Materials Science & Engineering A 557 (2012) 17–28

Contents lists available at SciVerse ScienceDirect

Materials Science & Engineering A

journal homepage: www.elsevier.com/locate/msea

Thermogravimetric behavior of natural fibers reinforced polymer

composites—An overview
Sergio N. Monteiro a,n, Veronica Calado b, Rubén Jesus S. Rodriguez c, Frederico M. Margem c
a
Military Institute of Engineering, IME, Materials Science Department, Prac- a General Tiburcio, 80, Praia Vermelha, Urca, CEP 22290-270, Rio de Janeiro, RJ, Brazil
b
Escola de Quı́mica, Universidade Federal do Rio de Janeiro, UFRJ, Rio de Janeiro, Brazil
c
State University of the Northern Rio de Janeiro, UENF/LAMAV, Brazil

a r t i c l e i n f o abstract

Available online 29 June 2012 Natural fibers obtained from plants, known as lignocellulosic fibers are environmentally friendly
Keywords: alternatives for synthetic fiber, as polymer composite reinforcement. Applications of natural fiber
Natural fiber composites are expanding in many engineering areas, from civil construction to automobile manufacturing.
Fiber treatment In recent years, a considerable number of scientific and technological works, including review papers, were
Polymer composite dedicated to the characterization and properties of natural fibers and their composites. The mechanical
Thermogravimetry behavior and the fracture characteristics are usually the most investigated and reviewed themes for the
TG/DTG purpose of comparison to corresponding polymer composites reinforced with synthetic fibers, mainly
fiberglass. The thermal behavior is also of practical interest for conditions associated with temperatures
above the ambient, as in fire damage, curing or process involving heating procedures. In fact, several works
also assessed distinct thermal responses, particularly in terms of thermogravimetric properties of natural
fiber polymer composites. As no general review was conducted so far on the thermogravimetric (TG)
behavior of these materials, this article presents an overview limited to temperature effects related to the
loss of mass by means of TG analysis and the related derivative, DTG, for different polymer composites
reinforced with the most common and relevant lignocellulosic fibers.
& 2012 Elsevier B.V. All rights reserved.

1. Introduction compared to glass fiber polymer composites (fiberglass) today

In the past few decades, environmental issues concerning
global scale pollution and climate changes renewed the interest
in natural materials including cellulose-rich fibers extracted from
cultivated plants, also known as lignocellulosic fibers, and their
polymer composites [1–11]. Actually, natural fibers are consid-
ered environmentally friendly not only for their saving in process
energy, which is an unavoidable problem for synthetic fibers, but
also for their renewable and biodegradable characteristics. Both
natural fibers and their polymer composites, if incorporated with
a reasonable amount of fiber, are neutral with respect to CO2
emissions that cause the earth greenhouse effect, a major respon-
sible for global warming [12]. In fact, at the end of these ‘‘green’’
materials life cycle, the release of CO2 due to combustion or
atmospheric degradation will be balanced by the content assimi-
lated during the fiber’s plant biological growth [1].
In addition to environmental, economical and social benefits,
some physical and mechanical properties found in natural fibers
and related composites [1–11] represent important technical
advantages over synthetic fibers. This is of special relevance when
n
Corresponding author. Tel.: þ55 21 2546 7042; fax: þ 55 21 2546 7049.
E-mail address: sergio.neves@ig.com.br (S.N. Monteiro).
extensively used in a wide variety of engineering systems in most
industrial sectors. Fiberglass is heavier, difficult to machine and
cannot be recycled even by incineration in thermoelectric plants.
It also has potential health hazards posed by fiber particulates
[13,14]. An expansion of industrial applications of natural fiber
composites is currently taking place aiming to substitute the
traditional uses of fiberglass in building construction, packaging,
sports, electrical parts, medical prosthesis and automobile com-
ponents [10]. In particular, the automobile industry is increas-
ingly adopting natural composites in numerous interior and
exterior components [15–19]. Among the technical advantages
of natural fibers and their composites, it stands the relatively low
density, which results in improved specific properties of signifi-
cant interest to the automobile industry. For instance, the energy
consumption of 9.6 MJ/kg to produce a flax fiber mat, including
cultivation, harvesting and fiber separation, is significantly lower
than the energy of 54.7 MJ/kg to produce a glass fiber mat [17].
Moreover, for composite processing, a non-abrasive natural
fiber is associated with less damage to tools and molding equip-
ments as well as a relatively better finishing in comparison to
glass fiber. Additionally, owing to an inherent flexibility and easy
delamination with respect to polymer matrices, natural fiber
composites are also tougher and resist impact loads without
shattering [9].

0921-5093/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.msea.2012.05.109
18 S.N. Monteiro et al. / Materials Science & Engineering A 557 (2012) 17–28

