, mass fraction of hemicellulose in bio-
the gas yields of any biomass and its composition in cellulose, masses ‘‘i”; ci, mass fraction of lignin in biomasses ‘‘i”.
hemicellulose and lignin. We have represented the experimental results of gas yields in
If cellulose, hemicellulose and lignin in a biomass undergo pyro- Fig. 5 using a ternary diagram. This was done for the three main
lysis without interaction between them, then gas yields should fol- gases: CO, CH4 and H2. The three vertexes are the pure components
low the additivity law of cellulose, hemicellulose and lignin. The content of a biomass (+
symbols) in one component corresponds with the value read on
the corresponding axis after perpendicular projection. In ordinate
the gas yields are plotted.
If an additivity law can be applied, the gas yields for the five bio-
masses should lie in a plane. The intersection of this plane with the
vertical lines above the vertexes C, H and L would define the gas
yields of ‘‘theoretical” pure components. We have searched this
plane (‘‘least squares plane”) minimising standard deviations with

Fig. 6. Effect of uncertainties concerning yields in CO (a), CH4 (b) and H2 (c) of
Fig. 5. Yields in CO (a), CH4 (b) and H2 (c) of biomasses and theoretical components. biomasses on yields of theoretical components.
all the biomass gas yields. Discrepancies between the plane and 3.2. Failure of the component additivity rule to predict gas yields of
the points corresponding to the biomass gas yields are materialised biomass
by continuous vertical lines. Fig. 5 shows that the intersection of
the plane with the vertical line above the pure hemicellulose oc- It has been shown that it is not possible to predict the gas
curs at a negative value for CO, CH4 and H2 yields, which is not a yields of any biomass from its composition in cellulose, hemicellu-
physically possible solution. At this stage, it would seem that we lose and lignin. A number of potential explanations are quoted
cannot find theoretical components able to represent the gas yields below:
of the five biomasses.
However, it is necessary to take into account the uncertainties – lignins in different biomasses may not be similar and may
on gas yield values, because they can modify the position of the form different quantities of gases. The same explanation
plane and thereby modify the gas yields of theoretical components. can be applied for hemicellulose and cellulose;
Therefore, two ‘‘least squares planes” are drawn in Fig. 6 – components interact during pyrolysis;
– ash influences pyrolysis reactions;
– the least squares plane corresponding to the higher values of – extractives contribute significantly to gas yields.
the biomass gas yields. These values are obtained by adding
uncertainties to average values of gas yields (represented in Hereafter, we will investigate the first three points in order to
Fig. 5). evaluate how far they can explain the discrepancies observed
– the least squares plane corresponding to the smallest values above.
of the biomass gas yields. These values are obtained by sub-
tracting uncertainties from average values of gas yields. 3.2.1. Gas yields of different lignins, hemicelluloses and celluloses
Fig. 7 shows the values of gas yields of the different lignins,
This shows that the intersections of the two planes with the hemicelluloses and celluloses that were selected for this work,
vertical line above the pure hemicellulose still occur at negative and characterized previously in detail. Again, the experiments
values for the gas yield of hemicellulose. Consequently, even when were repeated 3–5 times, which makes it possible to indicate error
uncertainties are taken into account, it is not possible to correlate bars on the figure.
gas yields of any biomass with its composition in cellulose, hemi- It can be seen that there are differences between the average
cellulose and lignin. This enables us to conclude that the idea that gas yields of the three different families of components. But these
biomasses composed of the same cellulose, hemicellulose and lig- differences are often less significant than the differences observed
nin all pyrolyse independently, does not correspond to reality. It is inside a given family of components. For example, concerning CH4,
also interesting to note that the least square planes always present the difference between yields of alkali lignin and lignin from Mead-
a significant slope when varying only the hemicellulose content. westvaco is comparable to the difference between the average
This indicates that the amount of hemicellulose has a significant yield for lignins and the average yield for xylans. Such differences
impact on all gas yields, while the impact of the ratio between lig- can be observed for the other gases yields too. However, the two
nin and cellulose seems less important. celluloses produce similar quantities of gases. This is not surprising
A remark can be made concerning the possibility of using this since the structure of cellulose is relatively well known and does
ternary diagram representation for practical applications. We can not depend on its origin. It is not the case for lignins and
see in Fig. 5 that the least squares plane is relatively close to the hemicelluloses.
five points corresponding to the biomass gas yields in a limited As was pointed out earlier, the differences between compo-
area of biomass composition (located around the five biomasses). nents’ gas yields could perhaps be explained by the origin of the
This plane can be used to predict gas yields of a biomass from its components, but may also be the result of the extraction process.
composition with a maximal discrepancy of 25%, whereas gas Indeed, the structure of a component of the biomass can be modi-
yields of biomasses vary by 60%. This result is interesting and fied by the extraction process, which may affect the gas yields dur-
needs to be confirmed with additional experiments of pyrolysis ing pyrolysis.
with alternative biomasses (with different compositions in cellu- The conclusion on this point is that celluloses extracted from
lose, hemicellulose and lignin). different biomasses lead to similar gas yields during pyrolysis. This

