You are on page 1of 15

J. Anal. Appl.

Pyrolysis 80 (2007) 151–165


www.elsevier.com/locate/jaap

Kinetic analysis of the thermal decomposition of cellulose:


The main step of mass loss
Vadim Mamleev a,b,*, Serge Bourbigot b, Jacques Yvon c
a
Institute of Polymer Materials and Technology, Kazakh-American University, Satpaev Str., 18a, Almaty 050013, Kazakhstan
b
Laboratoire des Procédés d’Elaboration de Revêtements Fonctionnels (PERF), LSPES UMR-CNRS 8008, Ecole Nationale Supérieure de Chimie de Lille
(ENSCL), Université des Sciences et Technologies de Lille (USTL), BP 90108, 59652 Villeneuve d’Ascq Cedex, France
c
Laboratoire Environnement et Minéralurgie, Centre de Recherche François Fiessinger, 15 Avenue du Charmois, 54500 Vandœuvre-lès-Nancy, France
Received 20 May 2006; accepted 16 January 2007
Available online 31 January 2007

Abstract
Modulated thermogravimetry (MTG) has confirmed the conclusion of Shafizadeh and Bradbury [F. Shafizadeh, A.G.W. Bradbury, J. Appl.
Polym. Sci. 23 (1979) 1431–1442] that the thermal decomposition of cellulose over the wide range of mass loss is essentially the same in both air
and nitrogen. At the beginning of the decomposition oxygen interacts only with surface cellulose, resulting in only ffi3% of mass loss. The two-
phase model of cellulose is proposed to explain all observable phenomena related to both the pyrolysis (inert atmosphere) and the oxidative
decomposition (air or oxygen). According to the model, the decomposition occurs through a migration of chain ends from the phase of polymer
cellulose into the phase of products (char, gases and high-boiling tar). The interface between two phases is the region of location of chain ends;
therefore unzipping of cellulose is preferable in comparison to the random scission of macromolecules. That is why the yield of levoglucosan can
reach 60%. It evolves as a result of the unzipping by the transglycosylation with activation energy of about 200 kJ/mol. Oxygen interacts with
products of the decomposition but does not penetrate into the matrix of polymer cellulose. The oxidation is the fast reaction subsequent to the
depolymerization; therefore, the main step of the degradation is independent of the oxidative reactions.
# 2007 Elsevier B.V. All rights reserved.

Keywords: Temperature-modulated thermogravimetry; Cellulose; Thermal decomposition; Kinetics

1. Introduction vegetable biomass yearly [4]. The resources of the energy


stored in it exceed the consumable energy expended every year in
Cellulose is the most abundant organic polymer (we may also the form of petroleum together with the natural gas and coal [4].
talk about biopolymer) on Earth and among all renewable Half of this biomass (ffi1011 tonnes [5]) is cellulose providing the
organic substances on the planet it has the maximum renewable main constituent of the renewable energy potentially accessible
mass, playing an exceptional role in forming vegetative from plants.
structures. Chemistry of cellulose is the special field of Moreover, cellulose is the substance of the extreme
knowledge [1,2]. Although the thermal degradation of cellulose importance for the textile industry. Earlier we studied the
is only a fragment of this knowledge, numerous publications are thermal decomposition of purified cotton [6–8] that is known to
dedicated to the investigation of its mechanisms. Biomass, in present the concentrated source of cellulose (about 94% even in
particular, wood can be used as a fuel and the source of renewable raw cotton).
energy. Therefore such investigations have the obvious applied We used temperature-modulated thermogravimetry in this
orientation [3]. Nature derives ffi2  1011 tonnes of dry research [7,8]. The advantages of the method in comparison
with classical thermogravimetry were discussed elsewhere [7–
10] and here we present only the final results proving the
benefits of using modulated thermogravimetry (MTG).
* Corresponding author. Permanent address: Institute of Polymer Materials
and Technology, Kazakh-American University, Satpaev Str., 18a, Almaty
Information by MTG allows a calculation of the activation
050013, Kazakhstan. energy, E, as a formal function of the degree of decomposition,
E-mail address: Vadim_Mamleev@yahoo.com (V. Mamleev). a, by the model-free method [8]. Plateaus in the dependence
0165-2370/$ – see front matter # 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jaap.2007.01.013
152 V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165

Nomenclature

a heating rate (8C/min)


Ai frequency factor (s1)
Ei activation energy (kJ/mol)
fi conversion function
Ki reaction rate constant (=Ai exp(Ei/RT)) (s1)
ð0Þ
mi initial mass of inactive cellulose being unable to
oxidizing at low temperatures
ð0Þ
mr and mr initial and current mass of reactive cellulose
being able to oxidizing at low temperatures
m0 initial mass of cellulosic sample
R gas constant (kJ mol1 K1)
t time (s)
T temperature (8C or K)
ðiÞ
Td peak temperature of ith reaction (8C or K)
T0 initial temperature (8C or K)
wi fraction of ith reaction in overall mass loss
wox fraction of cellulose able to oxidizing at low
temperatures
Yi yield of ith reaction (%)

Greek letters
a degree of conversion
ai partial degree of conversion
u parameter integration corresponding to time
t reduced time (s)

Subscripts
ac pathway of formation of ‘‘active cellulose’’
gas pathway of formation of light gases
Fig. 1. (a) The results of the approximation of activation energy and (b) of the
i ith reaction time derivative of a by using the three-step kinetic scheme with independent
ox oxidative decomposition reactions. Information represented was obtained from curve 1 in Fig. 2. As
tar pathway of tar formation established [8], step A is an artifact caused by the apparatus restrictions.

Superscript
i ith reaction where wB ¼ 0:055; wC ¼ 0:598; wD ¼ 0:347; AB = 8  105
s1, AC = 8.9  1014 s1, AD = 0.42  108 s1; EB = 90 kJ/
mol, EC = 195.9 kJ/mol, ED = 153.1 kJ/mol.
E(a) correspond to reliable values of activation energies for The derivative is approximated with some deviation
different steps of decomposition (Fig. 1a). After the approx- (Fig. 1b), although one can see that after integration (curve
imation of the derivative of mass loss the dependence E(a) can 1 in Fig. 2) the derivative leads to the theoretical kinetic curve
be described [8,10] theoretically (Fig. 1a). corresponding to the experimental one with the excellent
The data displayed in Fig. 1 correspond to the oxidative accuracy. Furthermore, the kinetic constants obtained allow the
decomposition of the cotton cellulose in synthetic air at the prediction of curve 2 (Fig. 2), which has been obtained at
heating rate of 2 8C/min. For the approximation of the another heating rate, 6 8C/min. We note that the latter curve was
derivative we used the three-step kinetic scheme with calculated on the basis of information hidden in curve 1. The
independent reactions [8]. The derivative da/dt and the accuracy of the prediction is comparable with the accuracy of
dependence E(a) were described [8] as follows: the approximation and is excellent as well.
    The predictive power of MTG and the objectivity of this
da EB EC method are connected with the fact that the determination of
¼ wB AB f B ðaB Þ exp þ wC AC f C ðaC Þ exp
dt RT RT E(a) does not require any model of decomposition [7,8]. This
 
ED function is merely experimental information that really reflects
þ wD AD f D ðaD Þ exp ; physicochemical features of the phenomenon and not para-
RT
  meters of its simulation by some theoretical equations.
wB EB daB =dt þ wC EC daC =dt þ wD ED daD =dt A cursory glance at Fig. 1a allows one to see that the

da=dt function E(a) has two pronounced plateaus (steps C and D).
V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165 153

Fig. 3. The Broido–Shafizadeh scheme for the main steps of cellulose pyr-
olysis.