Comparatively, important drawbacks are exhibited by natural 250 1C [29]. As a general comment, Nabi Sahed and Jog [2]
fibers that, to some extent, also affect their polymer composites mentioned that thermal degradation of natural fibers also results
[1–11]. As biomaterials, natural fibers have dimensions that are in production of volatiles at processing temperatures above
limited by anatomical restrictions with accentuated statistical 200 1C. This can lead to porous polymer composites with inferior
dispersion. Consequently, all lignocellulosic fibers present a mechanical properties. They stated that the real challenge for the
technically undesirable feature associated with heterogeneous scientists is to improve the thermal stability of these fibers, so
characteristics in contrast to uniform synthetic fibers fabricated that they can be used with engineering polymers for better
within precise dimensional values. Another shortcoming is the performance and thus widen the applications of natural fiber
fact that natural fibers are hydrophilic and tend to develop a weak composites.
bonding with hydrophobic polymers normally used as composite As final remarks, the reader should find relevant to know that
matrices [1–4,9–11]. Surface modifications [20,21] may improve the thermogravimetric behavior of natural fibers apparently bears
the fiber adherence to a polymer matrix and thus increase both a correlation with their chemical constituents. In fact, the TG/DTG
the composite mechanical and thermal resistance [1–4]. A surface curves of common lignocellulosic fibers such as jute, sisal, wood
treatment, however, results in additional cost and decreases the and cotton display similar aspects that could be correlated to the
economical competitiveness of the natural fiber composites. The thermal decomposition of their main constituents. Three stages of
thermal stability of any natural fiber composite may also impose weight loss are associated with the TG curve. A first stage, up to
limitations in applications at temperatures that cause degradation about 200 1C, corresponds to a maximum weight loss of 10% and
of the fiber organic structure. In principle, the temperature not is followed by a second stage with more than 70 wt% of loss, up to
only degrades the structure but also affects most properties of the about 500 1C. The final third stage extends to the usual ending
natural fiber composites. A complete understanding of these test temperature at about 800 1C in association with a loss that
effects requires a review on the composite basic thermogravi- may reach 20 wt%. The corresponding DTG curves display max-
metric, i.e., weight loss with increasing temperature characteriza- imum rate of thermal decomposition peaks as well as shoulder
tion. In spite of many specific works, no overview on thermal and tail peaks that could be ascribed to the fiber’s constituents.
behavior has been presented so far and, therefore, this is a Table 1 exemplifies important TG/DTG parameters obtained from
motivation for this overview article. figures of quoted review papers [7,22]. The values shown in this
table are similar to others reported for distinct natural fibers but
may vary depending on treatments performed on the fiber [7].
2. Thermal degradation of lignocellulosic fibers According to Nguyen et al. [22], the weight loss and DTG peak in
first stage can be attributed to water loss. The thermal degrada-
Among the review articles and books [1–11] on properties and tion of the main lignocellulosic constituents of the fiber begin to
structural characteristics of natural lignocellulosic fibers and their occur at the onset of the second stage. The main DTG peak in the
polymer composites, only that one of Nabi Sahed and Jog [2] second stage can be ascribed to the cellulose decomposition,
presented a section related to general aspects of the thermal while the shoulder peak to the hemicellulose and the tail peak to
stability. Works referred in that section [22–29] served for the the end of lignin decomposition. The weight remaining in the
brief summary on the state of art discussed by these authors [2]. third stage could be assigned to char or other products from
In two conference articles on thermal analysis of lignocellulosic decomposition reactions.
materials, Nguyen et al. [22,23] reviewed the effect of tempera- Another point worth discussing is the influence of different
ture on cellulose, hemicellulose, lignin, and other carbohydrates TGA atmospheres on the thermogravimetric results of natural
as well as different types of wood. It was indicated that the fibers. In principle, two distinct atmospheres, inert (helium and
thermal decomposition of cellulose begins to occur at 210–260 1C nitrogen) and oxidative (air and oxygen) may be used. Moreover,
by dehydration followed by major endothermic reaction of as gases conduct heat at different rates, thermograms obtained in
depolymerization with DTG peaks that vary from 310 to around nitrogen may be significantly different from those obtained in
450 1C. Hemicellulose was found to decompose at a maximum of helium. Under inert atmosphere, the thermal degradation of
290 1C and up to 150 kJ/mol for activation energy while lignin cellulose results in a main DTG peak associated with the forma-
would thermally decompose with peaks from 280 to 520 1C and tion of macromolecules containing rings bearing double bonds
up to 229 kJ/mol for activation energy. [22]. In an oxidative atmosphere there is partial overlapping of
The cellulose, hemicellulose and lignin degradation caused by this peak with the exothermic peak corresponding to the oxygen
temperature was considered by Nabi Sahed and Jog [2] a crucial reaction with the cellulose. As a consequence, the main DTG peak
aspect for the thermal stability of natural fiber reinforced polymer is shifted to lower temperatures in oxidative atmosphere as
composites. In the case of a thermoset matrix, this is a limiting compared to the inert one. For instance, the maximum rate of
factor for choosing the curing temperature, while in thermoplas-
tic matrix composites, a limitation for extrusion temperature.
Thermal protection of the fiber was attempted by grafting of Table 1
acrylonitrile on jute [25], which increased the degradation tem- Thermogravimetric parameters of common natural fibers [7,22].
perature to 280 1C, and on sisal [26], by lowering both the rate of
Natural 1st 1st 2nd 2nd 2nd 2nd 2nd 3rd
degradation as well as the total weight loss. The maximum rate of fiber stage stage stage stage stage stage stage stage
degradation was also reduced in wood flour by previous treat- weight DTG onset weight DTG DTG DTG weight
ment with phosphonate [25]. Works on the effect of thermal loss (%) peak T0 loss (%) shoulder main tail loss (%)
degradation of wood flour incorporated polymer composites (1C) (1C) (1C) peak (1C)
(1C)
[27,28] showed that mechanical properties deteriorate with
increasing temperature. Toughness and bending strength were Jute [7] 8 60 260 89 290 340 470 3
the most affected. This was attributed to changes in the surface Sisal [7] 9 52 250 76 275 345 465 15
chemistry of the wood flour, which might affect the bonding with Wood 2 107 290 85 270 367 400 13
the polymeric matrix and impair the composite properties. In [22]
Cotton 4 55 265 91 280 330 410 5
another work on wood flour incorporated polypropylene compo- [7]
sites, similar loss in properties were reported after extrusion at
 S.N. Monteiro et al. / Materials Science & Engineering A 557 (2012) 17–28
decomposition of wood occurs at 320–330 1C in air and 350–
370 1C in nitrogen [22].
3. Thermal analysis of natural fiber polymer composites
3.1. Preliminary considerations
Thermal analysis may be regarded as the experimental proce-
dures and tests used to evaluate chemical, physical and structural
changes occurring in a material under an imposed change in
temperature. In principle, temperature is a fundamental state
variable that affects most chemical reactions, physical properties
and structural transformations. Thus, as a general concept, any
scientific or technological characterization of a material, in which
temperature is varied as an experimental parameter, could be
considered as thermal analysis. However, this term has long been
limited to specific techniques related to thermogravimetric and
calorimetric effects [22]. It is now widely accepted that the main
techniques associated with a thermal analysis are the difference
in temperature between a sample and a reference (DTA), the loss
of weight measured by thermogravimetry (TG), its derivative
(DTG), and the determination of heat flow by differential scanning
calorimetry (DSC). Other techniques used to calculate the thermal
conductivity, specific heat, and thermal diffusivity are, in many
instances, embodied as thermal analyses. Among these techni-
ques, TG/DTG analysis has been extensively used to characterize
the thermal stability of natural fiber polymer composites. This
will be the main subject of the present overview. It is beyond the
scope and space of this work to cover the other techniques. Even
though the well known and extensively applied dynamic mechan-
ical analysis (DMA) might be understood as a type of thermal
analysis, it will also not be part of this overview.
TG/DTG results are usually displayed as curves of weight loss
variation with temperature and referred as thermograms. As
mentioned, lignocellulosic fibers are sensibly affected by tem-
perature and complete thermal degradation is expected to occur
above 400 1C [22]. Plants are mainly composed of cellulose,
hemicellulose, and lignin that are responsible for their fibers
physical properties [1]. Other volatile or partially stable consti-
tuents, such as pectin, waxes and water soluble substances, may
also exist in lignocellulosic fibers. As the major constituent,
cellulose conditions the physical properties of natural fibers and
has a significant contribution in their thermal degradation. The
linear polymeric chain of the cellulose begins to decompose at
relatively lower temperatures and, owing to a catalytic effect of
naturally existing inorganic ions, may form a higher amount of
char [30,31]. Hemicellulose corresponds to polysaccharides with
different sugar units associated with the cellulose. It has com-
paratively higher degree of chain branching but a much smaller
degree of polymerization than the cellulose. Thermal degradation
of hemicellulose precedes that of cellulose but its effect is
proportionally limited by its content in the fiber. Lignin is a
complex hydrocarbon polymer with both aliphatic and aromatic
constituents [1]. The thermal decomposition of lignin occurs in a
broader range that initiates earlier but extends to higher tem-
peratures than those of hemicellulose and cellulose degradation
[31]. However, its effect is also limited by the smaller content in
the fiber.
3.2. Relevant and common natural fibers
A natural fiber polymer composite is obviously composed of
two main parts, the natural fiber, as reinforcing phase, and the
polymer matrix. The present overview will focus attention on
natural fibers, as their already discussed peculiar characteristics
19
provide special properties to the composite material. This does
not mean that the polymer matrix lacks importance. However, the
reinforcement with synthetic fibers, such as glass, fiber, carbon,
aramid and others, places the polymer composite in an entirely
different category. In other words, the natural fiber is the factor
that differentiates the polymer composite as a more convenient
material in terms of specific environmental, economical, technical
(lighter, smoother, tougher) and social advantages [1–11].
A considerable number of natural fibers have been investigated
as a possible composite reinforcement. It would be a formidable
task to cover all existing works that either directly or indirectly
investigated the thermogravimetric behavior of polymer compo-
sites incorporated with these fibers. In addition to animal fibers
(silk, wool, hair, etc.), the reader may be surprised with the
reinforcement potential and specific interest of some exotic
lignocellulosic fibers, such as abaca, buriti, caroa, sansevieria,
palmyrah, kapok, maize, sponge gourd, artichoke, okra, curaua,
milkweed, sabei, phragmites, sparto, malva, piassava, date palm,
paina, communis, pita floja, fique and many others. Polymer
composites with both animal and the aforementioned less com-
mon lignocellulosic fibers will not be covered in the present
overview. Only composites incorporated with well known natural
fibers, such as jute, hemp, flax, sisal, coir, cotton, kenaf, wood,
pineapple, bamboo, ramie, banana and bagasse that are com-
monly investigated and even industrially applied will now be
overviewed. These relevant fibers serve as sub-titles for revisiting
related works.
3.3. Thermogravimetric analysis of commonly known natural fiber
composites
The growing interest for using natural fibers as reinforcement
of polymer composites has motivated investigations on the
thermal behavior of these materials. The reader will find that
many works here overviewed for the thermogravimetric behavior
of natural fiber composites were also dedicated to the thermal
behavior of the pure fiber encompassing, in some cases, not only
TG/DTG analysis but also DTA, DSC and DMA results. These works
are now revisited but the reader will be spared from the pure
fiber TG/DTG results as well as other techniques applied to
investigate the thermal behavior of natural fiber composites.
3.3.1. Jute fiber composites
Das et al. [32] showed TG/DTG curves obtained at a heating
rate of 10 1C/min in nitrogen for 7 wt% phenol formaldehyde
matrix composites (resin sprayed) reinforced with, supposedly,
93 wt% of both untreated and steam-stabilized jute fibers for 4
(SB-4) and 8 (SB-8) minutes. An initial DTG peak at 65–68 1C,
common to the untreated jute fiber, was attributed to the fiber
loss of moisture. The existing hemicellulose decomposition peak
for the pure fiber at 282 1C is missing in both composites. For the
authors [32], this means that during steam stabilization, hemi-
cellulose may have been modified to some other form. On the
other hand, the cellulose decomposition peaks for both compo-
sites at 348 1C, are slightly higher than that of jute at 345 1C.
Finally, the steam stabilized fiber composites were indicated to
yield almost the same amount of char, which is higher than that
of the untreated fiber.
Dash et al. [33] showed TG/DTG curves of polyester compo-
sites reinforced with 60 wt% jute fibers (sliver) modified by
carding operation, in condition of both untreated and NaCl
bleached treated. The authors [33] indicated the existence of an
initial DTG peak at 59 1C for the untreated and 65 1C for the
bleached fibers because of the loss of moisture. It was concluded
that the bleached fiber is less hydrophilic than the untreated one.
20 S.N. Monteiro et al. / Materials Science & Engineering A 557 (2012) 17–28
A second major peak at 337 1C for the untreated and 328 1C for
the bleached indicated that the untreated fiber/polyester compo-
site is thermally more stable than the bleached fiber/polyester
composite, although the difference is not remarkable.
Ray et al. [34] presented TG/DTG of vinyl ester composites
reinforced with untreated and alkali treated (5% NaOH for
immersion times up to 8 h) jute fibers. Just like the simple
untreated fiber (37 1C), all composites display an initial DTG peak
around 42–58 1C owing to the evaporation of moisture. The
hemicellulose degradation temperature, second shoulder peak,
remained the same in all composites, 300 1C. This is very close to
the corresponding of untreated fiber, 297 1C, but much lower than
that of the neat vinyl ester, 419 1C. By contrast, the cellulose
degradation temperature, third peak, decreased from the
untreated fiber composite, 364 1C, to the 8 h alkali treated fiber
composite, 357 1C. The main degradation peak decreased from
416 to 410 1C, for untreated and 8 h-alkali treated fiber compo-
site, respectively. The authors concluded that the changes occur-
ring in the fiber because of the alkali treatment, such as the fiber
splitting into finer filaments, the increase in the crystallinity of
the fiber and the improved bonding between the fiber and the
resin, had a considerable effect on the thermal degradation
behavior of the composites. The incorporation of alkali-treated
jute fibers reduced the thermal stability of the composite.
Manfredi et al. [35] investigated the thermogravimetric beha-
vior of both unsaturated polyester (UP) and unsaturated polyester
with acrylic acid (Modar) as matrices for untreated jute fibers.
Although not presenting quantitative parameters for the compo-
sites, the authors [35] indicated that the Modar matrix exhibits a
higher thermal resistance than the unsaturated polyester. In fact,
the TG curve of jute reinforced Modar composite is shifted up to
70 1C to higher temperature, as compared to the jute-UP curve.
Both jute composites display much lower thermal stability, by TG
curves, than glass fiber composites with same matrices.
Mohanty et al. [36] presented TG/DTG curves of both untreated
and maleic anhydride grafted polyethylene (MAPE) modified jute
fiber as reinforcement for high density polyethylene (HDPE) com-
posites. They found that the decomposition of virgin HPDE started
at 430 1C and practically finished 515 1C. This was comparatively
higher than that of jute fibers. For the composites, apparently, no
Fig. 1. TG/DTG curves of jute/PP þ 2 wt% MAHg PP composites in nitrogen and air with different fiber contents. Reproduced from [37].
peak below 200 1C could be seen in the DTG curves. The authors
[36] indicated that for the untreated jute composite, an initial DTG
peak at 352 1C was probably because of dehydration from cellulose
unit and thermal cleavage of glycosidic linkage. The second peak at
522 1C was attributed to aromatization, involving dehydration
reactions. Above 515 1C, while the virgin HDPE got completely
decomposed, in the untreated fiber composite a charred residue of
3.2% was left. By contrast, the treated fiber composite displayed an
initial peak of 364 1C and a main decomposition peak at 523 1C,
almost the same as in the untreated jute fiber composite. However,
the charred residue was about 5.0% above 530 1C. These results
ensure, in the opinion of the authors [36], a higher thermal stability
of the composites reinforced with MAPE modified jute fibers.
Doan et al. [37] investigated the thermogravimetric behavior
of polypropylene (PP) matrix composites reinforced with 9.4;
19.8; and 31.3 vol% of jute fiber with or without addition of 2 wt%
of maleic anhydride grafted polypropylene (MAHg PP) modifier.
The TG/DTG analysis was carried out at a heating rate of
10 1C/min in either nitrogen or air. The authors [37] found that,
as compared to pure jute fiber, the thermal degradation of PP is
much more influenced by the atmosphere. In nitrogen, a single
DTG peak at 431 1C was said to be initiated by thermal scissions of
carbon bonds accompanied by transfer of hydrogen. By contrast,
in air the peak is prominently reduced to 299 1C, indicating that
thermal degradation of PP in air is easier and faster than in
nitrogen. Doan et al. [37] stated that the thermal stability of the
composites was found to be higher than those of either neat PP or
pure jute fiber in both nitrogen and air. Fig. 1, reproduced from
[37], shows that the composite thermal resistance decreases with
increasing fiber content in nitrogen, but the reverse occurs in air.
The authors’ [37] explanation was based on the difference in
fraction of PP that decomposes in nitrogen or air.
Yu et al. [38] showed TG curves obtained at a heating rate of
20 1C/min in nitrogen, for polylactic acid (PLA) matrix reinforced
with up to 50 wt% of untreated jute fibers. According to the
authors [38], the thermal degradation of PLA takes place in a
single stage with a peak at 356 1C. Composites show lower
degradation temperatures than that of PLA. No discussion was
offered on possible degradation mechanisms or on the thermal
stability of the composites.
 S.N. Monteiro et al. / Materials Science & Engineering A 557 (2012) 17–28 21