14

12

10
mgaz / mo (% daf)

0
H2 C2H4 C2H2 CH4 CO/10 CO2 total/10
white: from left to right: microcrystalline cellulose and Sigmacell microcrystalline cellulose
grey : from left to right: xylan from beechwood, xylan from birchwood and xylan from oat spelts
black: from left to right: alkali lignin, lignin from Meadwestvaco and lignin from Borregaard.

Fig. 7. Gas yields from all the components at 950 !C.

is not the case for hemicellulose and for lignin, which can present the mass-weighted average yields of the three components as
very different yields in a same family. However, it cannot be estab- follows:
lished from this work whether differences are due to differences of
original structures or if the extraction process has affected the Y mix ¼ 0:4Y cel þ 0:39Y xyl þ 0:21Y lig ð1Þ
structure of hemicellulose and lignin.
where Ymix is the gas yield of the mix, Ycel is the gas yield of the cel-
3.2.2. Interactions between components lulose, Yxyl is the gas yield of the xylan, Ylig is the gas yield of the
In this study, we have carried out experiments to establish if lignin.
interactions between components – inside or outside particles – The intimacy of the mixing might play a role on the interactions
occur or not. For these experiments, microcrystalline cellulose, lig- between the components. In order to take this into consideration,
nin from Borregaard and xylan from birch wood were first pyroly- the differences between gas yields calculated with Eq. (1) and
sed separately. They were then mixed in three-component mixes. gas yields analysed during pyrolysis of simple and intimate mix-
The mixes were prepared in two ways: the simple mix and the inti- ings (described previously) have been plotted. Results – not pre-
mate mix, as detailed before. The proportion of each component in sented here – have shown that it is difficult to conclude because
the mix is similar to those found in natural birch wood: 40% cellu- of uncertainties (mean deviations obtained by repeating experi-
lose, 39% hemicellulose and 21% lignin. If no interaction occurs, the ments three to five times) that are sometimes larger than the ob-
gas yield (for a given gas species) of a mixture should be equal to served difference. However, for certain gas species, clear
differences are observed between experimental yields and calcu-
lated yields. As an example, Fig. 8 presents the CO2 yields calcu-
14 lated with Eq. (1) and the CO2 yields measured during pyrolysis
of simple and intimate mixes. Results show that there are differ-
12
ences between CO2 yield calculated with Eq. (1) and experimental
CO2 yields. Moreover, there are significant differences in the CO2
CO2 yield (daf mass %)

10
yields of simple and intimate mixes: the more intimate the mix,
8 the higher the CO2 yield.
A more detailed analysis of the situation is proposed before try-
6 ing to interpret these results. During the pyrolysis of simple mixes,
the three components devolatilize separately. Interactions are
4
likely to occur outside the particles. Some potential interactions
2 are described in Fig. 9 and quoted below:

0 1. Homogeneous gas phase reactions: gases and condensable

calculated simple mix intimate mix
vapours formed by one component may react with gases and
Fig. 8. CO2 yield calculated by Eq. (1) (considering no interaction) and CO2 yields condensable vapours formed by another component. Reform-
measured during pyrolysis of simple and intimate mixes. ing, oxidation or polymerisation reactions are presumed.