been handled quantitatively in the laboratory of Shafizadeh and


co-workers [27,28], remains the classical model [13,19,21–23]
of cellulose pyrolysis. This mechanism named in the scientific
literature [13,23] as the Broido–Shafizadeh model is symbo-
lically depicted in Fig. 3.
Fig. 2. The kinetic curves of the decomposition of the cotton cellulose in air at According to this scheme, the first step, Kac, is the formation
the different heating rates. Curve 1 was used in determining the kinetic of the so-called ‘‘active cellulose’’. This step presumably is
constants obtained by handling the perturbations caused by the temperature connected only with a scission of glycosidic bonds, probably by
modulation [7,8]. The reliability of those constants is proven by the accurate
transglycosylation when cellulose does not lose its mass. Note
prediction of experimental curve 2 measured at another heating rate.
that in some publications this step is attributed to the formation
of the so-called ‘‘anhydrocellulose’’ [23], although ‘‘active
Step D related to oxidation of the carbonized residue [8] is less cellulose’’ is a material without mass loss, while ‘‘anhydro-
interesting for chemistry of cellulose in comparison to step C. cellulose’’ is a material with a lost mass due to the water
Indeed, step D can be expected for many carbonized polymers removal resulting in cross-linking of cellulose.
[8], while steps B and C are peculiarities intrinsic only for The initial idea proposed by Kilzer and Broido [24] included
cellulose. the step responsible only for the formation of the ‘‘anhy-
The constancy of the dependence of E(a) on step C over the drocellulose’’ (see, e.g., the comments of Capart et al. [19]).
wide ranges of a and T (0.2 < a < 0.62, 295 8C < T < 320 8C) The more detailed scheme, which combines the ideas of both
is a surprising fact. Indeed, even for pyrolysis (decomposition Broido [26] and Shafizadeh [28], should include both activation
in an inert atmosphere) one might expect many competing and dehydration (Fig. 4). Unfortunately, these processes are
reactions. All the more, their variety should rise for the supposed to occur only at rather low temperatures, and so their
oxidative decomposition in air (our experiment). Nevertheless, experimental discrimination and interpretation are compli-
we see the constancy of EC meaning that the decomposition on cated.
this step is controlled by one individual chemical reaction. This Note that we have purposely replaced the symbols Kc and Kv
effect surely needs a detailed explanation. The main purpose of [27] by the symbols Kgas and Ktar, respectively. Bradbury et al.
the present article is to find it by using stronger argumentation [27] implied the abbreviations ‘‘c’’ and ‘‘v’’ in their famous
in comparison to our previous article [8]. publication to denote ‘‘char’’ and ‘‘volatiles’’. However, such a
We could doubt of this effect but the cellulose behavior designation suffers from a few defects.
described in the article of Shafizadeh and Bradbury [11] has the First, a mass loss in the thermogravimetric experiments is
strong resemblance to our observations. The authors have connected with gasification and not with charring.
established the following features, literally. Oxidative reactions Second, it is hardly possible to consider ‘‘char’’ in Fig. 3 as
are responsible for acceleration in the rates of weight loss and being an individual chemical substance. Indeed, char can be
depolymerization of cellulose on pyrolysis in air at tempera- considered either as a partly cross-linked polymer resin
tures below 300 8C (obviously step B in Fig. 1a) but at (‘‘anhydrocellulose’’) or as a product resembling pure carbon.
temperatures above 300 8C (obviously step C in Fig. 1a) the rate Obviously, both definitions are far from exact physical
of pyrolysis is essentially the same in both air and nitrogen, chemistry. In this context Fig. 4 is more understandable.
indicating that thermal degradation is independent of the Third, according to the unconventional researches of Li and
oxidative reactions. co-workers [20–22], step Kgas (or Kc) has no direct relation to
Believing the conclusions of Shafizadeh, we might interpret
step C by comparing the rate-limiting steps within the
corresponding temperature range for the oxidative decomposi-
tion (our experiment) and for cellulose pyrolysis (inert
atmosphere) studied, in turn, in numerous publications [e.g.
12–23].
Having read the recent literature [12–23], one can make sure
that despite many achievements reached during the last four
decades, up to now the kinetic scheme, which has been
proposed by Broido and co-workers [24–26] and which has Fig. 4. The generalized Broido–Shafizadeh mechanism.
154 V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165

charring at all. Besides, these authors have established that CO2 The condition d2a/dt2 = 0 and Eq. (2) lead to the transcendental
is an exception with respect to other light gases. Its yield and equation with respect to the peak temperature Td with a preset a:
yields of other light gases change in the inverse manner; an  2  
increase of the CO2 yield corresponds to a decrease of yields of RTd E
a¼A exp (3)
other light gases. E RT d
Nevertheless, the assumptions underlying the scheme in Such an evaluation of the peak temperatures [7] should give
Fig. 3 are quite understandable. The cross-linking competes good accuracy not only for first-order reactions but also for
with depolymerization. If cellulose at the beginning of other types of kinetics corresponding to bell-shaped derivatives,
pyrolysis is cross-linked it tends to conversion toward light da/dt, being approximately symmetrical on the scale of time
gases and char, rather than toward heavy molecules of tar. (temperature).
Actually, such simple considerations exhaust the general The temperature integral (reduced time) is approximately
hypothesis underlying the Broido–Shafizadeh model and at first calculated [31] as:
sight these arguments seem quite convincing. Z t    2  
As for us, we might arbitrarily ascribe the char formation to E RT E
t¼ exp du  exp
any of the pathways, Ktar or Kgas. Only the fact of the domination 0 RTðuÞ aE RT
of one of two pathways is important for interpreting step C.
The aforesaid considerations predetermine the sequence of Thus, for low conversions (a  1, a  At) one can readily
the necessary actions to explain the observable facts and to obtain the useful expression for the experimental determination
elucidate the mechanism of cellulose degradation. of E for kinetics of any order:
First of all we have to evaluate the kinetic parameters of the    
a AR E
oxidation (Section 3.1) to explain why it influences the cellulose ln 2
¼ ln  (4)
T aE RT
degradation at the low temperatures (step B) but ceases to affect
the decomposition at higher temperatures (step C).
Using the Broido–Shafizadeh model for a rational orienta- 2.2. The two-step competition model
tion, we have to confront the published results (Section 3.2) of
the pyrolytic experiments to find a rate-limiting step and, thus, In the case of two competing reactions one should correct
to identify step C chemically (Section 4.1). This will give Eq. (3), since the reciprocal dependence of the reactions
information to pass through the following stages in under- influences the peak temperatures. We shall proceed from the
standing of cellulose decomposition. model:
Using a number of indirect arguments (Section 4.2) along    
1 da E1 E2
with the microphotographs of cellulose [29], we have to ð1  aÞ ¼ A1 exp þ A2 exp (5)
approach to its mechanistic model consistent with all dt RTðtÞ RTðtÞ
considered experimental facts (Section 4.3). Ultimately, we For generalization to any two competing reactions in nature we
have to explain (Section 4.4) why we do not see in Fig. 1a any used the subscripts ‘‘1’’ and ‘‘2’’. Referring to a particular
indications of the existence of ‘‘active cellulose’’, just like its system, the subscripts can be changed, for example by ‘‘tar’’
existence was not being detected in the other experiments and ‘‘gas’’ (see below).
[13,14,19]. Analysis of contradictions of the Broido–Shafiza- One can readily obtain the analytical solution of Eq. (5) in
deh model will be continued in the subsequent article [30]. the form:
 Z t 
2. Simulation 1  a ¼ exp  SðuÞ du
0
2.1. The one-step model with single activation energy
where
   
The simplest equation of a one-step first-order reaction has E1 E2
the form: SðtÞ ¼ A1 exp þ A2 exp
RTðtÞ RTðtÞ
 
da E The differential relationships for the partial conversions, a1
¼ Að1  aÞ exp (1)
dt RTðtÞ and a2, are written as:
   Z t 
We shall consider below mainly the dynamic mode of heating da1 E1
¼ A1 exp exp  SðuÞ du (6)
with linear temperature–time relationships: dt RTðtÞ 0
   Z t 
TðtÞ ¼ T 0 þ at da2 E2
¼ A2 exp exp  SðuÞ du (7)
dt RTðtÞ 0
After integration, Eq. (1) can be regrouped as follows:
   Z t    The partial conversions can be found by the integration of
da E E Eqs. (6) and (7) with separable variables, although the Runge–
¼ A exp exp  A exp du (2)
dt RTðtÞ 0 RTðuÞ Kutta [32] method for the direct solution of Eq. (5) seems
V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165 155

