Composites Part B 88 (2016) 1e10

Contents lists available at ScienceDirect

Composites Part B
journal homepage: www.elsevier.com/locate/compositesb

Mechanisms of thermal decomposition of natural bre composites

Mizi Fan a, b, *, Alexander Naughton b
a
College of Material Engineering, Fujian Agriculture and Forestry University, Fuzhou, 350002, PR China
b
College of Engineering, Design and Physical Sciences, Brunel University, UB8 3PH, United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents research pertaining to the thermal decomposition of natural bre composites (NFC).
Received 28 July 2015 The behaviour and damage prole of NFCs have been determined, and comprehensive mechanisms of
Received in revised form thermal decomposition interpreted at both macroscopic and microscopic levels. The results show that
15 September 2015
the process of thermal decomposition of NFCs underwent complex processes from the initial light dis-
Accepted 6 October 2015
Available online 14 November 2015
colouration to blackening and in some cases blistering, and from the decomposition of polyester to bres
to delamination of lamina with the dictated inuence of natural bre and interface structure. Ignition
was closely related to the Vf of NFCs. The damage prole of NFCs included a combust region, a char
Keywords:
A-Polymer-matrix composites
region, a degradation region and a region of virgin material. This paper concludes that the thermal
B-Thermal properties decomposition of NFC is different from that of GFRP and hence its thermomechanical modelling will
D-Optical microscopy require the development of specic terms to account for the inclusion of natural bres.
E-Heat treatment 2015 Elsevier Ltd. All rights reserved.
Thermal decomposition mechanisms

1. Introduction number of methods have been adopted to assess the re perfor-

Natural bre composites (NFC) have been identied as a po-
tential low impact alternative to glass bre reinforced polymers
(GRP) for use in the construction industry. GRPs have, up until
recently, been more economically viable than NFCs. However, with
the economic viability of construction materials set to become
increasingly inuenced by their embodied impact, natural bres
that are less harmful to humans, machinery and the environment
are a realistic alternative to the energy intensive production of glass
bre. If with research and development, the unique technical
problems posed by natural bre reinforcement can be resolved,
NFCs have vast potential as construction materials and as replace-
ment of GRPs in certain applications. The assimilation of NFCs into
the construction industry will depend on their proven performance
in use, their economic viability and the condence with which
structural and architectural components can be designed.
Fire performance of polymer composites has been a great in-
terest of the composite researches, e.g. the heat release and carbon
monoxide yield of thermoset matrix composites [1,2]; re retardant
and other protection of composites [3,4], and thermal decomposi-
tion is the most important consideration in terms of re [5]. A
* Corresponding author. College of Material Engineering, Fujian Agriculture and
Forestry University, Fuzhou, 350002, PR China.
E-mail address: fanfafu@gmail.com (M. Fan).
mance of FRP composites such as ISO 9705 (full-scale room test for
surface products), ISO 13784-1 and ISO 13784-2 (large- and small-
room tests for sandwich panels) and ISO 13785-2 (large-scale test
for faades) (BS ISO 25762: 2009, 2009). The thermal decomposi-
tion of a solid material generates ignitable gasses that burn above
the surface of the material. If the burning gases feedback sufcient
heat to propagate the further decomposition of the solid state
material, the process becomes cyclic, leading to the complete
degradation of the material [5], and contributes to the re load in a
building. There appears to be some disparity between different
researchers on the thermal and re performance of NFCs. Some
researchers have found an improved thermal stability [6,7], while
others have found diminished thermal stability with the inclusion
of natural bres [8]. Some researchers have found ignition time to
be reduced [8] while others found ignition times increased [9]. The
disparity suggests that the bre type, matrix type and the inter-
action between the matrix and bre has a signicant inuence on
the mechanisms of thermal decomposition. It also suggests that
these mechanisms are not yet fully understood.
The aim of this paper is to develop a better understanding of the
mechanisms of thermal decomposition for NFCs. Flat panel com-
posites have been uniformly heated on one surface to present a
clear distribution of damage through the depth of the composites,
referred to as a damage prole, which is very different to that of
GRPs. The damage prole has been characterised in detail in order
to characterise the mechanisms for the thermal decomposition.