Pyrolysis of each
Pyrolysis of mixes
component

cellulose * * * 1
* * *
* * *
4 *
* * 2
*
*
* * * *
xylan 3 * Simple mix
*
* * *
* *

*
* *
* *
* *
Zoom * *
lignin * *
* * * *
* * *
* * * *
*
*
* * Intimate mix
Mineral matter
Condensable vapours
* * Gases

Fig. 9. Simplified schema of the pyrolysis of components taken separately and of the pyrolysis of component mixes.
 washed wood mix "washed wood + 1 % ashes" mix "washed wood + 10 % ashes" impregnated wood

6
mgas / mo (% daf)
5

0
C2H4 C2H2 CH4 CO/10 CO2 H2 total/10

Fig. 10. Gas yields of ‘‘washed wood”, ‘‘impregnated wood” and simple mixes ‘‘washed wood + ashes”. Wood used is beech wood II.

2. Gas phase reactions: catalysed by char or/and ashes (or mineral the gas phase in quantities of 6.10#4 g/g gas and 6.10#3 g/g gas only
matter), like for example catalysed cracking and polymerisation slightly affects the observed gas yields after pyrolysis.
reactions. Char could modify CO–CO2 equilibrium too. On the other hand, ‘‘impregnated wood” forms less CO, much
3. Heterogeneous reactions: gases formed by one component may more CO2 and less CH4 than ‘‘washed wood”. The H2 yield is not af-
react with char formed by another component. For example, fected by ash enrichment. We can conclude that minerals signifi-
char oxidation reactions by O2, H2O or CO2 may occur. These cantly influence the pyrolysis reactions occurring inside the particle.
reactions could be catalysed by ashes or mineral matter.
4. Influence of pyrolysis atmosphere: gases formed by one compo- 4. Conclusions and perspectives
nent may influence the devolatilization process of another com-
ponent by modifying the atmosphere under which it pyrolyses. The main result from this study is that it is not possible to pre-
dict – from an additivity law – pyrolysis gas yields of any biomass
During the pyrolysis of intimate mixes, reactions can occur out- from its composition in cellulose, hemicellulose and lignin. This
side the particles in the same way as during the pyrolysis of simple work shows that two phenomena may explain this:
mixes, but additional interactions may occur inside the particles.
As one component devolatilizes inside the particle, it is submitted – components interact between themselves during pyrolysis. It
to an atmosphere with very high concentrations in gas and con- was shown in this work that interactions occur outside the
densable vapours; the gases formed are in close contact with the particle – i.e. in the gas phase – and probably also inside
solids of other components. the particles since the intimacy of the mix has an impact
In the present results, there are interactions during the pyroly- on the CO2 yields and
sis of simple mix. Consequently, it is likely that some interactions – minerals influence pyrolysis reactions occurring inside the
occur outside the particles because particles are not in contact dur- particle. They favour the formation of CO2 and decrease the
ing this pyrolysis. There are also probably interactions inside the formation of CO and CH4; an addition up to 6.10#3 g/g gas
particles because CO2 yield of intimate mix is higher than CO2 yield of ash in gas phase does not influence the reactions occurring
of simple mix. Similar results were obtained in a previous work outside the particle.
[23] with two component mixes instead of three-component mixes
in this work. The approach consisting of extracting pure components from
biomass in order to characterize their behaviour during pyrolysis
3.2.3. Influence of mineral matter is not recommended. Indeed, it remains impossible to establish
In this part, the aim is to evaluate the potential influence of how far the extraction process affects the structure of components.
mineral matter on gas yields in the conditions of high temperature This research work also yields practical information. It seems
rapid pyrolysis. An attempt is made to differentiate the effect of possible – in a restricted domain of composition – to predict the
minerals located inside a particle and the effect of minerals when gas yields of any biomass from its composition in cellulose, hemi-
dispersed in the gas phase. cellulose and lignin, with a maximal discrepancy of 25%, whereas
A ‘‘washed wood” containing 0.2% of minerals and an ‘‘impreg- gas yields of biomasses vary by 60%. This result needs to be con-