simpler. The final yields, Y1 and Y2, of the corresponding and 0.005 < a < 0.02. In an inert atmosphere [19] cellulose
pathways are calculated as the limits Y1 = a1(t ! 1), should not lose its mass in this temperature region at all. In other
Y2 = a2(t ! 1). words, the effects shown in Fig. 5 are caused solely by the
By differentiating Eqs. (6) and (7) and equating the second presence of oxygen in the ambient gas.
derivatives of a1 and a2 to zero, one can find the symmetrical The calculations lead to Eox = 58.74 kJ/mol, Aox = 14.87 s1
transcendental equations for the peak temperatures of the two and Eox = 91.95 kJ/mol, Aox = 8.16  104 s1 for the approx-
pathways: imation inside the intervals with lower and higher values of a,
respectively. Note that the value of 58.74 kJ/mol corresponds
 ð1Þ 2     
RðTd Þ E1 E2 ideally to the activation energy of 58.14 kJ/mol determined for
a¼ A1 exp ð1Þ
þ A2 exp ð1Þ
(8) the low-temperature step of oxidation of cotton in the research
E1 RTd RTd
of Rychlý et al. [34] by the method of chemiluminescence.
      However, the pre-exponential factor, 14.87 s1, calculated on
ð2Þ 2
RðTd Þ E1 E2 the basis of thermogravimetry is considerably lower than the
a¼ A1 exp þ A2 exp (9)
E2 ð2Þ
RTd
ð2Þ
RTd value of 4.06  103 s1 published in their article [34].
The most noteworthy effect observed in the experiments of
Similar equations are of great value when concentrations of Rychlý et al. [34] is that the different samples of pure cellulose of
chemical products of two competing pathways in the experi- equal mass evince the different chemiluminescent activity. At the
ment are measured separately and independently [21,22]. same time, the calculated activation energies for the different
samples are approximately equal. Unfortunately, they did not
3. Results discus these observations [34], although these facts unambigu-
ously mean that different fragments of cellulose matrix are
3.1. The kinetic constants of the oxidation of cellulose in essentially distinguishable as to their ability to the oxidation.
air In Fig. 6 the calculated reaction rate constants of the
oxidation calculated by the method of thermogravimetry are
Shafizadeh and Bradbury [11] have concluded that the confronted with those tabulated in the article [33]. Excepting
oxidative decomposition of cellulose in air includes the for the thermogravimetric results at a < 0.004, the dependen-
mechanism of autooxidation. Much later on, Rychlý et al. cies obtained by the different methods in air and in oxygen
[33,34] noted the experimental facts indicating that the auto- present almost parallel lines, meaning that in both atmospheres
acceleration inherent in the autooxidation of organic liquids, the oxidation includes the same mechanism. The data presented
gases and fusible polyolefins is absent in the oxidation of in the article [33] lead to Eox = 96.78 kJ/mol (see the dotted
cellulose. Unfortunately, this important conclusion was not gray parallel lines in Fig. 6) corresponding well to 91.95 kJ/mol
discussed in detail in their articles [33,34]. Using the data of in our evaluations.
thermogravimetry, we shall try to substantiate our own The reaction in oxygen passes faster by a factor of 1.73 in
viewpoint concerning this effect. comparison to air. If kinetics with respect to oxygen obeys the
Since the oxidation is expected to be the low-temperature step
[8,11], it is reasonable to evaluate the kinetic constants of this
reaction by handling very short initial fragments of the kinetic
curves. The approximation of the kinetic curves shown Fig. 5 was
implemented by Eq. (4) for very low values of a inside two
different intervals of the mass loss, namely, 0.0005 < a < 0.004

Fig. 6. The logarithms of the reaction rate constants of the oxidation vs. the
inverse temperatures. One can clearly see to what extent the constants deter-
mined by thermogravimetry are lower than those obtained by the methods with
Fig. 5. The initial fragments of the kinetic curves within the range of the the normalization of the initial reaction rate [11,33]. The dotted gray lines are
comparatively low temperatures. parallel and coincide as a result of vertical translation.
156 V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165

first order, this factor would be approximately equal to 100/ The translation shown in Fig. 6 leads to wox ¼ 0:0273, instead
21 = 4.76. Thus, the simple first-order kinetics is not fulfilled of wB ¼ wox ¼ 0:055 used earlier [8] (see also Section 1)
for the oxidation of cellulose. However, taking into account the without a rigorous argumentation.
parallelism of the lines (Fig. 6), one can conclude that the Using only one fitted constant, wox , and the constants
ðtrueÞ
difference in the reaction rates in oxygen and in air is caused Eox = 96.78 kJ/mol, Aox ¼ 1:09  107 s1 determined inde-
mainly by the concentration effect and not by a change of a pendently on the basis of the results published [33] we can
chemical mechanism. predict the kinetic curves over the range 0.005 < a < 0.02
Let us assume that the initial mass of cellulose, m0, includes surprisingly accurately (Fig. 5). At the same time, the
regions of two types. The regions of the first type, with the unsatisfactory approximation of the curves over the range
ð0Þ
initial mass mr , are able to oxidizing by oxygen, while the a < 0.004 means that the decomposition at a very low
ð0Þ
regions of the second type, with the initial mass mi , are inert conversion obeys another mechanism.
with regard to oxygen. The autooxidation occurs due to the accumulation of
When calculating the values of a shown in Fig. 5 we peroxides in a system [35]. These are able to initiate subsequent
proceeded from the assumption that all repeat units of cellulose oxidation by means of their fast destruction with the formation
macromolecules are equally accessible for oxygen, namely, we of either alkoxyl or hydroxyl radicals. The energy of the O–O
assumed that at very small time and a the decomposition is bonds in peroxides and the homolysis temperature (half-
described by the equation: decomposition for one hour) of these bonds [36] are equal to
  about 140 kJ/mol and 160 8C, respectively.
m0 daðapparentÞ dmr ð0Þ Eox
¼ ¼ ðmð0Þ
r þ m i ÞA ðapparentÞ
ox exp The quasi-stationary approximation predicts [e.g. 37] that
dt dt RTðtÞ the decomposition activated by radical reactions has an
(10) apparent activation energy larger than half energy of the bonds
The calculation of the correct pre-exponential factor should most predisposed to homolysis. Thus, if the autooxidation
include only the reactive fraction of cellulose: really controls the decomposition at low temperatures, the
activation energy of the overall process should be greater than
ð0Þ  
mr daðtrueÞ dmr ð0Þ ðtrueÞ Eox 140/2 = 70 kJ/mol. Activation energy ffi90–100 kJ/mol is
¼ ¼ mr Aox exp (11) consistent with the autooxidation mechanism, but the value
dt dt RTðtÞ
of 58–59 kJ/mol is too low to accept it.
We can evaluate the fraction of reactive cellulose by comparing The mass loss in the region a < 0.004 can be ascribed not
ðapparentÞ
the constant Aox with the published constants [11,33] only to chain ends of cellulose macromolecules, structural
determined by using the normalization of the decomposition defects, adsorbed oxygen and other chemically reactive
rate with respect to its initial value. The absolute method [11,33,34] regions accompanying pure cellulose in its real
is equivalent to the measurement of the derivative specimens. It can be caused also by chemical contamination of
ð0Þ ðtrueÞ
dðmr =mr Þ=dt and allows evaluating Aox . Dividing the cellulose. Anyway, one may expect that impurities in cellulose
right-hand side of Eq. (10) by the right-hand side of have an organic nature. Therefore, the oxidation should obey a
Eq. (11) we obtain: common mechanism with defined activation energy.
ð0Þ ðapparentÞ ðapparentÞ From general considerations, the initial step of oxidation is
mr Aox Kox
wox ¼ ¼ ¼ the abstraction of hydrogen from organic molecules by oxygen.
ð0Þ ð0Þ ðtrueÞ ðtrueÞ
ðmr þ mi Þ Aox Kox Probably this step is characterized by activation energy of about

Fig. 7. The scheme of initiation of the cellulose autooxidation. It begins from the abstraction of hydrogen by oxygen at a very low degree of conversion (<0.4%).
Subsequent abstraction of hydrogen is effected by either hydrogen peroxyl or hydroxyl radicals; their concentrations are significantly lower than that of oxygen but
these are much more reactive.
V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165 157