http://dx.doi.org/10.1016/j.compositesb.2015.10.038
1359-8368/ 2015 Elsevier Ltd. All rights reserved.
2 M. Fan, A. Naughton / Composites Part B 88 (2016) 1e10
Several types of observable post-re induced damages that are
associated with thermal decomposition and degradation of the
composites, e.g. charring, overheating of matrix phase, cracking
and delamination, were also examined.
2. Materials and methods
2.1. Raw materials and NFCs panels
Hemp bre was chosen as a representative natural bre for
reinforcement and hemp bre polyester composites as NFCs for the
investigation of thermal decomposition in this research.
A needle punched non-woven hemp bre mat was used as
reinforcement. A non-woven mat was chosen for ease of manu-
facture and for comparison with literature. The mat was purchased
as a 1 m by 25 m roll at approximately 800 g/m2 from Hemp-
technology Ltd. A Loyds approved polyester resin, Crystic 2-414pa
was used as the matrix phase for the composite samples. Crystic 2-
414pa is a pre accelerated, thixotropic polyester resin with low
styrene emission.
Hemp mats were cut from the hemp rolls with the dimensions
of 420
420 mm2, which were laminated one layer by layer in an
aluminium mould measuring 420
420 mm2 and applied (roller
coated) with polyester resin. The number of the layers is 3, 5, 7, 9
respectively for the volume of 11, 16, 22, 25% hemp bres. The
mould with the resin-bre mixtures were then compressed under
constant pressure in an electrically heated hydraulic press and the
temperature of the platens in the press were set to 50  C. The
pressing time and pressure were 2 min/mm and 3 MPa respectively.
The mould was then removed from the press to cool to ambient
temperature and the sample was removed. The samples were then
stored in an environmentally controlled room at 20  C/65% relative
humidity before use.
2.2. Methods
The thermal decomposition was investigated through compre-
hensive bench scale re exposures and tests, which are based on a
method which is developed as an adapted version of a cone-
calorimeter test and is referred to in this paper as the reduced
cone test. It is so named because, as much as the apparatus is
similar to cone-calorimetry experiments, the caloric data is not
recorded. Only the cone heater, heat ux meter and rig frame are
retained in the reduced cone tests. Additional sample boundary
conditions are applied and thermocouples are inserted to record
the temperature prole in samples. 200 mm2 at panel samples
were cut from the 420
420 mm2 panels and exposed to a constant
radiant heat on one side to induce thermal damage. The tempera-
ture is recorded at the exposed surface and through the depth of
the material with Type K Inconel alloy thermocouples. The data
output from the reduced cone tests is a temperature prole through
the depth of the samples which can be correlated with the degree
of visible damage though depth (damage prole). A full description
of the method is described in a previous separate paper [10].
Macroscopic analysis has been conducted with an Optika ster-
eozoom microscope [6] to characterise the cross-section of re
damaged samples. Fire damaged samples were cut in half to reveal
the damage prole. The cut surfaces were sanded with a 120 gauge
sandpaper and cleaned with white spirit to remove imperfections
and debris. The surfaces were then inspected under enhanced light
conditions. Colour transitions through the depth were measured
with digital callipers. Photographs were taken with a top mounted
digital camera.
Microscopic analysis has been conducted with a Zeiss Axioplan
[6] microscope to characterise the composition of virgin material
and to characterise the damage prole of re damaged samples
based on physical changes and chromophoric development in the
material. Micrographs were prepared with 30 mm thick which is
thin enough for transmitted light analysis but thick enough to
capture technical hemp bre bundles (10e80 mm) lying in the plane
of the micrograph sample. For re damaged material, micrographs
were extracted at 25 mm from the centre of the exposed surface in
order to accurately correlate the damage prole with the temper-
ature prole measured at that location.
Photographs of magnied micrographs were taken with a top
mounted digital camera and analysed using Adobe Photoshop. The
digital images were calibrated by correlating accurate measure-
ments from the microscope (1 mm) with the pixel count in the
Photoshop analysis environment. This meant that accurate mea-
surements could be automatically recorded. In this manner, the
depth of observed thermal damage was measured.
2.3. Experimental program
The experimental program has been designed in conjunction
with statistical analysis software Design Expert 8. The cone test
experiments are separated into the parameters: Fibre volume
fraction (Vf), thickness (h) and exposure time (t). The radiant heat
ux has been set at 25 kW/m2, which corresponds with the lower
bound heat ux for a domestic building re [11]. The experiments
are based on an optimum factorial design which reduces the
number of replicates required to derive meaningful statistical
conclusions (Table 1).
3. Results and discussion
3.1. Behaviour NFCs under re exposure