3.3.2. Hemp fiber composites
Sgriccia and Hawley [39] investigated the effectiveness of oven
and microwave curing of diglycidyl ether of bisfenol-A (DGEBA)
epoxy composites reinforced with 15 wt% of several natural fibers
by means of thermogravimetric analysis conducted at a heating
rate of 25 1C/min in nitrogen. Table 2, adapted from the tables
shown in Sgriccia and Hawley [39] paper, presents the TG
degradation temperatures for 5, 25, 50 and 70 wt% weight loss
of fibers and composites. In the particular case of the hemp fiber
and epoxy composites, this table shows that the limit values for
the degradation temperatures for the pure hemp fiber are 105 1C
(5 wt%) and 458 1C (75 wt%). These degradation temperatures
increased, respectively, to 348 1C (5 wt%) and 472 1C (75 wt%)
for the oven cured and further to 352 1C (5 wt%) and 474 1C
(75 wt%) for the microwave cured epoxy composites. As a main
conclusion, the authors [39] indicated that natural fiber compo-
sites have lower degradation temperature as compared to corre-
sponding glass fiber ones, which affects processing of composites
and usage temperatures.
Pracella et al. [40] showed TG curves obtained at a heating rate
of 10 1C/min in nitrogen for polypropylene (PP) as well as for
various compositions of PP/hemp fibers, PP/glycidyl metacrylatte
(GMA) modified hemp fibers, and GMA grafted PP/hemp fiber.
As a summary of their results, the authors [40] indicated that all
composites started degradation earlier than the PP matrix. More-
over, for modified fiber composites, the weight loss at a given
temperature was higher than that of untreated fiber composite. In
the DTG curves, not shown in the work, the maximum degrada-
tion rate was always shifted to higher temperature with respect
to the neat PP and cellulose. Composites with GMA modified PP
matrix resulted in higher thermal stability. Furthermore, the
temperature corresponding to the maximum decomposition rate
was found to be almost independent of hemp fiber content.
Beckermann and Pickering [41] investigated the thermogravi-
metric behavior of polypropylene (PP) matrix composites rein-
forced with up to 40 wt% of both untreated and NaOH/Na2SO3
treated hemp fibers. Results of TG, obtained at a heating rate of
2 1C/min in air, showed that both type of fibers reinforced PP
composites start to lose weight at lower temperatures, 214 1C,
Table 2
Degradation temperatures at different levels of TG weight loss for glass and
selected natural fibers as well as cured epoxy composites. Adapted from [39].
when compared to neat PP, around 240 1C. This indicates that the
addition of hemp fiber reduces the thermal stability of the compo-
site. It can also be seen that there is an apparent improvement in
thermal stability of the untreated fiber composite in comparison to
the alkali treated composite. This is in contrast to what was
expected. However, the authors [41] attributed the maleic anhy-
dride grafted polypropylene (MAPP) coupling agent, used in the
untreated fiber, as able to bond with the hemicellulose. The authors
[41] indicated, quoting [34,36,42], that the poor thermal stability of
hemicellulose and pectin in a composite can be negated by the
inclusion of a coupling agent. As conclusions, the authors [41] stated
that both untreated hemp fiber composites and NaOH/Na2SO3
treated hemp fiber composites, each with a matrix of PPþ 4%
anhydride modified polypropylene (MAPP), were less thermally
stable than the neat PP. Moreover, the thermal stabilities of
composites containing untreated fiber and NaOH/Na2SO3 treated
fiber, were found to be similar to each other.
Moriana et al. [43] presented TG/DTG results on three ligno-
cellulosic fibers: cotton, kenaf and hemp, 10 wt%, reinforced
biocomposites using a commercial starch-based, Mater-Bi KE,
thermoplastic biopolymer as the matrix. Fig. 2 reproduces the
TG/DTG curves of the neat Mater-Bi KE and lignocellulosic fiber
composites. The thermal decomposition of the studied reinforced
biocomposites presents two main weight loss regions, similarly to
the neat matrix. The thermogravimetric parameters, including the
corresponding activation energies, are presented in Table 3,
adapted from tables in the work of Moriana et al. [43].
The authors [43] indicated that, when the pure Mater-Bi KE is
reinforced, the region associated with the synthetic component of
this matrix is not significantly affected. In addition, the peak
owing to starch degradation is completely overlapped with
hemicellulose/pectin as well as with cellulose degradation of
the fiber. This causes a slight displacement to higher tempera-
tures and is an indication of the improved thermal stability of the
biocomposites. As for activation energies of hemp fiber compo-
sites (Table 3), the values for the first, 175–177 kJ/mol, and
second, 219–222 kJ/mol, thermal decomposition processes are
comparatively higher than the corresponding ones, 98–102, and
172–175 kJ/mol for the Mater-Bi KE matrix. The second (main)
thermal decomposition activation energy is also greater than the
corresponding, 201–202 kJ/mol, for the pure hemp fiber. This also
corroborated the fact that the biocomposites have a higher
thermal stability than either the fiber or the matrix.

Weight loss (%) T (1C) at T (1C) at T (1C) at T (1C) at

5 wt% 25 wt% 50 wt% 75 wt% (1)
(2)
Hemp 105 329 341 458
(2)
Kenaf 64 313 341 371 (1) (3) (3)
Henequen 61 301 337 392 (4) (4)
Flax 88 323 339 441
Glass 586 – – –
DTG, (/°C)

Hemp/epoxy/oven 348 384 392 472

Hemp/epoxy/ 352 390 393 474
microwave (1) (4)
Kenaf/epoxy/oven 321 370 390 479
Kenaf/epoxy/ 336 382 392 479
microwave (3)
Henequen/epoxy/ 306 375 391 456 (2)
oven
Henequen/epoxy/ 355 389 392 483
microwave
Flax/epoxy/oven 364 389 393 474
Flax/epoxy/ 351 385 391 472
microwave 112 187 262 337 412 487 562 637
Glass/epoxy/oven 400 N/A 402 539
Glass/epoxy/ 400 N/A 402 482 T(°C)
microwave
Neat epoxy/oven 497 N/A 398 490 Fig. 2. Thermogravimetric analysis of the biocomposites: DTG curves of (1) Mater-
Neat epoxy/oven 403 N/A N/A 484 Bi KE, (2) Mater-Bi KE/cotton, (3) Mater-BI KE/kenaf and (4) Mater-Bi KE/hemp at
the heating rate of 20 K/min. Adapted from [43].
22 S.N. Monteiro et al. / Materials Science & Engineering A 557 (2012) 17–28

Table 3
Thermogravimetric parameters (a) and activation energies (b) for neat matrix as well as cotton, hemp and kenaf biocomposites. Adapted from [43].