60 kJ/mol, which is lower than the value resulting from the observation of two steps in the degradation. The first one,
autooxidation mechanism. A speculative scheme of the oxidation Eox ffi 60 kJ/mol, is related to the initiation of autooxidation by
of cellulose at a < 0.004 is shown in Fig. 7. According to the the abstraction of hydrogen mainly by oxygen. The rate of the
hypotheses of Shafizadeh and Bradbury [11], the positions ‘‘1’’ second step is controlled by the quasi-stationary decomposition
and ‘‘4’’ are most vulnerable for the oxidation (Fig. 7), although of peroxides. This step is characterized by activation energy of
other positions are believed to be able to react with oxygen as about 90–100 kJ/mol or maybe somewhat higher. In particular,
well [33]. the measurements of chemiluminescence [33,34] lead to
The primary formation of the radicals subsequently leads Eox ffi 100–130 kJ/mol.
either to the formation of peroxides or to the b-scission. The The seeming absence of the auto-acceleration is explained
latter, in turn, can cause either the dehydration or the by the same cause. In order to detect it only the reactions in
depolymerization of cellulose (Fig. 7). Both the reactions surface layers of cellulose should be taken into account;
should be fast in comparison to the reaction of the generation of however, the implemented experiments do not possess the
radicals. If these reactions have close activation energies, inside precision necessary for such measurements.
a narrow temperature interval their yields are determined by a A worthwhile note is the fact that oxygen reacts with
constant proportion. Then the methods of thermogravimetry cellulose all the time [38] from the beginning up to the end of
[8], chemiluminescence and the measurement of degree of the decomposition. The influence of oxygen is detected as the
polymerization (DP) [33,34] should give the close results considerable increase of intensity of the water evolution in
(Fig. 6). comparison to an inert atmosphere [38].
The dehydration seems more preferable in comparison to the The overall amount of water evolved in air increases
depolymerization. Indeed, DP in the course of the oxidation approximately by a factor of two in comparison to the
changes [33] from 3000 to 500 over the range 160–220 8C decomposition in nitrogen [23,38], although the peak
during the time acceptable for a real experiment (ffi24 h). Thus, temperatures of the water evolution in both atmospheres differ
the number of cleavages per one chain is rather small (3000/ inconsiderably. From here we come to the additional
500 = 6). This corresponds to (2  6/3000)  100 = 0.4% of conclusion; namely, oxygen does not affect appreciably the
mass loss, at most. Inasmuch as the real mass loss is much continuous phase of polymer cellulose but it intensively
larger, one can conclude that the dehydration predominates. interacts with products of the degradation. These products form
Probably the b-scission with the release of the hydroxyl new phase (see Section 4.3) presenting the char impregnated
radicals is the major pathway of the dehydration promoting the with high-boiling tar.
charring of cellulose in air (see the next section). Oxygen plays We have revealed that the reaction rate constant of the
the role of the promoter of this process accelerating it by the oxidation is rather high. Therefore, in the case of their
radical mechanism. competition the oxidative depolymerization would suppress the
ðoxÞ
Using Eq. (3) we can evaluate the peak temperature, Td , of purely thermal depolymerization. Thus, we conclude that the
ðtrueÞ
the oxidation. For the sets Eox = 58.74 kJ/mol, Aox ¼ 14:87= oxidation occurs basically as a secondary reaction subsequent
ðtrueÞ
0:0273 s1 ; Eox = 91.95 kJ/mol, Aox ¼ 8:16  104 =0:0273 to the thermal depolymerization passing, in turn, in air or
ðtrueÞ
s1 and Eox = 96.78 kJ/mol, Aox ¼ 1:09  107 s1 at oxygen just like in an inert atmosphere.
a = 2 8C/min we obtain 255, 241 and 239 8C, respectively. So, we find a well-reasoned explanation why oxygen
Thus, if all fragments of cellulose were equally accessible for influences the cellulose degradation (Fig. 1) at the low
oxygen, at 255 8C approximately half of cellulose would be temperatures (step B) but ceases to affect the decomposition
oxidized in air with the considerable mass loss. However, the at higher temperatures (step C). Further analysis should clarify
measured mass loss [8] of the cellulosic sample at 255 8C is less to what extent the above considerations conform to the
than 3% of its initial mass. Using the method of proof by pyrolytic experiments carried out in an inert atmosphere.
reduction to absurdity, we conclude once again that only a small
fraction of cellulose is able to oxidation at low temperatures. 3.2. The data on cellulose pyrolysis
Probably only surface cellulose is accessible for oxygen at
the initial moment of cellulose degradation, while crystalloid In Table 1 and in Figs. 8 and 9 we compare our results [8]
blocks and even amorphous (vitreous) cellulose blocks with the results of a few publications [12,16,19,21,27] devoted
possessing a structure of the solid continuum are impenetrable specially to the approximation of kinetic information concern-
(hermetic) for oxygen. ing the decomposition of pure cellulose in an inert atmosphere.
In our opinion, no new experimental facts, which appeared In two publications [12,16] the authors used the one-level
during last three decades, disprove the conclusion of Shafizadeh model with the single activation energy. The corresponding
and Bradbury [11] about the mechanism of the autooxidation in peak temperatures were calculated by Eq. (3).
the cellulose degradation. However, the concentration of The authors of three publications [19,21,27] implied the
peroxides should be much lower than it is reached in the existence of two reaction pathways, Ktar and Kgas. Their peak
course of oxidative degradation of liquid organic substances or temperatures were calculated by Eqs. (8) and (9). The step of
fusible organic polymers such as polyolefins. Indeed, only the initiation connected with the formation of ‘‘active
surface pyranose rings of cellulose are able to the formation of cellulose’’ was disregarded in these calculations, because it
peroxides. The low rate of accumulation of peroxides allows the has much lower peak temperature (see Table 1). The conclusion
158
Table 1
The kinetic constants for the different steps of the cellulose decomposition and the corresponding peak temperatures at different heating rates
MTG in air [8] Banyasz et al. Bradbury et al. Capart et al. Milosavljevic and Milosavljevic Antal et al. [12] Kashiwagi and Kashiwagi and
[21] in N2 at [27] in vacuum [19] in N2 Suuberg [16] in N2 and Suuberg [16] in N2 for Avicel Nambu [38] Nambu [38]
fast pyrolysis at slow pyrolysis in N2 at fast PH-105 for water peaks for water peaks
for CF-11 pyrolysis for CF-11 in air in N2
Eox (Eac?) (kJ/mol) 96.78 242.8
Aox (Aac?) (s1) 1.09  10 7 2.8  1019
ðoxÞ ðacÞ
ðTd ?Þ ð CÞ

V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165


Td
2 8C/min 239.1 303.7
6 8C/min 262.8 316.0
Etar = EC (kJ/mol) 195.9 151.1 198.0 202.7 218 145 195 176.4 259.3
Atar = AC (s1) 8.9  10 14 4.0  10 10 3.2  1014 1.46  1015 1.58  1016 7.37  109 2.69  1014 6.22  1013 6.92  1020
ðtarÞ ðCÞ
Td ¼ Td ð CÞ, at 2 8C/min
Calculated 308.7 313.3 325.2 316.6 331.8 325.2 322.9 287.2 306.8
Experimental 310.0
Ytar (%) 81.1 72.1 72 100 100 100
ðtarÞ ðCÞ
Td ¼ Td ð CÞ, at 6 8C/min
Calculated 324.1 332.3 342.1 330.8 346.8 347.0 339.1 303.1 318.5
Experimental 328.0
Ytar (%) 76.3 76.8 66.7 100 100 100
ðtarÞ ðCÞ
Td ¼ Td ð CÞ, at 65 8C/min 360.5 376.6 381.9 363.2 382.0 400.1 377.5 340.7 345.4
Ytar (%) 63.7 84.9 54.1 100 100 100
Egas (kJ/mol) 195.9 153.2 255.0
Agas (s1) 1.0  10 14 1.3  1010 2.66  1019
ðgasÞ
Td ð CÞ, at 2 8C/min 317.7 321.3 319.5
Ygas (%) 18.9 27.9 28 0 0 0
ðgasÞ
Td ð CÞ, at 6 8C/min 336.9 338.1 333.8
Ygas (%) 23.7 23.2 33.3 0 0 0
ðgasÞ  381.7 377.4 366.5
Td ð CÞ, at 65 8C/min
Ygas (%) 36.3 15.1 45.9 0 0 0
V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165 159