nated wood” containing 1.2% of minerals were prepared as detailed firmed with additional experiments of pyrolysis of alternative
previously to study the effect of minerals inside the particle. In par- biomasses.
allel, simple mixes ‘‘washed wood + 1% straw ash” and ‘‘washed
wood + 10% straw ash” were prepared by manual mixing in order Acknowledgements
to investigate the second effect.
Fig. 10 shows gas yields of ‘‘washed wood”, ‘‘impregnated We thank the GEM (Groupe des Ecoles des Mines) for support-
wood” and the two simple mixes. ing the ‘‘H2-fuel cell” project and CNRS for supporting the ‘‘Biomas-
‘‘Washed wood”, the simple mix ‘‘washed wood + 1% straw ash” ters” Project.
and the simple mix ‘‘washed wood + 10% straw ash” form approx-
References
imately the same quantities of gases even if a slight decrease in the
yields of CH4 and CO is observed when the amount of ash is in- [1] Ferdous D, Dalai AK, et al. Production of H2 and medium Btu gas via pyrolysis
creased in the phase gas. It can be concluded that adding ash in of lignins in a fixed-bed reactor. Fuel Processing Technol 2001;70:9–26.
 [2] Demirbas A. Gaseous products from biomass by pyrolysis and gasification: [13] Di Blasi C, Russo G. Modeling of transport phenomena and kinetics of biomass
effects of catalyst on hydrogen yield. Energy Convers Manage 2002;43: pyrolysis. In: Bridgwater AV, editor. Advances in thermochemical biomass
897–909. conversion, vol. 2. New York: Blackie Academic and Professional; 1994. p. 906–21.
[3] Corella J, Monzon A, et al. Ultra-fast biomass pyrolysis in a high-temperature [14] Miller RS, Bellan J. A generalized biomass pyrolysis model based on
(2200 !C), fluid-wall reactor. J Solar Energy Eng 1988;110:10–3. superimposed cellulose, hemicellulose and lignin kinetics. Combust Sci
[4] Zanzi R, Sjöström K, et al. Rapid pyrolysis of agricultural residues at high Technol 1997;126:97–137.
temperature. Biomass Bioenergy 2002;23:357–66. [15] Di Blasi C, Branca C, et al. Degradation characteristics of straw and washed
[5] Li S, Xu S, et al. Fast pyrolysis of biomass in free-fall reactor for hydrogen-rich straw. Thermochim Acta 2000;364:133–42.
gas. Fuel Processing Technol 2004;85:1201–11. [16] Das P, Ganesh A, et al. Influence of pretreatment for deashing of sugarcane
[6] Bitowft B, Andersson LA, et al. Fast pyrolysis of sawdust in an entrained flow bagasse on pyrolysis products. Biomass Bioenergy 2004;27:445–57.
reactor. Fuel 1989;68:561–6. [17] Yang H, Yan R, et al. Influence of mineral matter on pyrolysis of palm oil
[7] Zanzi R, Sjöström K, et al. Rapid high-temperature pyrolysis of biomass in a wastes. Combustion Flame 2006;146:605–11.
free-fall reactor. Fuel 1996;75(5):545–50. [18] Nik-Azar M, Hajaligol MR, et al. Mineral matter effects in rapid pyrolysis of
[8] Rajeswara Rao T, Sharma A. Pyrolysis rates of biomass materials. Energy beech wood. Fuel Processing Technol 1997;51:7–17.
1998;23(11):973–8. [19] Bru K, Blin J, et al. Pyrolysis of metal impregnated biomass: an innovative
[9] Yang H, Yan R, et al. In-depth investigation of biomass pyrolysis based on three catalytic way to produce gas fuel. J Anal Appl Pyrolysis 2007;78:291–300.
major components: hemicellulose, cellulose and lignin. Energy Fuels [20] Vamvuka D, Troulinos S, et al. The effect of mineral matter on the physical and
2006;20:388–93. chemical activation of low rank coal and biomass materials. Fuel 2006;85:
[10] Biagini E, Barontini F, et al. Devolatilization of biomass fuels and biomass 1763–71.
components studied by TG/FTIR technique. Ind Eng Chem Res 2006;45: [21] Energy Research Centre of the Netherlands. <http://www.ecn.nl/phyllis/single.
4486–93. html>; 2006.
[11] Caballero JA, Font R, et al. Comparative study of the pyrolysis of almond shells [22] Williams PT, Nugranad N. Comparison of products from the pyrolysis and
and their fractions, holocellulose and lignin Product yields and kinetics. catalytic pyrolysis of rice husks. Energy 2000;25:493–513.
Thermochim Acta 1996;276:57–77. [23] Couhert C, Commandré JM, Salvador Sylvain. Failure of the component
[12] Bradbury AG, Sakai Y, Shafizadeh F. A kinetic model for pyrolysis of cellulose. J additivity rule to predict gas yields of biomass in flash pyrolysis at 950 !C.
App Polym Sci 1979;23:3271–80. Biomass Bioenergy, in press. doi: 10.1016/j.biombioe.2008.07.003.