(a)
Composite First weight loss First weight loss
Residue (%)
Onset (1C) Peak (1C) Onset (1C) Onset (1C)

Mater-Bi KE (neat) 319 336 413 435 1.2

Cotton/Mater-Bi KE 324 344 415 434 2.5
Hemp/Mater-Bi KE 327 351 414 432 4.0
Kenaf/Mater-Bi KE 334 356 416 433 4.3

(b)
Composite Thermal decomposition process (region) Activation energy (kJ/mol) obtained by different methods

Friedman Flynn-Wall-Ozawa Coats-Redfern-Criado

Mater-Bi KE (neat) First Weight Loss 98 102 101

Second Weight Loss 173 175 172
Cotton/Mater-Bi KE First Weight Loss 112 113 111
Second Weight Loss 295 297 295
Hemp/Mater-Bi KE First Weight Loss 175 175 177
Second Weight Loss 219 222 220
Kenaf/Mater-Bi KE First Weight Loss 163 159 162
Second Weight Loss 200 199 201

3.3.3. Sisal fiber composites
Albano et al. [44] investigated the thermogravimetric behavior
of untreated and acetylated sisal fibers reinforced neat polypro-
pylene, PP, of polypropylene/high density polyethylene blend
(80PP/20 HDPE) matrix composites. The polymer blend was also
modified with 5 wt% of ethylene–propylene copolymer (EPR). TG/
DTG analysis was carried out at a heating rate of 10 1C/min in
nitrogen. All sisal composites (blends with filler) display three
decomposition stages that are attributed to the decomposition of
the different components of the blend, i.e., fiber and polymer. The
authors [44] presented the values of the starting decomposition
temperature (Ti), temperature of the maximum decomposition
rate (Tmax), associated with the main DTG peak, and final decom-
position temperature (Tf) for the sisal fibers and composites
investigated. As relevant examples, for the acetylated sisal fiber,
Ti ¼250 1C; Tmax ¼385 1C; and Tf ¼410 1C. For the PP/HDPE/nf-EPR:
Ti ¼320 1C; Tmax ¼ 445 1C; and Tf ¼470 1C. For the PP/HDPE/nf-EPR
composite reinforced with acetylated fiber: Ti ¼280 1C; Tmax ¼
448 1C; and Tf ¼470 1C. The authors [44] indicated that when
the sisal fiber is added to the matrix, the starting decomposition
temperature, Ti, decreases by 40 1C. However, both the maximum,
Tmax, and the final, Tf, decomposition temperatures are not much
affected. These temperatures represent the long-term stability of
the decomposition process. An extensive study on the activation
energy led Albano et al. [44] to conclude that when acetylated
fiber is mixed with polymers, a greater polymer–filler interaction
takes place, which slightly favors the stability of these composite
materials as compared to blends of non-acetylated fiber.
Nair et al. [45] studied the thermogravimetric of polystyrene
(PS) composites reinforced with short sisal fibers. These fibers
were used in the condition of both untreated as well as treated
with benzoylation, polystyrene maleic anhydride coating and
acetylation. TG analysis was carried out at a heating rate of
20 1C/min. It was reported that the treatments improved the
fiber/matrix adhesion and the PS/sisal composites are thermally
more stable than unreinforced neat PS and pure sisal fiber. In fact,
for neat PS decomposition starts at 288 1C and is finished at
435 1C. In the case of the 20 wt% untreated sisal fiber composite,
the major degradation temperature begins at 32 1C and is almost
complete at 447 1C, which indicates a better thermal stability
than neat PS or pure sisal fiber. Moreover, TG curves show that
the thermal stability of sisal fiber treated composites is higher
than that of the untreated fiber composite. The authors [45]
attributed this higher composite thermal stability not only to the
inherent greater thermal stability of the treated fibers but also to
the improved fiber/matrix interactions, which produce intermo-
lecular boding between the sisal fiber and PS.
Xie et al. [46] investigated the thermogravimetric behavior of
composites consisting of polypropylene (PP) and maleic anhy-
dride grafted styrene–ethylene–co-butylene–styrene copolymer
(MA-SEBS) reinforced with 20 wt% of untreated sisal fibers.
Hybrid PP/MA-SEBS matrices were compounded with 0 wt%
(PP-0); 8 wt% (PP-8); and 16 wt% (PP-16) of MA-SEBS. TG/DTG
analysis was carried out at a heating rate of 10 1C/min in helium.
The neat MA-SEBS presented only one DTG peak at 442 1C while
the composites display two peaks. The first small peak occurred,
for the different MA-SEBS contents, at approximately the same
temperatures: 363 1C (PP-0); 367 1C (PP-8); and 365 1C (PP-16).
The existence of this first peak was an indication to the authors
[46] that the thermal stability of neat MA-SEBS is better than that
for the composites. The second major peak was observed at
slightly higher temperatures, when MA-SEBS is introduced in
the composite. In fact, this second peak occurred at 451 1C (PP-0);
459 1C (PP-8); and 457 1C (PP-16). Xie et al. [46] indicated that the
incorporation of MA-SEBS improves the thermal stability of the
composites, which was attributed to MA-SEBS interactions
between PP and sisal fibers. No discussion was presented on
degradation mechanisms.
Joseph et al. [47] studied the thermal behavior by TG analysis,
carried out at a heating rate of 10 1C/min in inert atmosphere, of
polypropylene (PP) composites reinforced with sisal fibers that
were both untreated as well as treated with urethane derivative
of PP glycol (PPG/TDi) and maleic anhydride modified PP (MAPP)
in order to improve the fiber/matrix interfacial adhesion. The
authors [47] found that the treated fiber composites show super-
ior thermal stability comparing with the untreated fiber compo-
sites as well as the neat PP and the pure sisal fiber. DTG curves
display main decomposition peaks at: 350 1C—pure sisal fiber;
400 1C—neat PP; 476 1C—30 wt% untreated fiber composite;
478 1C—PPG/TDi treated fiber composite; and 785 1C—MAPP
treated fiber composite. The authors [47] indicated that, in
general, the degradation temperature is shifted to a slightly
higher value, in the case of treated fiber composite, than that of
the untreated sisal fiber composite. This was attributed to the
improved fiber/matrix adhesion due to specific chemical mechan-
isms discussed by Joseph et al. [47].
 S.N. Monteiro et al. / Materials Science & Engineering A 557 (2012) 17–28 23