The integration of the differential equations was imple-


mented by the Runge–Kutta [32] method. For the calculation of
the kinetic curves displayed in Figs. 8 and 9 we used the first-
order kinetics equation for each of the reaction channels.
To make the correction for the oxidation we used the
assumption that 2.73% of cellulose decomposes by the
ðtrueÞ
mechanism with Eox = 96.78 kJ/mol, Aox ¼ 1:09  107 s1 ,
while the remaining fraction, 97.27%, decomposes just like
cellulose in an inert atmosphere.
The model used by Capart et al. [19] and the Broido–
Fig. 8. Comparison of the experimental curve of the decomposition of cotton in Shafizadeh model [27] imply non-zero char yield at the end of
air at the heating rate of 2 8C/min with the curves calculated by the one-step the pyrolysis (Fig. 9b and c). In the first case [19] the yield of
model. char is assumed unchangeable during all the pyrolysis and is
accepted as 8.66%. In the Broido–Shafizadeh model [27] the
of Bradbury et al. [27] that at temperatures above 300 8C this char formation combines together with the light gases in the
step may be ignored seems correct (Fig. 9c). route Kgas (Fig. 3) and corresponds to 35% with respect to a
In Table 1 we used three heating rates. Two of these, 2 and total yield of this pathway. Two approaches give the close
6 8C/min, correspond to our experiment (Fig. 2). The third rate, results (Fig. 9b and c).
65 8C/min, was selected to investigate the increase of the peak In the publication of Capart et al. [19] the fractions wtar ; wgas
temperatures when going from slow to fast pyrolysis. We note and the pre-exponential factors Atar, Agas for two steps of
that the heating rate of 65 8C/min was used repeatedly in the decomposition, which they postulated in their study, are given
experiments [12,16] as a suitable value to trace changes in separately. They used the model of parallel reactions and not the
kinetic information in the course of the different pyrolysis competition model; the latter includes only the products wtar Atar
regimes. and wgas Agas . In the case of two reactions the competition model
involves four ðwtar Atar ; wgas Agas ; Etar ; Egas Þ and not six
ðwtar ; wgas ; Atar ; Agas ; Etar ; Egas Þ sought parameters. For
the construction of Table 1 we used the products wtar Atar and
wgas Agas instead of Atar and Agas. After this remark we can
compare the data of Capart et al. [19] with the other results.
The calculated peak temperatures for the pathway Ktar are
situated closely with respect to both each other and to the peak
temperature for the main step of the oxidative decomposition.
Unfortunately, the deviations in the peak temperatures
corresponding to the different publications do not allow
identifying the pathways. However, together with the calcu-
lated yields of the pathways, Ytar and Ygas, the identification
becomes reliable and unequivocal.
Only in two [21,27] of three publications [19,21,27] the rate-
limiting pathways were identified chemically. However, the
comparison of their yields allows the assignment of the
prevailing step in the research of Capart et al. [19] to the process
of the tar formation as well. One may conclude also that the
authors employing the one-step model (Fig. 8) actually
evaluated the kinetic constants of the tar formation.
In order to show that the evolution of water is related to the
tar formation, we have inserted in Table 1 the corresponding
peak temperatures calculated on the basis of the data of
Kashiwagi and Nambu [38]. The calculation of the kinetic
constants was organized by using the experimental peak
temperatures and Eq. (3) inverted with respect to E and A.
Within the experimental errors the approximated peak
temperatures at 2 and 6 8C/min coincide with the data of the
measurements [38]. At the same time, the formal calculation of
the activation energy by using the peak temperatures does not
Fig. 9. Comparison of the experimental curve of the decomposition of cotton in
air at the heating rate of 2 8C/min with the curves calculated by the two-step have a rigorous physical sense, inasmuch as a yield of water
model of the competition (a–c) and by the Broido–Shafizadeh model with the depends on heating rate and changes for different cellulosic
step of initiation (c). samples [23]. Thus, the theoretical peak temperature of
160 V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165

345.4 8C at 65 8C/min predicted by means of the far cellulose pyrolysis. Thus, we can depict the main step of the
extrapolation of the data at lower temperatures most probably cellulose degradation by the scheme shown in Fig. 10.
is erroneous. Levoglucosan is a thermally stable compound [43]; at 300–
Generally, the belonging of the water evolution to the 400 8C it leaves the pyrolysis zone before an appreciable
process of the tar formation seems proven, inasmuch as the decomposition. Neither the dehydration (evolution [23,38] of
peak temperatures for water are separated from the peaks of the H2O) nor the decarboxylation (evolution [20–22] of CO2) can
overall mass loss only by 7–20 8C. Using the data in Table 1 we be attributed to its destruction. The only remaining possibility is
can reliably identify the main step of the cellulose degradation to ascribe the source of these gases to the non-reducing hydroxy
chemically. glycosidic chain ends with pyranose (Fig. 10). These appear at
each elementary step of the depolymerization by transglyco-
4. Discussion sylation.
The non-reducing ends can be removed by the process called
4.1. The main step of the decomposition by Boon and co-workers [44–46] as ‘‘Ei-elimination’’. The latter
process should be slower than transglycosylation, since
Table 1 is an illustrative example of manifesting the so- otherwise the yield of levoglucosan could not reach 60%.
called compensation effect [7]. An error in a pre-exponential Consequently, the pathway of the Ei-elimination does not lead to
factor compensates an error in the corresponding activation the CO2 evolution. Rather, this pathway leads to production of
energy. In this situation only the peak temperatures together light gases due to the fragmentation of the unstable reducing
with the contributions of the pathways can serve for the chain ends (the ends with the semi-acetal groups) resulting from
objective chemical identification of a process. the Ei-elimination. The reducing chain ends are able to fast ring
Banyasz et al. [21,22] calculated the constants Atar and Etar cleavage with the formation of chain-end glucose and probably
by using information about the evolution of the only gas, CO2. different variants of its dehydrated and cyclic derivatives such as
Table 1 really proves that the pathway for the CO2 formation 5-hydroxymethyl-2-furfural [43]. Discussion of this mechanism
coincides with the pathway of the tar formation. However, the remains beyond the scope of the present article. Nevertheless, if
value of activation energy alone cannot serve as a criterion for accepting the scheme in Fig. 10, we have to explain here a few
chemical identification of a pathway. Indeed, both Banyasz additional features of the phenomenon.
et al. [21,22] and Bradbury et al. [27] correctly attributed the
dominant pathway to the process of the tar formation but the 4.2. Indirect information reflecting a hidden mechanism
calculated activation energies, 151.1 and 198 kJ/mol, corre-
sponding to the route Ktar considerably differ (see the polemics One of evident questions needing an answer relates to the
in the articles [21,22]). The constant Etar = 195.9 kJ/mol high possible yield of levoglucosan. In particular, for cotton it
obtained in our calculations coincides ideally with that of can reach 57% [40]. The cotton cellulose possesses the
Banyasz, Egas = 195.9 kJ/mol, but these fall into the opposite maximal degree of polymerization, namely, 9  103 to
sides of Table 1. 1.5  104 monomer units per chain [47]. The dehydration of
Antal et al. [12,13] reasonably explained the apparent the non-reducing chain ends seems to occur really fast, since
decrease of calculated activation energies at fast pyrolysis by the water peaks are situated not only close by the peaks of mass
the thermal lag inherent in large cellulose samples. In two cited loss but even precede these (see Table 1). The dehydration
publications [16,21] the energy Etar was evaluated for the blocks the transglycosylation (Fig. 10). In other words,
conditions of fast pyrolysis and its value is really lower than that dehydrated pyranose does not convert to levoglucosan.
determined for slow pyrolysis. If the thermal lag is maximally Therefore, random scission of macromolecules should not
diminished (see Table 1), the activation energy Etar calculated lead to the yield of levoglucosan exceeding 50%.
tends to its correct value, 200 kJ/mol. We have to explain why the unzipping of cellulose via the
The major fraction of volatile tar presents levoglucosan (1,6- chain ends is preferable in comparison to the random scission.
anhydroglucose) that is derived in cellulose during pyrolysis by Another question is connected with the displacement of the
the reaction of transglycosylation. Shafizadeh et al. [39,40] have kinetic curves on the scale of temperature for different
shown that the yield of levoglucosan from pure cellulose can cellulosic samples [12].
reach 60%. Transglycosylation additionally leads to the dimer Let us scrutinize the differences in the curves displayed in
and the trimer, namely, to cellobiosan [41] and cellotriosan [42], Figs. 8 and 9 once more. Although we used the theoretical
respectively. In particular, cellobiosan is derived with the yield of constants instead of experimental data, the accurate approx-
up to 6–15% [41]. The compounds consisting of two and more imation reached in all publications cited in Table 1 allows us to
pyranose rings are components of non-volatile tar; their regard the calculated dependencies as the replicas of the
volatilization into a gas phase is possible only after their experimental ones.
additional fragmentation with the evolution of light gases. The smaller activation energy is, the more sensitive kinetic
Therefore, the transglycosylation followed by the evolution of curves are to heating rate. Therefore, if using the kinetic
light gases leads to the summarized mass loss exceeding 60% constants obtained for fast pyrolysis to simulate slow pyrolysis,
(see Table 1). The above facts allow one to conclude that the theoretical curves of mass loss predict the experimental
transglycosylation is the essentially prevailing pathway of ones with an outstripping. The data of Banyasz et al. [21]
V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165 161