It is very interesting that the initial light discolouration of the
exposed surface of NFCs was closely combined with the emission of
a light white smoke prior to blackening of the surface and in some
cases; the blackening of the surface was combined with blistering.
Blistering is caused by a pressure build up in a gas pocket just below
the surface of the material [11]. The pocket could contain a number
of gases associated with the thermal decomposition of polyester
and hemp bre, the detail of which is present in the Section 3.3. In
addition, for NFC the steam may be created by moisture within the
natural bre due to its hydrophilic nature, by considering the
temperature of chart front of around 200  C. It is evident that
within 3e4 min exposure, although the surface of the NFCs was
completely black and glossy with small cracks appearing in the
surface, the surface is less glossy with an increase in Vf. In some
cases, ignition was preceded by the release of ignitable gases from
the blisters. For NFC, auto-ignition was observed for the materials
with between 11 and 25% Vf. Flames reached approximately
40e50 cm in height (Fig. 1A) and died back to a sustained ame of
approximately 30 cm for the duration of the test. Where ignition
did occur, the surface of the residual material was mostly composed
of exposed charred bres with minimal evidence of polyester
Table 1
Experimental design.
Study type: Factorial runs: 247
Design type: D-optimal-Coordinate Exchange
Factor Type Minimum Maximum Levels
A Vf (%) Categoric 11 47 6
B h (mm) Categoric 6 18 5
C t (min) Categoric 0 17.5 7
 M. Fan, A. Naughton / Composites Part B 88 (2016) 1e10
Fig. 1. Behaviour of ames (A) and NFC (B): A Flames immediately after ignition and B Layers within the char region of NFC.
residue (Fig. 1B). Ignition did not occur for samples with Vf higher
than 25%. For these samples, the surface of residual material was
black and glossy with minimal exposed bres and evidence of
polyester residue. The char region is composed of several layers of
char depending on the number of bre laminas included in the char
region (Fig. 1B). Each of the layers in the char displays a similar
composition of charred bres and polyester residue. The surface of
the residual solid material is less blackened than the char and the
polyester is discoloured but intact. The results indicated that the
natural bre, in common with other natural materials, may be able
to build up a layer of char which served as an insulator to heat
transition and hence re protection of NFCs, this could be very
different from those of GRPs, the glass bre of which cannot form
an insulated char layer.
3.2. Understanding damage prole of NFCs
The damage prole dened for FRPs typically includes three
regions: a char region, a degradation region and a region of virgin
material (heat affected and unaffected regions). However, an
additional region has been identied in this research. It has been
shown experimentally that during the process of pyrolysis, the
volatiles produced by the matrix phase in GFRPs are given up to
combustion leaving a surface region of exposed bres [12], which
has also been observed of the NFCs in this investigation. The
composition of this region differs slightly from FRPs in that the
bres are completely charred and there is evidence of residual
polyester in globules around bres. Given that the residual poly-
ester in this region is not homogeneous it can be assumed that
there is no mechanism by which the polyester can transfer shear
stress to the bres. Therefore, the denition of this region as
separate to the char region is important for the calculation of the
residual mechanical properties of the composite as the mechanical
efcacy has been completely degraded. For this paper, this region
was dened as the pyrolysis region since it contains a combination
of pyrolysis and charring (Fig. 2A). The thickness of the pyrolysis
region is dened as (hp) and the depth is dened as (dp). The char
region is composed of heat affected polyester and charred bres,
characterised by deeply black discolouration and a conglomerate
mass of charred material (hc). The degradation region is charac-
terised by a slight discolouration of the material and in some cases
3
includes evidence of delamination and thermally induced voids in
the composite (hd). Finally, a region of virgin material that pre-
sents no visual evidence of thermal degradation can be denoted as
(hv). By virtue of the characteristics of the above dened regions,
the thickness of residual solid material (hr) is dened as the
original thickness of the sample (ho) minus the pyrolysis depth
(dp) (Fig. 2B).
Given the gradual progression of the thermal decomposition
through the depth of the material, it can be difcult to accurately
dene the boundary between two adjoining regions. The boundary
between the pyrolysis region and the char region is known as the
pyrolysis front and is the easiest to dene. The exposed bres in the
pyrolysis region are easily removed to reveal the residual solid
material which can be measured quite accurately to within a tenth
of a millimetre with a set of electronic callipers. The boundaries
between the regions within the residual solid material are much