Gañán et al. [48] showed TG/DTG curves for epoxy (DGEBA/ lignocellulosic fiber/epoxy composites, oven or microwave cured,
TETA) composites reinforced with 30 wt% of sisal fibers that were fall between those for the neat epoxy and the fibers.
both untreated or mercerized, silanized, or silanized with pre-
vious mercerization. According to the authors [48], the composite
3.3.5. Coir fiber composites
present a region at 210–350 1C associated with the sisal fiber
Rosa et al. [51] presented TG/DTG curves for starch/ethylene
constituents decomposition. Treated fiber composites show a
vinyl alcohol copolymer (EVOH) matrix composites reinforced
slight increase on thermal stability with respect to that of
with 15 wt% of both untreated as well as washed, alkali and
untreated fiber composite. In fact, a shoulder peak around
bleached treated coir fibers. Glycerol was added as plasticizer
310 1C for the untreated fiber composites is shifted to about
during the composite melting extruded processing. The authors
330 1C for the silanized sisal fiber composite. These results, as
[51] indicated that changes occurring in the coir fibers because of
indicated by the authors, are possibly related to the treatment.
the treatment led to a positive effect on the thermal stability of
Additionally, for composites whose fibers have been subjected to
the composites. For instance, the moisture loss peak at 134 1C for
mercerization, only a single decomposition region, 380–400 1C, is
the untreated fiber composite shifted to 142–144 1C for the
observed owing to the presence of the sisal fiber.
treated coir fiber/starch EVOH composite. Slight increases were
Paiva and Frollini [49] presented results of thermogravimetric
also found in the hemicellulose degradation DTG peak. Although
analyses conducted at a heating rate of 10 1C/min in air of both
not mentioned by the authors, DTG peaks around 450 1C could
phenolic and lignophenolic matrices composites reinforced with
probably be associated with the polymeric matrix degradation.
70 vol% of sisal fibers that were unmodified as well as modified by
From the main DTG peak results, it is possible to conclude that
mercerization (10% NaOH for 1 h); esterification (merceriza-
coir fiber treated composites present improving thermal stability.
tionþsuccinic anhydride in xylene); and ionization (7.5 kV elec-
Santafé Jr. et al. [52] characterized the thermogravimetric
tric discharge in air). From a list of weight losses at certain
behavior of untreated coir fiber incorporated in different compo-
temperature levels (no TG curves given), the authors [49] indi-
sition (up to 30 vol%) into polyester matrix composites. TG/DTG
cated that the composites present lower thermal stability than
curves were obtained at a heating rate of 10 1C/min in nitrogen.
their respective thermosets, as the sisal fibers decompose at lower
Fig. 3, reproduced from their work [52], shows the composites as
temperatures. Events responsible for the different levels of weight
well as the neat polyester TG/DTG curves. Apparently, the neat
loss were discussed, and the authors proposed that, over a certain
polyester displays a major DTG peak at 385 1C and a tail peak
temperature, the polymer suffers continuous weight losses as a
around 415 1C. The authors [52] attributed these polyester ther-
consequence of these successive processes. In particular, Paiva
mal degradation events to the evolution of volatile compounds
and Frollini [49] state that the lignophenolic composites show at
followed by depolymerization of macromolecular chains, quoting
400 1C a larger weight loss than the phenolic sisal fiber compo-
Mark et al. [53]. Similar curves are observed in Fig. 3 for the coir
sites, probably because of the fragmentation of lignin interunit
fiber/polyester composites with two additional DTG peaks. An
bonds.
initial peak around 57 1C was assigned to the release of absorbed
Kim and Netravali [50] investigated the thermal degradation
humidity, mainly from the coir fiber surface as in the other
profiles of sisal fiber reinforced composites with soy protein
overviewed pure lignocellulosic fibers and related composites.
concentrate modified by blending with gelatin (SG) resin matrix.
Two shoulder peaks around 285 and 348 1C were clearly detected,
Actually, two types of matrices SG-0 (no gelatin) and SG-20 (with
as shown in Fig. 3 for the 30% coir fiber composite. The authors
20% gelatin) were used, both containing 20% of sorbitol. The
attributed these shoulder peaks respectively to the degradation of
thermogravimetric results were obtained at a heating rate of
hemicellulose and cellulose of the coir fiber.
20 1C/min in nitrogen. The thermal degradation profiles of the
neat matrices, SG-0 and SG-20, were almost identical and show
three weight loss steps associated with DTG peaks. The first peak 3.3.6. Cotton fiber composites
around 110 1C is due to water evaporation. The second, a shoulder Silva et al. [54] presented TG/DTG curves obtained at a heating
peak around 200 1C, is attributed to the degradation of sorbitol, rate of 10 1C/min in air for phenolic thermoset (PT) matrix
which was added as a plasticizer. The third broad main peak, composites reinforced with up to 70 wt% of cotton textile fibers.
around 280 1C, is assigned to the degradation of soy protein and The composites were prepared with randomly oriented fibers in a
gelatin. The authors [49] mentioned that, although the thermal mold but cured at higher pressure. The DTG curve for the neat PT
stability of SG-0 and SG-20 resins was identical, the composite displays two peaks at 540 and 700 1C that were not interpreted by
with SG-20 matrix displays a major DTG peak around 313 1C, a
higher temperature than that of the SG-0 composite, around
304 1C. For Kim and Netravali [50], this result, together with a
final char residue at 600 1C for the SG-20 composite, indicates an
improvement in thermal stability owing to the better interaction TG curves
between the sisal fiber and the SG-20 matrix.

3.3.4. Flax fiber composites DTG curves

Sgriccia and Hawley [39] evaluated the effectiveness of oven
and microwave curing of epoxy composites reinforced with
several lignocellulosic fibers. The degradation temperatures from
TG weight losses of 5–75 wt% were determined. The hemp/epoxy
composite was already evaluated in Section 3.3.2. In the case of
the flax/epoxy composite, Table 2 shows that the limit values for the
degradation temperature (weight loss %) for the pure flax fiber, 88 1C
(5 wt%) and 441 1C (75 wt%) increased to 351 1C (5 wt%) and 472 1C
(75 wt%) for the microwave cured epoxy composite. The authors Fig. 3. TG/DTG curves for neat polyester and coir fiber composites. Reproduced
[39] indicated that the degradation temperatures of the evaluated from [52].
24 S.N. Monteiro et al. / Materials Science & Engineering A 557 (2012) 17–28