Fig. 10. The kinetic scheme that clarifies the domination of transglycosylation with Etar  200 kJ/mol on the main step of cellulose pyrolysis. The rate limitation is
observed [19] over the range 0.1 < a < 0.7, although the maximum yield of levoglucosan [39,40] does not exceed 60%. Thus, the pathway of transglycosylation leads
to the evolution of other gases in addition to levoglucosan, for example such as H2O and CO2.

demonstrate this effect with respect to the experimental curve Worthy of note is the fact that Antal et al. [12] demonstrated
(Fig. 9a). The data of Milosavljevic and Suuberg [16] even larger shift between kinetic curves for different cellulosic
illustrate it for the simulated dependencies (Fig. 8). For samples, up to 40 8C. Ultimately the authors wrote: ‘‘Each
reliability of conclusions the data obtained with the constants cellulose is a unique material!’’. Such an explanation seems
for fast pyrolysis should be disregarded, since these include unsatisfactory. Indeed, pure cellulose is a substance with the
the expectable error. At the same time, the difference invariable chemical formula and the spatial pattern. Thus, it is
between the curves of mass loss for different samples is not natural to expect that in all cases the predominating pathway is
an artifact. unchangeable.
The distinction between decomposition of Avicel PH-105 Either the scheme in Fig. 10 is fallacious or the distinctions
and Whatman CF-11 was shown and discussed in the article of between different types of cellulose have physical rather than
Antal et al. [12]. At a = 1 8C/min the experimental curve of chemical nature. One should solve this dilemma by considering
mass loss for Avicel cellulose is shifted by 15 8C with respect to the degradation with due regard for peculiarities of aggregated
the curve for CF-11 cellulose toward lower temperatures. The states of cellulose.
same effect is observed for the simulated curves, although the
predicted difference is somewhat lower and equals to 331.8– 4.3. Two-phase model of the cellulose degradation
322.9 = 8.9 8C (see Table 1 and Fig. 8). The small discrepancy
results from simulation errors. In the experiments of Antal et al. [12] Avicel PH-105
The distance of 15 8C between the curve for cotton and the cellulose pyrolyzed faster than other samples. Cotton pyrolyzes
curve reconstructed via the constants of Bradbury et al. [27] is even faster than Avicel cellulose (Fig. 8). That is why cotton is
large to such an extent (Fig. 9c) that it cannot be ascribed to convenient as a benchmark for pyrolysis of other cellulosic
simulation errors and surely reflects experimental behavior of materials. However, at first one should find an explanation why
the samples. cotton decomposes faster than other samples.
162 V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165

Cotton [48] consists of cellulose of several types: surface hydrolyze cellulose. The dehydration of the monomer units can
cellulose (7.8%), Ib-cellulose (27.6%), paracrystalline cellu- start only after the hydrolysis. The hydrolysis is a rather slow
lose (33.1%), amorphous cellulose (24.1%), Ia-cellulose process that starts exceptionally from outer surface of
(4.2%), and hemicellulose (3.2%). microfibrils; therefore charring of cellulose does not occur
On the basis of the results in Section 3.1 we conclude so efficiently as charring of sugar.
that only surface cellulose undergoes the oxidation at the low Similarly, solid structures of cellulose are inaccessible for
degree of conversion. According to our assessment (2.73/ oxygen (Section 3.1). One may conclude also that these should
7.8)  100 = 35% of surface cellulose turns into gas at the be impenetrable for molecules with larger van der Waals
beginning of the oxidative degradation. Such an evaluation radii.
seems realistic, although the percentages in both numerator and We note that during and after the decomposition in an inert
denominator are determined rather roughly. atmosphere the skeleton of cellulose is completely preserved
The school experiment with sugar placed in concentrated (see Fig. 11), and so the decomposition occurs in internal pores
sulfuric acid showily illustrates charring of sugar at the room of the fibers. The thermal destruction of cellulose should lead
temperature. This reaction takes place at the low temperature to the formation of cavities inside each of microfibrils.
due to acid catalysis that is followed by the strong hydration of As temperature grows, hydrogen bonds in cellulose begin to
H2SO4. collapse and the microfibrils soften. The cavities begin to
Taking into account the chemical likeness of cellulose and aggregate into a net of pores and decomposition gases leave the
sugar (sucrose), one may expect the same reactions for cellulose matrix, resulting in mass loss.
cellulose. Indeed, kinetics of its pyrolysis appreciably changes Capart et al. [19] have shown the results of the isothermal
after impregnation of cellulose only with 1% H2SO4 [29,49]. experiments that unambiguously prove that cellulose pyrolysis
However, the action of H2SO4 is not as spectacular as in the is an auto-accelerated process. The initial step of the process,
case of sugar. The cause is the special aggregation of cellulose according to their conclusions, conforms to the nucleation
that forms crystalloid domains in vegetable structures. model. For the approximation of the kinetic curves we selected
Atalla [48] wrote: ‘‘The unusual nature of cellulose in its the one-parameter nucleation model as well [6–8]. It is
aggregated states is perhaps best represented by its relationship noteworthy that the parameter of nucleation is the same as in the
to water; a primary structure with three hydroxyl groups per article of Capart et al. [19], although they have additionally
pyranose ring would, in the normal course of events, be improved the approximation accuracy by using the model of
expected to be quite soluble in water and aqueous media. But it Prout–Tompkins. However, our conclusions are mainly based
is, in fact, essentially insoluble, and this characteristic is but one on the function E(a) that does not depend on a model used [8].
of the many unusual patterns of behavior of the aggregated Appearance of gases, liquids and solid char in the volume of
states of cellulose’’. microfibrils should somehow deform the cellulose structure.
The stability of cellulose is caused by the intramolecular and We can assume that because of such deformations crystalline
intermolecular hydrogen bonds regularly repeating along each cellulose turns into amorphous one [52]; however, cellulose is
cellulose chain. Cellulose chains are aggregated into either not able to melting (see Fig. 11) even at very high temperatures.
crystalline or vitreous blocks. In particular, the crystalline The energy of cohesion of cellulose chains with each other
polymorphs Ia and Ib of native cellulose have the sheet-like seems to be very large even after considerable depolymeriza-
structure. The pyranose rings lie in the planes of the sheets. The tion.
sheets are connected exceptionally by the van der Waals forces. At the beginning of the decomposition the appearance of
Hydrogen bonds exist only within the sheets [50,51]. low-molecular extraneous substances probably can be likened
In cotton the crystals form linear microfibrils with the to point defects in crystals. According to the kinetic model, it is
diameter ffi10–20 nm. An agglomeration of the microfibrils the stage of nucleation. However, the stage of nucleation is
forms a cotton fiber. followed by the stage of the growth of nuclei. Continuous
The chemically resistant crystalloid structure prevents the nucleation results in an increase of the surface free energy
penetration of chemical agents, even such aggressive as H2SO4, corresponding to the interface between polymer cellulose and
into the microfibrils of cellulose. At first sulfuric acid should the extraneous cavities in its structure. Thus, the formation of

Fig. 11. Scanning electron micrographs of cotton cellulose and the carbonized residue derived by 800 8C treatment in nitrogen: (a) original cotton, (b) after
carbonization with H2SO4 and (c) after carbonization without H2SO4 (from Kim et al. [29]).
V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165 163

The model shown in Fig. 12 gives one of possible


explanations for the distinctions between different types of
cellulose [12]. Different surface area of the cavities leads to
different rates of decomposition. We should revert to the data of
Bradbury et al. [27] to consider these from the viewpoint of the
model of nucleation.