the authors. The DTG curve for the 50% cotton fiber composite polypropylene (MAPP) matrices. Thermogravimetric results
also shows two clear peaks. The first at 349 1C was attributed to obtained at a heating rate of 10 1C/min in nitrogen were pre-
the decomposition of cellulose. The main second peak at 445 1C sented in terms of the onset and peak temperatures, although the
was assigned to events related to the matrix. In fact, the TG/DTG authors did not show TG curves and characteristics peaks. For
curves of the neat PT reveal a beginning of thermal decomposition instance, the 20 wt% untreated fiber/PP composite is associated
around 350 1C. Although not mentioned by Silva et al. [54], a faint with the onset range of 307–453 1C and a peak at 476 1C, while
peak at about 50 1C in the DTG curve of the composite and not the 20 wt% A1100 silane treated and subjected to propylene
seen in the neat PT, might be ascribed to the release of moisture polymerization wood fiber/PP composite has onset range of
from the cotton fiber. The authors [54] indicate that all cotton 325–447 1C and peak around 471 1C. By comparing these results
fiber composites have lower thermal stability than the neat to the neat PP, with onset at 459 1C and peak around 473 1C, also
phenolic thermoset. presented in the work of Coutinho et al. [57], the reader sees that
only minor changes occurred with wood fiber addition to the
investigated composites. Nevertheless, the authors stated that
3.3.7. Kenaf fiber composites
thermal parameters could be influenced by the presence of wood
In an already presented work (see Section 3.3.2 and 3.3.4),
fibers.
Sgriccia and Hawley [39] studied the degradation temperature,
Doh et al. [58] investigated the thermogravimetric behavior of
associated with TG weight losses of 5–75 wt%, of oven and
wood fiber in the form of liquefied wood (LW) added in different
microwave cured epoxy composites reinforced with hemp, flax,
amounts, up to 40 wt%, into three distinct polymeric matrices: low
kenaf and henequen fibers. In the particular case of kenaf/epoxy
density (LDPE) and high density polyethylene (HDPE) as well as
composite, Table 2 shows that the values for the degradation
polypropylene (PP). TG curves were obtained at a heating rate of
temperature (weight loss %) of the pure kenaf fiber, 64 1C (5 wt%)
5–50 1C/min in nitrogen. According to the authors [58], the addition
and 392 1C (75 wt%), increased to 336 1C (5 wt%) and 479 1C
of 10 wt% LW into LDPE, HDPE and PP showed no significant effect
(75 wt%) for the microwave cured epoxy composite. The authors
on the polymer matrix thermal decomposition. However, as the LW
[39] mentioned that the degradation temperature for both micro-
amount increased, the thermal stability of the investigated polymer
wave and convection oven cured fiber composites investigated,
composites decreased. This behavior was attributed to the decrease
including kenaf/epoxy, fall between the degradation temperature
in the compatibility and interfacial bonding between LW/polymer in
of the epoxy matrix and that of the pure kenaf fiber.
the composite. After thermal degradation of the composites above
Julkapli and Akil [55] performed thermogravimetric analysis on
500 1C, the tar and ash content increased with the amount of added
chitosan matrix biocomposites reinforced with up to 28 wt% of kenaf
LW. As relevant examples of the thermogravimetric parameters
fiber. The ground kenaf fiber, 30–40 mm mesh size, was dispersed
obtained by Doh et al. [58], the decomposition temperature was
together with the chitosan powder in acetic acid solution, which
determined at the starting point of severe weight loss along with the
renders the molecules more reactive with the presence of NH3þ . Thus,
activation energy for, HDPE with 472 1C, 240 kJ/mol; and PP with
compatibility between kenaf fiber and chitosan matrix is expected to
450 1C, 75 kJ/mol. These values change with the addition of LW into
improve. TG/DTG curves were obtained at a heating rate of 20 1C/min
the polymers: LW/HDPE with 474 1C, 223 kJ/mol; and LW/PP with
in nitrogen. The authors [55] reported two DTG peaks in all curves for
452 1C, 55 kJ/mol. As main conclusions, the authors [58] indicate
both neat chitosan and kenaf fiber composites. The first degradation
that the thermal stability of LW composites decreased with addition
step, 25–140 1C, with an apparent DTG peak around 120 1C was
of LW. Moreover, higher heating rates provided better thermal
attributed to evaporation of water or volatile compounds, like
stability resulting from the decelerated decomposition rate.
solvent. The second main peak around 280 1C and tail peaks around
Bhardwaj et al. [59] conducted thermogravimetric analysis
360 and 500 1C were generally attributed to the major degradation of
at a heating rate of 20 1C/min, atmosphere not indicated, of a
the polymers including the pyrolysis of chitosan and decomposition
polyhydroxybutyrate-co-hydroxyvalerate (PHBV) as well as
of kenaf fiber. Julkapli and Akil [55] concluded that the addition of
polypropylene (PP) matrices composites reinforced with 40 wt%
kenaf fiber (dust) in chitosan matrix (film) does not give any
of recycled (wood) cellulose fiber (RCF) from selected news-
significant change in thermal stability of the biocomposite, except
papers, magazines or Kraft paper stock. The authors [59] indicated
that the TG/DTG decomposition curves were recorded in two stages.
that the onset of thermal degradation in the PHBV-based compo-
For the reader, this exception is rather surprising as a conclusion. In a
site was comparable to that of PP-based composite. However, the
review of kenaf fiber reinforced composites, Akil et al. [56] confirmed
degradation rate of the former was more drastic after 250 1C,
the conclusion expressed in the afore-mentioned work of Julkapli and
which was attributed to the degradation (lower thermal stability)
Akil [55]. The authors [56] concluded on the potential of kenaf fiber
of both the fiber in RCF and the PHBV.
as an alternative to replace conventional material or synthetic fibers
Lei et al. [60] studied the thermal stability of 30 wt% bagasse or
as composites reinforcement.
pine wood fiber reinforcing recycled high density polyethylene
Although the basic results and discussion of the work of
(RHDPE) by TG analysis, at a heating rate of 10 1C/min in nitrogen.
Moriana et al. [43] have been presented in Section 3.3.2, the
In the case of wood fiber composites, two active coupling agents
reader may find worthwhile a comment on the results of kenaf
were used: a maleated polyethylene (MAPE) and a titanium-
fiber/ Mater-Bi KE biocomposite. As shown in Fig. 2 and Table 3,
derived mixture (TDM) of chemical agents. It was reported that
the first DTG peak of this biocomposite at 356 1C is sensibly
the onset degradation for the neat RHDPE occurs at 441 1C and the
higher than that, 336 1C, of the neat Mater-Bi KE. Furthermore, as
maximum decomposition rate at about 471 1C. The introduction
presented in Table 3, the activation energy 162–163 kJ/mol for
of 30 wt% wood fiber generates another intermediate decomposi-
this peak in the composites is significantly higher than the
tion peak. These three peaks, onset, stage I and stage II, were
corresponding 99–101 kJ/mol of the matrix. This indicates a
respectively found as: pure fiber/RHDPE with 262, 353 and
comparatively higher thermal stability.
469 1C; 1.2% MAPE coupled fiber/RHDPE with 263, 353
and 469 1C; and 0.9% TDM coupled fiber/RHDPE with 260, 349
3.3.8. Wood fiber composites and 468 1C. As indicated by the authors [60], the coupling agents
Coutinho et al. [57] studied the thermal behavior of both seem to have little influence on the thermal degradation of the
untreated and treated (silane and methanol) aspen wood fibers composites. The reader could also conclude that the addition of
reinforced composites with polypropylene (PP) or maleated pine wood fiber to the RHDPE significantly decreased the onset
 S.N. Monteiro et al. / Materials Science & Engineering A 557 (2012) 17–28
degradation temperature and thus the beginning of the composite
thermal stability.
3.3.9. Pineapple fiber composites
George et al. [61] investigated the thermal behavior of both
untreated and treated pineapple leaf fiber (PALF) reinforced
polyethylene composites by thermogravimetric analysis. The
PALF treatment was carried out using isocyanate, silane and
hydrogen peroxide to improve the fiber/matrix interfacial adhe-
sion. The authors [61] indicated that at the high temperatures the
PALF degrades before the polyethylene matrix. Moreover, the
treated PALF composites impart better properties as compared to
untreated PALF polyethylene composites.
Luo and Netravali [62] presented thermogravimetric results of
poly (hydroxybutyrate-co-valerate) (PHBV) matrix biocomposite
reinforced with up to 28 wt% of pineapple leaf fiber. TG curves
obtained at a heating rate of 10 1C/min in both nitrogen and air
confirmed that the PHBV easily undergo thermal degradation.
Moreover, the thermal decomposition of PHBV is very similar,
regardless the atmosphere. As indicated by the authors [62], the
main decomposition temperatures were observed at 270 1C and
273 1C, in air and nitrogen, respectively, with a final thermal
degradation residue, mostly composed of crotonic acid. The
‘‘green’’ biocomposites display, not only the main decomposition
temperatures associated with the PHBV matrix, at 268 1C in air,
and 273 1C in nitrogen, but also practically the same main
decomposition temperatures, as a second peak, associated with
the pineapple fiber at 327 1C in air and 334 1C in nitrogen. It was
concluded that the presence of pineapple fibers did not affect the
thermal decomposition of PHBV.
Threepopnatkul et al. [63] showed TG curves (no conditions
given) for pineapple leaf fiber (PALF) as well as polycarbonate (PC)
and related composites reinforced with up to 20 wt% of both
alkali and silane treated PALF. It was commented, without
quantitative parameters, that the PALF has the lowest, while the
PC the highest thermal stability with all composites lying in
between. The authors [63] indicated that the PALF composites
showed the onset for thermal degradation at lower temperature,
approximately 270 1C. It was then interpreted that composites
have lower stability than the neat PC because of the low thermal
stability of the PALF.
3.3.10. Bamboo fiber composites
Lee and Wang [64] investigated the thermogravimetric beha-
vior of 30 wt% bamboo fibers (BF) reinforcing two distinct
biocomposites with polylactic acid (PLA) and polybutylene succi-
nate (PBS) as matrices. TG/DTG analyses were performed at a
heating rate of 10 1C/min in nitrogen. The interfacial adhesion of
both composites was improved by the addition of different
amounts of lysine-based diisocyanate (LDI) as a coupling agent.
TG/DTG curves show in either case neat PLA or neat PBS, the onset
of thermal degradation and the main (single) DTG peak, 376 1C for
PLA and 405 1C for PBS, occurring at higher temperatures than
those for pure BF and related composites. In the case of BF/PLA, a
two-stage loss of weight was observed with degradation in the
range of 280–340 1C attributed to PLA and a small shoulder peak
350 1C, owing to fiber degradation. Thermal degradation was
increased with LDI content. The authors [64] indicated that, in
general, the increase in the molecular weight by cross-linking
reactions between matrix and fiber, or molecular chain extension
of the matrix itself, could increase the thermal stability. In the
case of BF/PBS composites, an intermediate thermal stability was
verified between those of BF and PBS. The addition of LDI also
improved the thermal stability of the composite.
25
Shih [65] showed TG curves obtained at a heating rate of 10 1C/
min in nitrogen for composites of both powder and fiber of water
bamboo husk incorporated, in formulation up to 10 wt%, into
epoxy matrix. The bamboo fibers were either untreated or treated
with distinct silane coupling agents. Thermogravimetric para-
meters were not presented by Shih [65], only mentioning that the
char yields of the composites were larger than that of the epoxy.
This reveals that the addition of bamboo powder or fiber into
epoxy would effectively raises the char yield and thus limits the
production of combustible gases, decreasing the exothermicity of
the pyrolysis reaction. This, therefore, inhibits the thermal con-
ductivity of the burning material.
Singh et al. [66] showed TG/DTG curves obtained at a heating
rate of 20 1C/min in nitrogen for pure bamboo fiber as well as 30%
bamboo fiber reinforcing poly(hydroxybutyrate-co-valerate)
(PHBV) biocomposite. The DTG curve for the neat PHBV presents
a major, single, peak, at 316 1C, while the composite has its main
peak around 320 1C followed by a small peak at about 375 1C,
practically coinciding with the main pure bamboo fiber peak.
According to the authors [66], the thermal degradation of the
biocomposite appears to be a cumulative phenomenon of matrix
and fiber alone. The maximum degradation temperature peak of
the composite shifted to a higher temperature from that of PHBV
indicating the composite to be slightly more thermally stable than
the matrix. Thus, the presence of bamboo fiber did not have any
practical degradation effect on PHBV.
3.3.11. Ramie fiber composites
In the already reviewed (see Section 3.3.1) work of Yu et al.
[38], in addition to jute fiber/polylactic acid (PLA) composite, a TG
curve of PLA reinforced with 30 wt% ramie fibers was also
presented. From this curve, one may infer that the composite
onset thermal degradation temperature, around 220 1C, is sig-
nificantly lower than that for the neat PLA, at about 310 1C. The
authors [38] indicate that the thermal degradation of PLA occurs
completely in a single stage at 356 1C, while the ramie fiber
composite shows a lower degradation temperature. It was also
mentioned that this might be a consequence of the decrease of
relative molecular mass of PLA. Moreover, Yu et al. [38] stated
that when the composite is mixed by the two rolls operation, the
incorporation of ramie fiber into PLA matrix has also an effect on
the thermal degradation temperature.
3.3.12. Banana fiber composites
Zainudin et al. [67] showed TG/DTG results obtained at a
heating rate of 10 1C/min in nitrogen for banana pseudo-stem
(BPS), in amounts up to 40 wt%, reinforcing unplastisized poly-
vinyl chloride (UPVC) composites in both unmodified and acrylic
modified condition. For the reader information, the authors [67]
always referred to the BPS as a composite filler not a fiber. The
physical aspect of the BPS was not revealed, according to the
authors, because of the confidential nature of these data. The TG/
DTG curves for the neat UPVC display a main DTG peak at
apparently 268 1C, but not mentioned, combined with a shoulder
around 300 1C. Another smaller peak was reported at 435 1C. The
authors attributed the main peak to the dehydrochlorination of
the UPVC with succeeding formation of double bonds and release
of HCl as the main volatile product of degradation, quoting
Semsarzadeh et al. [68]. The 435 1C peak was assigned to the
separation of polyene structure sequence formed during the first
main degradation stage. For all composites, an initial DTG peak
occurred around 63 1C, owing to the volatilization of moisture,
which was correlated to loss of water in BPS. Zainudin et al. [67]
indicated that the addition of acrylic increased the thermal
stability of the composite. In fact, from the DTG curves, one can
26 S.N. Monteiro et al. / Materials Science & Engineering A 557 (2012) 17–28
see that the main peak temperatures (and amount of BPS) for the
acrylic modified at 259 1C (40 wt%) and 293 1C (10 wt%) are
sensibly higher than the corresponding for the unmodified com-
posites, respectively 256 1C (40 wt%) and 281 1C (10 wt%). The
authors [67] also concluded that the thermal stability of
the composites decreased with increasing BPS loading because
the thermal stability of the filler (fiber) is much lower than that of
the matrix.
3.3.13. Bagasse fiber composites
Mothé et al. [69] investigated the thermogravimetric behavior
of polyurethane (PU) matrix composites reinforced with up to
20 wt% of untreated sugarcane bagasse fibers. TG/DTG analysis
was carried out at a heating rate of 10 1C/min in nitrogen. DTG
curves for the pure bagasse fiber as well as for composites with
different (5, 10 and 20 wt%) fiber contents show small initial
peaks at a common temperature around 40 1C. These peaks are
probably due to fiber moisture release. Three other peaks could
also be seen. A clear shoulder peak for the pure bagasse around
310 1C shifts to a faint shoulder peak at about 280 1C for the
20 wt% fiber composite. Although not mentioned, these shoulder
peaks could be a result of the hemicellulose degradation. Other
clear peaks at about 355–370 1C might be caused by cellulose
degradation in the bagasse fiber. A major peak around 420 1C for
all composites, but not seen in the pure bagasse fiber, might be
attributed to PU degradation. As the main objective of Mothé et al.
[69] was to investigate the thermal decomposition kinetics, the
free interpretation of degradation mechanisms was an additional
contribution to the reader.
In the already reviewed (see Section 3.3.8) work of Lei et al.
[60], in addition to pine wood fiber/RHDPE composites, TG curves
were also presented for RHDPE matrix composites reinforced with
30 wt% of sugarcane bagasse fiber. DTG measurements were
carried out at a heating rate of 10 1C/min in nitrogen. Three active
coupling agents were applied to the fiber: a maleated polyethy-
lene (MAPE), a carboxylated polyethylene (CAPE) and a titanium
100µm
100µm
Fig. 4. Impact fracture surface of bagasse fiber (30 wt%)/RHDPE composites containing: (a) no coupling agent; (b) 1.5% MAPE; (c) 1.5% CAPE; and (d) 0.9% TDM. Reproduced
from [60].
derived mixture (TDM). The TG curves revealed that the pure
bagasse has a significantly lower onset degradation temperature,
248 1C, than the RHDPE, 441 1C. With the incorporation of 30 wt%
of bagasse fiber, this onset temperature becomes 254–256 1C.
In addition to the onset degradation temperature, the authors [60]
reported two more weight loss stages for the composites. The first
with fastest decomposition temperature (TdI) around 341–342 1C,
was close to the only corresponding stage for the pure bagasse at
327 1C. The second (TdII) around 469–470 1C, was very close to the
only corresponding stage for the neat RHDPE at 470 1C. Based on
these results, Lei et al. [60] concluded that the first stage of the
composite thermal degradation, at much lower temperature, was
related to the bagasse fiber influence, while the second stage to the
RHDPE influence. It was also concluded that the applied coupling
agents had little effect on the thermal stability of the composites,
although SEM fractographs in Fig. 4, reproduced from [60], show
that the gap (white arrows) between bagasse fiber and matrix is
reduced when the coupling agents were added.
3.4. Remarks and conclusions
More than an extensive list of results acknowledging the effort
of researchers, this overview aims to bring into attention relevant
advances in the comprehension of thermogravimetric behavior of
natural fiber composites. As a first remark, the reader should keep
in mind that, due to space limitation, a relatively small number
(13) of the most commonly known lignocellulosic fibers was
selected. Moreover, the published works, at least in easily avail-
able and well recognized sources, are also restricted to limited
types of polymers as composites matrices. Therefore, a rather
fragmented picture exists regarding both fibers and polymers.
Consequently, the reader should expect uncertainties not only in
a complete view but also in reliable conclusions on the thermo-
gravimetric stability of natural fiber composites. In spite of these
restrictions, some general trends appear to hold and are worth
discussing.
100µm
100µm
 S.N. Monteiro et al. / Materials Science & Engineering A 557 (2012) 17–28
3.4.1. Initial water loss
Normally, an initial weight loss (o10 wt%) is observed in
natural fiber polymer composites below 200 1C. This is usually
associated with a relatively small DTG peak assigned to evapora-
tion of water from the fiber surface. The peak indicates a
maximum rate of released water (moisture, humidity). Common
synthetic polymers give only a minor contribution, if any, to this
low temperature water loss. However, biocomposites with natu-
rally containing water matrices may also provide a significant
amount to this initial DTG peak. In several works [43,48,54,66,69],
this initial peak has been overlooked or just ignored. Whenever
relevant, its existence in this overview was inferred from pre-
sented DTG curves. Although in relatively small amount, the
water loss from inside a composite may cause porosity and
impair the properties. Actually, its DTG peak temperature may
be considered as a first stability limit for the composite. In
practice, even if a natural fiber undergoes oven-drying before
incorporation into a polymer composite, the total elimination of
water is not possible because of the hydrophilic characteristic of
the fiber. Fiber treatments such as alkalinization [34,49] and