4.4. Another explanation of the lag in mass loss at the


beginning of pyrolysis

We note that the activation energy Eac = 242.8 kJ/mol


calculated by Bradbury et al. [27] for the step of activation is
abnormally high, despite the low temperatures predicted for
this step (see Table 1 and Fig. 1a).
Although, according to Bradbury et al. [27], the formation of
‘‘active cellulose’’ does not lead to a mass loss, this step
Fig. 12. The model of a pore (cavity) inside the volume of undecomposed influences considerably the decomposition kinetics. Therefore,
cellulose. Depending on a stage of degradation and heating rate, the pore can
contain gas, liquid (tar or products of depolymerization) or solid (char). Any
if a chemical activation of cellulose would really take place, one
combination of these products is possible as well, although coexistence of gas should observe a deviation of activation energy from the
bubbles inside the liquid tar is problematic because of the large Laplace pressure constant level in Fig. 1a. The absence of any indications of an
in such bubbles. The pore is surrounded by the surface consisting of the chain additional step in both our study and model-free computations
ends of polymer cellulose. At the initial step of degradation (a) polymer of Capart et al. [19] means that the additional step is
cellulose contains inclusions of the extraneous cavities, while at the end of
degradation (b) the polymer cellulose itself presents inclusions inside the
unnecessary in the model and can be excluded by a proper
microporous carbonized residue. selection of a conversion function distinct from the first-order
kinetics equation.
large cavities is preferable. The cavities of a macroscopic size In Fig. 13 one can see the results of the use of the model of
should be considered as a new phase (Fig. 12). nucleation instead of first-order kinetics. The parameter of
If we cut off any woven fabric, we see the butts of fibers on nucleation 1.5 was used in these calculations by analogy with
the boundary of the torn fabric. By analogy to this figurative other researches [7,19]. The selection of such a parameter
example, the surface of the cavities (entirety ruptures) is the seems successful, since after a small correction of the pre-
region of location of chain ends of cellulose (Fig. 12). If our exponential factors the theoretical curves might approximate
considerations are correct, depolymerization is a migration of the experimental data of Bradbury et al. [27] ideally.
the chain ends from one phase (structured polymer cellulose) We find an explanation why the curve E(a) does not contain
into another phase (char and other decomposition products). any anomalies in the region of the rate limitation on step C
The scission of only chain ends seems preferable in comparison (Fig. 1a). The delayed decomposition at the beginning of the
to a random scission of bonds in polymer cellulose, since the pyrolysis is connected with the accumulation of nuclei in
latter process leads to the formation of new cavities. We find a cellulose.
reasonable explanation why the yield of levoglucosan can An attainable size of nuclei should depend on size of
exceed 50%. particles of cellulose. Rate of the pyrolysis increases with
Solid char inside the cavities appears in the form of a very increasing surface area of the interface between the phase of
fine dispersion suspended in the liquid tar. The particles of char
are fastened with each other rather weakly, therefore, as tar
gasifies, the fibers of cotton become thinner (compare Fig. 11a
and c). Their mean radius decreases from ffi7 to ffi3 mm.
Penetrating into the cavities, sulfuric acid dissolves in tar and
promotes its charring. In the latter case the particles of char are
firmly fastened with each other and the radius of the fibers
remains almost unchangeable (compare Fig. 11a and b),
although in any case the greater part of cellulose turns into gas.
Just like sulfuric acid, oxygen does not penetrate into
polymer cellulose but it intensively interacts with products of
cellulose degradation promoting their charring. Indeed, Fig. 9b
and c clearly illustrates the increase of the char yield from 8% to
30% in comparison to the decomposition in an inert
Fig. 13. Comparison of the kinetic curves calculated by the three-step Broido–
atmosphere. Probably the intensification of charring under Shafizadeh model and by the two-step model of nucleation. One can see that the
action of oxygen is caused by the promotion of dehydration (see effect of nucleation leads to the lag described in the Broido–Shafizadeh model
our considerations in Section 3.1). by the additional step of the formation of ‘‘active cellulose’’. a = 2 8C/min.
164 V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165

polymer cellulose and the phase of products. In other words, characterized by the activation energy of about 60 kJ/mol. At
many nuclei with small sizes provide faster pyrolysis in higher conversions the rate of the decomposition is character-
comparison to few nuclei with large sizes. That is why cotton ized by the autooxidation with activation energy of about
pyrolyzes faster than other samples. In cotton the maximum 100 kJ/mol.
attainable size of nuclei is equal to the diametrical size of Oxygen intensively reacts with products of the thermal
microfibrils. In cellulose consisting of larger crystalloid blocks depolymerization of cellulose, although it does not penetrate
this size should be larger, correspondingly, the surface area of into the matrix of polymer cellulose. In other words, the
the interface should be smaller. oxidation is a fast process subsequent to the thermal
Let us assume that the decomposition of the samples used in depolymerization. Since gasification of cellulose caused by
the research of Bradbury et al. [27] occurred just like the the evolution of volatile substances and by the oxidation of non-
decomposition of cotton but the cavities had the surface area volatile substances is determined by the thermal depolymer-
smaller than that for cotton by a factor of 2.2. The kinetic curve ization, the rate of cellulose decomposition is essentially the
recalculated with the pre-exponential factors multiplied by 2.2 same in both air and nitrogen.
approaches the kinetic curve for cotton (Fig. 13). A decrease of The proposed two-phase model of cellulose is able to
the surface area by a factor of 10 leads to the increase of the explain all experimental observations. According to the model,
peak temperature from 310 to 350 8C. Despite the similarity of the thermal decomposition is a migration of chain ends from the
the mechanism of nucleation, one may assume that the 10-fold phase of polymer cellulose to the phase presenting char
change of the surface area of the interface is possible for impregnated with high-boiling tar. Unzipping of cellulose at the
cellulose of different origin. Thus, the two-phase model interface between two phases is preferable in comparison to the
explains the distinction of cellulosic samples shown by Antal random scission of the macromolecules. That is why the yield
et al. [12]. of levoglucosan can reach 60%.
The approximation of the kinetic curve for cotton by the Under conditions usual for thermogravimetry the trans-
constants of Bradbury et al. [27] seems unsatisfactory (Fig. 13). glycosylation with activation energy of about 200 kJ/mol is
Probably conversion functions for different cellulosic samples the predominating reaction. An increase of the surface area of
should be selected individually. the interface between two phases accelerates the decomposi-
Correction for oxidation is not accurate either. In particular, tion. Although the nucleation resulting in appearance of the
we ignored the fact that the true boiling point of levoglucosan is phase of products is the general phenomenon for all types of
about 260 8C [17,18]. The lag of mass loss below such a cellulose, cellulose consisting of smaller particles decom-
temperature can be explained by non-volatility of this poses faster than cellulose consisting of larger particles. One
substance; this distorts the nucleation model [19]. Oxygen may expect that cotton, where the structural formations
undoubtedly abridges the lag [38] due to the oxidative determining the size of nuclei are microfibrils with diameters
gasification of levoglucosan and other non-volatile products ffi10–20 nm, decomposes faster than most of other types of
of the depolymerization [8]. cellulose.
We disregarded the interaction of oxygen with decomposi-
tion products at all, although this interaction surely influences Acknowledgements
kinetic curves. Such an influence includes two opposite effects.
On the one hand, oxygen promotes the gasification of high- Vadim Mamleev is grateful to ‘Region Nord-pas-de-Calais’
boiling tar, in particular, the water evolution [38]. On the other (France) for the financial support and to the laboratories PERF
hand, it promotes charring of cellulose. Therefore, kinetics of (ENSC, Lille) and LEM (ENSG, Nancy) for his visits in Lille
oxidative decomposition only approximately corresponds to and Nancy for the joint work on thermal analysis. The authors
kinetics of pyrolysis. However, when considering the general thank Doctor Dae-young Kim and Professors Shigenori Kuga,
tendencies of the process the differences between these regimes Yoshiharu Nishiyama, Masahisa Wada from Department of
of cellulose degradation at the step of the rate limitation are Biomaterials Science (University of Tokyo, Japan) for their
small enough and can be disregarded. kind permission to use their unique microphotographs as well
as Professor Takashi Kashiwagi from National Institute of
5. Conclusions Standards and Technology (Gaithersburg, USA) for his data
about the water evolution. They deeply appreciate the help
Before the decomposition, cellulose consists of solid of Doctor Joseph Banyasz from Philip Morris Company
particles with the crystalloid or amorphous structure impene- (Richmond, USA) in the interpretation of the rate-limiting steps
trable for chemical reagents. Oxygen at low temperatures and in the detailing of the kinetic scheme of the cellulose
interacts only with surface cellulose. Only ffi3% of cellulose decomposition.
gasifies as a result of the oxidation at the early stage of the
decomposition preceding considerable thermal depolymeriza-
References
tion.
At a very low conversion (<0.4%) oxygen initiates the [1] D. Klemm, B. Philipp, T. Heinze, Comprehensive Cellulose Chemistry:
abstraction of hydrogen from cellulose molecules with the Fundamentals and Analytical Methods., vol. 1, Wiley–VCH, Weinheim,
formation of the hydrogen peroxyl radicals. This reaction is 1998.
V. Mamleev et al. / J. Anal. Appl. Pyrolysis 80 (2007) 151–165 165

[2] D. Klemm, B. Philipp, T. Heinze, U. Heinze, W. Wagenknecht, Compre- [28] F. Shafizadeh, in: R.P. Overend, T.A. Milne, L.K. Mudge (Eds.), Funda-
hensive Cellulose Chemistry: Functionalization of Cellulose, vol. 2, mentals of Thermo-Chemical Biomass Conversion, Elsevier, 1985, pp.
Wiley–VCH, Weinheim, 1998. 183–217.
[3] A.V. Bridgwater (Ed.), Fast Pyrolysis of Biomass: A Handbook, vol. 2, [29] D.-Y. Kim, Y. Nishiyama, M. Wada, S. Kuga, in: Proceedings of 2nd
CPL Press, Newbury, 2002. Annual Partnerships for Environmental Improvement and Economic
[4] B.N. Kusnetsov, Soros Educ. J. 12 (1996) 47–55 (in Russ.). Development Conference: ‘‘Wood and Cellulose: Building Blocks for
[5] L.S. Gal’braikh, Soros Educ. J. 11 (1996) 47–53 (in Russ.). Chemicals, Fuels and Advanced Materials’’, Syracuse, New York, 2000.
[6] S. Bourbigot, S. Chlebicki, V. Mamleev, Polym. Degrad. Stabil. 78 (2002) [30] V. Mamleev, S. Bourbigot, J. Yvon, J. Anal. Appl. Pyrol. 80 (2007) 141–
57–62. 150.
[7] V. Mamleev, S. Bourbigot, Chem. Eng. Sci. 60 (2005) 747–766. [31] A.W. Coats, J.P. Redfern, Nature 201 (1964) 68–69.
[8] V. Mamleev, S. Bourbigot, M. Le Bras, J. Yvon, J. Lefebvre, Chem. Eng. [32] G.F. Forsite, M.A. Malcolm, C.B. Moler, Computer Methods for Math-
Sci. 61 (2006) 1272–1288. ematical Computations, Prentice-Hall, Englewood Cliffs, NJ, 1977.
[9] V. Mamleev, S. Bourbigot, J. Therm. Anal. Calorim. 70 (2002) 565–579. [33] J. Rychlý, M. Strlič, L. Matisová-Rychlá, J. Kolar, Polym. Degrad. Stabil.
[10] V. Mamleev, S. Bourbigot, M. Le Bras, J. Lefebvre, J. Therm. Anal. 78 (2002) 357–367.
Calorim. 78 (2004) 1009–1027. [34] J. Rychlý, L. Matisová-Rychlá, M. Lazár, K. Slovák, M. Strlič, D. Kočar, J.
[11] F. Shafizadeh, A.G.W. Bradbury, J. Appl. Polym. Sci. 23 (1979) 1431– Kolar, Carbohydr. Polym. 58 (2004) 301–309.
1442. [35] R.B. Grossman, The Art of Writing Reasonable Organic Reaction
[12] M.J. Antal Jr., G. Várhegyi, E. Jakab, Ind. Eng. Chem. Res. 37 (1998) Mechanisms, Springer, New York, 2003.
1267–1275. [36] W.H. Reusch, An Introduction to Organic Chemistry, Holden-Day, San
[13] G. Várhegyi, M.J. Antal Jr., E. Jakab, P. Szabó, J. Anal. Appl. Pyrol. 42 Francisco, 1977.
(1997) 73–87. [37] H. Bockhorn, A. Hornung, U. Hornung, J. Anal. Appl. Pyrol. 50 (1999)
[14] M.J. Antal Jr., G. Várhegyi, Ind. Eng. Chem. Res. 34 (1995) 703–717. 77–101.
[15] M.J. Antal Jr., M. Grønli, Ind. Eng. Chem. Res. 42 (2003) 1619–1640. [38] T. Kashiwagi, H. Nambu, Combust. Flame 88 (1992) 345–368.
[16] I. Milosavljevic, E.M. Suuberg, Ind. Eng. Chem. Res. 34 (1995) 1081– [39] F. Shafizadeh, J. Anal. Appl. Pyrol. 3 (1982) 283–305.
1091. [40] F. Shafizadeh, R.H. Furneaux, T.G. Cochran, J.P. Scholl, Y. Sakai, J. Appl.
[17] I. Milosavljevic, V. Oja, E.M. Suuberg, Ind. Eng. Chem. Res. 35 (1996) Polym. Sci. 23 (1979) 3525–3539.
653–662. [41] D. Radlein, A. Grinshpun, J. Piskorz, D.S. Scott, J. Anal. Appl. Pyrol. 12
[18] E.M. Suuberg, I. Milosavljevic, V. Oja, in: Proceedings of 26th Sympo- (1987) 39–49.
sium (International) on Combustion, The Combustion Institute, Pitts- [42] P. Arisz, J.A. Lomax, J.J. Boon, Anal. Chem. 62 (1990) 1519–1522.
burgh, 1996. [43] E.B. Sanders, A.I. Goldsmith, J.I. Seeman, J. Anal. Appl. Pyrol. 66 (2003)
[19] R. Capart, L. Khezami, A.K. Burnham, Thermochim. Acta 417 (2004) 79– 29–50.
89. [44] J.A. Lomax, J.M. Commandeur, P.W. Arisz, J.J. Boon, J. Anal. Appl.
[20] S. Li, J. Lyons-Hart, J.L. Banyasz, K.H. Shafer, Fuel 80 (2001) 1809– Pyrol. 19 (1991) 65–79.
1817. [45] P.W. Arisz, G.B. Eijkel, J.J. Boon, J. Anal. Appl. Pyrol. 33 (1995) 21–38.
[21] J.L. Banyasz, S. Li, J. Lyons-Hart, K.H. Shafer, Fuel 80 (2001) 1757– [46] P.W. Arisz, J.J. Boon, J. Anal. Appl. Pyrol. 25 (1993) 371–385.
1763. [47] M. Joseph, Introductory Textile Science, Holt Reinhold and Winston, Fort
[22] J.L. Banyasz, S. Li, J. Lyons-Hart, K.H. Shafer, J. Anal. Appl. Pyrol. 57 Worth, 1986.
(2001) 223–248. [48] R.H. Atalla, in: B.M. Pinto, D. Barton, K. Nakanishi, O. Meth-Cohn
[23] J. Scheirs, G. Camino, W. Tumiatti, Eur. Polym. J. 37 (2001) 933–942. (Eds.), Comprehensive Natural Products Chemistry, vol. 3, Elsevier,
[24] F.J. Kilzer, A. Broido, Pyrodynamics 2 (1965) 151–163. Oxford, 1999 (Chapter 16).
[25] A. Broido, M.A. Nelson, Combust. Flame 24 (1975) 263–268. [49] S. Julien, E. Chornet, R.P. Overend, J. Anal. Appl. Pyrol. 27 (1993) 25–43.
[26] A. Broido, in: F. Shafizadeh, K.V. Sarkanen, D.A. Tillman (Eds.), Thermal [50] Y. Nishiyama, P. Langan, H. Chanzy, J. Am. Chem. Soc. 124 (2002) 9074–
Uses and Properties of Carbohydrates and Lignins, Academic Press, New 9082.
York, 1976, pp. 19–35. [51] Y. Nishiyama, J. Sugiyama, H. Chanzy, P. Langan, J. Am. Chem. Soc. 125
[27] A.G.W. Bradbury, Y. Sakai, F. Shafizadeh, J. Appl. Polym. Sci. 23 (1979) (2003) 14300–14306.
3271–3280. [52] J.B. Wooten, J.I. Seeman, M.R. Hajaligol, Energy Fuels 18 (2004) 1–15.

You might also like