bleaching [33] that partially extract the highly hydrophilic hemi-
cellulose, considered as greatest responsible for water absorption
[49], tend to increase the corresponding DTG peak temperature
and reduce weight loss. In the present overview, water loss peaks
were found at temperatures as low as 37 1C [34] and inferred at
about 140–144 1C [48,51].
3.4.2. Onset of degradation
From a practical standpoint, the thermal stability of a natural
fiber composite is related to the onset of a massive weight loss.
This is clearly observed as a sharp downward inclination in the TG
curve and has been reported in terms of an onset temperature
(T0). The determination of T0, however, can be done by distinct
methods, based on transition from an initial baseline [57],
tangents intercepts [58] or a percentage of weight loss deviation
from baseline [67]. A more visually reliable method is to consider
the first decomposition DTG peak, either single or as a shoulder.
The works overviewed have shown that this first DTG peak
occurred in a range from 250 1C [50] to 365 1C [46], depending
on the fiber and polymer matrix. Apparently, the onset of
thermogravimetric degradation in a natural fiber composite is a
complex process, which depends not only on the comparative
thermal stability of fiber vs. matrix but also on the experiment
atmosphere. In petroleum-based polymers that are traditionally
used as engineering materials and possess a high degree of
biodegradability resistance (non-degradable), such as high den-
sity polyethylene [36], epoxy [48], polyester [52], and polyur-
ethane [69], a single DTG peak under nitrogen is observed at
relatively higher temperatures: 515, 380, 385, and 420 1C, respec-
tively. Comparative decrease may occur in air [37]. The introduc-
tion of a lignocellulosic fiber in these polymers normally reduces
T0 and in particular the first DTG peak to 352, 310, 285 and 280 1C,
respectively. As an exception, the reverse was reported by Doan
et al. [37] in jute/polypropylene. It is accepted [41,52] that earlier
decomposition of hemicellulose, known to have a maximum rate
around 250–290 1C [22,70], is probably responsible for these
lower T0 and first DTG peaks of thermal degradation in the
composite as compared to neat polymer matrix. Decomposition
of other fiber components, such as waxes, glycosidics, and pectin,
is also taking place at this lower temperatures, including the
beginning of lignin degradation at about 110–220 1C [22,70]. By
contrast, environmentally friendly biodegradable polymers, such
as starch [43], soy protein [50], chitosan [55], and PHBV [66], with
more than one DTG decomposition peak, show the first at 336,
270, 300, and 316 1C, respectively. Introduction of lignocellulosic
27
fiber results in composites with very close or higher first DTG
peak: 356, 304, 300, and 320 1C, respectively. In this case, the
biocomposite thermal stability is limited by the matrix. From all
these results, the reader may conclude that applications of natural
fiber composites are, in general, thermally restricted to a safe
temperature of 250 1C, or to a maximum of 365 1C in case more
stable specific composites, such as sisal/polypropylene [46] are
selected.
3.4.3. Higher temperatures degradation
In addition to a first DTG peak, most natural fiber polymer
composites display other higher temperature peaks, the second
usually with higher intensity. In common non-degradable polymer
composites, this second main DTG peak is related to matrix
degradation [33,44,52,60], while biocomposites depict second peaks
associated with cellulose/lignin decomposition [50,55,62,64,66].
Other peaks at even higher temperatures are attributed to thermal
degradation of substances resulting from earlier decomposition
stages. For engineering applications, these higher temperature peaks
are not as important as the first. However, they may be used to
calculate the activation energy (Table 3) [43], which is another
parameter associated with thermal stability [58,70].
Acknowledgments
The authors thank the financial support of the Brazilian
agencies: CNPq, CAPES and FAPERJ.
References
[1] A.K. Bledzki, J. Gassan, Prog. Polym. Sci. 4 (1999) 221–274.
[2] D. Nabi Sahed, J.P. Jog, Adv. Polym. Technol. 18 (1999) 221–274.
[3] A.K. Mohanty, M. Misra, G. Hinrichsen, Macromol. Mater. Eng. 276/277
(2000) 1–24.
[4] A.K. Mohanty, M. Misra, L.T. Drzal, J. Polym. Environ. 10 (2002) 19–26.
[5] A.N. Netravali, S. Chabba, Mater. Today 6 (2003) 22–29.
[6] A.K. Mohanty, M. Mishra, L.T. Drzal (Eds.), Natural Fibers, Biopolymers and
Biocomposites, Serial Publications, CRC Press, Boca Raton, USA, 2006 , 2005
and Taylor and Francis, London,.
[7] K.G. Satyanarayana, J.L. Guimara~ es, F. Wypych, Compos. Part A 38 (2007)
1694–1709.
[8] J. Crocker, Mater. Technol. 2–3 (2008) 174–178.
[9] S.N. Monteiro, F.P.D. Lopes, A.S. Ferreira, D.C.O. Nascimento, JOM 61 (2009)
17–22.
[10] S. Kalia, B.S. Kaith, I. Kaurs (Eds.), first ed,Springer, New York, 2011.
[11] S.N. Monteiro, F.P.D. Lopes, A.P. Barbosa, A.B. Bevitori, I.L.A. Silva, L.L. Costa,
Metal. Mater. Trans. A 42 (2011) 2963–2974.
[12] A. Gore, An Inconvenient Truth. The Planetary Emergency of Global Warming
and What We Can do About it, first ed, Rodale Press, Emmanaus, 2006.
[13] P. Wambua, I. Ivens, I. Verpoest, Compos. Sci. Technol. 63 (2003) 1259–1264.
[14] S.V. Joshi, L.T. Drzal, A.K. Mohanty, S. Arora, Compos. Part A 35 (2004)
371–376.
[15] H. Larbig, H. Scherzer, B. Dahlke, R. Poltrock, J. Cell. Plast. 34 (1998) 361–379.
[16] G. Marsh, Mater. Today 6 (2003) 36–43.
[17] J. Holbery, D. Houston, JOM 58 (2006) 80–86.
[18] R. Zah, R. Hischier, A.L. Lea~ o, I. Brown, J. Clean. Prod. 15 (2007) 1032–1040.
[19] C. Alves, P.M.C. Ferra~ o, A.J. Silva, L.G. Reis, M. Freitas, L.B. Rodrigues, J. Clean.
Prod. 18 (2010) 313–327.
[20] J. George, M.S. Sreekala, S. Thomas, Polym. Eng. Sci. 41 (2001) 1471–1485.
[21] S. Kalia, B.S. Kaith, I. Kaurs, Polym. Eng. Sci. 49 (2009) 1253–1272.
[22] T. Nguyen, E. Zavarin, E.M. Barral, J. Macromol., Sci.—Rev. Macromol. Chem. C
20 (1981) 1–65.
[23] T. Nguyen, E. Zavarin, E.M. Barral, J. Macromol. Sci.—Rev. Macromol. Chem. C
21 (1981) 1–60.
[24] A.K. Mohanty, S. Patnaik, B.C. Singh, J. Appl. Polym. Sci. 37 (1989) 1171–1181.
[25] M.W. Sabaa, Polym. Degrad. Stabil. 32 (1991) 209–217.
[26] M.G.S. Yap, Y.T. Que, L.H.L. Chia, H.S.O. Chan, J. Appl. Polym. Sci. 43 (1991)
2057–2065.
[27] C. Gonzalez, G.E. Myers, Int. J. Polym. Mater. 23 (1993) 67–85.
[28] G.E. Myers, I.S. Chahyadi, C. Gonzalez, C.A. Coberly, D.S. Ermer, Int. J. Polym.
Mater. 15 (1991) 171–186.
[29] S. Takase, N.J. Shiraishi, J. Appl. Polym. Sci. 37 (1989) 645–659.
[30] C. Di Blasi, C. Branca, G. D’Errico, Thermochim. Acta 364 (2000) 133–142.
[31] E. Mészáros, E. Jakab, G. Várhegyi, J. Anal., Appl. Pyrol. 79 (2007) 61–70.
28 S.N. Monteiro et al. / Materials Science & Engineering A 557 (2012) 17–28
[32] S. Das, A.K. Saha, P.H. Chourdhury, R.K. Basak, B.C. Mitra, T. Todd, S. Lang,
R.M. Rowell, J. Appl. Polym. Sci. 76 (2000) 1652–1661.
[33] B.N. Dash, A.K. Rana, H.K. Mishra, S.K. Nayak, S.S. Tripathy, J. Appl. Polym. Sci.
78 (2000) 1671–1679.
[34] D. Ray, B.K. Sarkar, R.K. Basak, A.K. Rana, Appl. Polym. Sci. 94 (2004) 123–129.
[35] L.B. Manfredi, E.S. Rodriguez, M. Wladyka-Przybylak, A. Vazquez, Polym.
Degrad. Stabil. 91 (2006) 255–261.
[36] S. Mohanty, S.K. Verma, S.K. Nayak, Compos. Sci. Technol. 66 (2006) 538–547.
[37] T.-T.-L. Doan, H. Brodowsky, E. Mäder, Compos. Sci. Technol. 67 (2007)
2707–2714.
[38] T. Yu, Y. Li, J. Ren, Trans. Nonferrous Met. Soc. China 19 (2009) s651–s655.
[39] N. Sgriccia, M.C. Hawley, Compos. Sci. Technol. 67 (2007) 1986–1991.
[40] M. Pracella, D. Chionna, I. Anguillesi, Z. Kulinski, E. Piorkowska, Compos. Sci.
Technol. 66 (2006) 2218–2230.
[41] G.W. Beckermann, K.L. Pickering, Composites A 39 (2008) 979–988.
[42] P. Gañán, I. Mondragon, J. Therm. Anal. Cal. 73 (2003) 783–795.
[43] R. Moriana, F. Vilaplana, S. Karlsson, A. Ribes-Greus, Composites A 42 (2011)
30–40.
[44] C. Albano, J. Gonzalez, M. Ichazo, D. Kaiser, Polym. Degrad. Stabil. 66 (1999)
179–190.
[45] K.C.M. Nair, S. Thomas, G. Groeninckx, Compos. Sci. Technol. 61 (2001)
2519–2529.
[46] X.L. Xie, K.L. Fung, R.K.Y. Li, S.C. Tjong, Y.-W. May, J. Polym. Sci. Part B: Polym.
Phys. 40 (2002) 1214–1222.
[47] P.V. Joseph, K. Joseph, S. Thoas, C.K.S. Pillai, V.S. Prasad, G. Groeninckx,
M. Sarkissova, Composites A 34 (2003) 253–266.
[48] P. Gañán, S. Garbizu, R. Llano-Ponte, I. Mondragon, Polym. Compos. 26 (2005)
121–124.
[49] J.M.F. Paiva, E. Frollini, Macromol. Mater. Eng. 291 (2006) 405–417.
[50] J.T. Kim, A.N. Netravali, J. Biobased Mater. Bioenergy 4 (4) (2010) 338–345.
[51] M.F. Rosa, B.-S. Chiou, E.S. Medeiros, D.F. Woods, T.G. William, L.H.C. Mattoso,
W.J. Orts, S.H. Iman, Bioresource Technol. 100 (2009) 5196–5202.
[52] H.P.G. Santafé Jr., R.J.S. Rodriguez, S.N. Monteiro, T.E. Castillo, Characterization of
thermogravimetric behavior of polyester composites reinforced with coir fiber.
In: Characterization of Minerals, Metals and Materials Symposium—TMS 2011
Annual Conference, San Diego, CA, USA, 2011, pp. 1–6.
[53] H.F. Mark, N.,.M. Bikales, C.G. Overberger, G. Menges, Encyclopedia of
Polymer Science and Engineering, John Wiley & Sons, New York, 1988.
[54] C.G. Silva, D.B. Benaducci, E. Frollini, BioResources 7 (1) (2011) 78–98.
[55] N.M. Julkapli, H.M. Akil, Polym.-Plast. Technol. Eng. 49 (2010) 147–153.
[56] H.M. Akil, M.F. Omar, A.A.M. Mazuki, S. Safiee, Z.A.M. Ishak, A.A. Bakar, Mater.
Des. 32 (2011) 4107–4121.
[57] F.M.B. Coutinho, T.H.S. Costa, D.L. Carvalho, M.M. Gorelova, L.C. de Santa
Maria, Polym. Test. 17 (1988) 299–310.
[58] G.H. Doh, S.-Y. Lee, I.-A. Kang, Y.-T. Kong, Compos. Struct. 68 (2005) 103–108.
[59] R. Bhardwaj, A.K. Mohanty, L.T. Drzal, F. Pourboghrat, M. Misra, Biomacro-
molecules 7 (2006) 2044–2051.
[60] Y. Lei, Q. Wu, F. Yao, Y. Xu, Composites A 38 (2007) 1664–1674.
[61] J. George, S.S. Bhagawan, S. Thomas, J. Therm. Anal. 47 (1996) 1121–1140.
[62] S. Luo, A.N. Netravali, Polym. Compos. 20 (3) (1998) 367–378.
[63] P. Threepopnatkul, N. Kaerkitcha, N. Athipongarporn, Composites B 40 (2009)
628–632.
[64] S.-H. Lee, S. Wang, Composites A 37 (2006) 80–91.
[65] Y.-F. Shih, Mater. Sci. Eng. A 445–446 (2007) 289–295.
[66] S. Singh, A.K. Mohanty, T. Sugie, Y. Takai, H. Hamada, Composites A 39 (2008)
875–886.
[67] E.S. Zainudin, S.M. Sapuan, K. Abdan, M.T.M. Mohamad, Mater. Des. 30 (2009)
557–562.
[68] M.A. Semsarzadeh, M. Mehrabzadeh, S.S. Arabshahi, Eur. Polym. 38 (2002)
351–358.
[69] C.G. Mothé, C.R. de Araujo, M.A. de Oliveira, M.I. Yoshida, J. Therm. Anal. Cal.
67 (2002) 305–312.
[70] J.J.M. Órfa~ o, F.J.A. Antunes, J.L. Figueiredo, Fuel 78 (1999) 349–358.