more difcult to dene. For FRPs, the degradation region is typically
considered to be the gradual transition between charred material
and virgin material. It therefore occupies a relatively narrow region
in the composite and has either been neglected or considered as the
char front for modelling purposes in other research [11]. For the
NFCs in this study, degradation appears to occupy a relatively wide
region which is comparable in thickness to the char region and it
can therefore not be considered as the char front. Instead, the
boundary between the degradation region and the char region will
be referred to as the char front and is characterised as a colour
change from black to dark red in micrographs (discussed below).
The boundary between the degradation region and the virgin ma-
terial is referred to as the degradation front. It is very difcult to
dene the degradation front from visual observation as the colour
transition is too vague. This boundary is thought to occur within a
temperature range which is discussed in more detail in the Section
4. It has been assumed in this section that all material beyond the
char front has undergone thermal degradation to some extent and
that a virgin region does not exist.
The depth of the pyrolysis front (dp), the depth of the char front
(dc) and the depth of degradation front (dd) have been measured
relative to the exposed surface of the composite. For instance, the
depth (dc) is the distance between the exposed surface of the ma-
terial and the char front as shown in Fig. 2B. The boundaries be-
tween each region are identied as a change in colour.
4 M. Fan, A. Naughton / Composites Part B 88 (2016) 1e10
Fig. 2. Damage prole of NFC: A Cross-section of thermal degradation and B Damage regions through depth.
3.3. Mechanisms of NFC decomposition under re exposure
Specic decomposition and degradation have been identied
through analysis of the cross-section of heat damaged samples. It is
apparent that while there is a modest transition between pyrolysis
and char phenomena, the char region contains a combination of
decomposed bre, mass loss of polyester, charred bres, deeply
discoloured polyester and evidence of the degradation of bre/
matrix interface (Figs. 1B and 2B). In general, thermally damaged
polyester in the char region is characterised by dark red, becoming
a pale orange colour with the direction of heat transfer until it re-
sembles the undamaged pale grey colour of virgin material. The
bre in the char region is characterised as a deep black colour,
becoming a dark ochre red with the direction of heat transfer until
it becomes an orange/red.
There is a differential colour transition between the bres and
polyester in the char region. The transition from discoloured red to
pale grey in the polyester is more rapid than the transition from
black to orange in the bres. The polyester begins to decompose at a
higher temperature than hemp bre which could explain the dif-
ferential colour transition between the two. It should be noted that
the material in the char region is not strictly char but is a combi-
nation of overheated polyester and bres in an extreme state of
decomposition.
The degradation region is characterised by pale grey polyester,
red bres becoming orange and pale with the direction of heat
transfer, the occurrence of thermal voids as observed in the char
region and evidence of delamination. The colour of the bres sug-
gests that they have undergone some thermal damage but are not
charred. The red colour of the bres is more indicative of extreme
dehydration and the onset of decomposition.
Polyester in the virgin state is characterised as a pale grey colour
and has a homogeneous appearance. Hemp bres in the virgin state
are characterised as a darker grey/green with dark grey contours.
They appear perpendicular to the direction of heat transfer with
some bres lying at an angle. They also appear as circles or irregular
shapes which are the bres in cross-section.
The epoxy resin used to stabilise the samples prior to cutting is
characterised as a pale green colour. The epoxy resin is apparent in
the location of a void in the composite. Voids in virgin material
occur as air bubbles in the polyester, trapped air in bre and
dislocation between bre and matrix. Air bubbles are characterised
under the microscope as green circles with a neat border between
epoxy and polyester. Trapped air in the bre is generally dispersed
in small bubbles through the bres. Dislocations between bre and
matrix which represent a poor interfacial adhesion have a neat
border between epoxy and polyester. Voids created by thermal
treatment are discussed below.
3.3.1. Decomposition of polyester
When NFCs are exposed to re, the decomposition and degra-
dation of the matrix (polyester) will be taking place rst, as the
 M. Fan, A. Naughton / Composites Part B 88 (2016) 1e10
matrix is normally the rst direct contact to re by considering the
bres within the NFC are usually wrapped with resin. A small de-
gree of mass loss of polyester beyond the pyrolysis front is theo-
retically expected. The mass loss of polyester due to decomposition
can be identied by an irregular void created in a location where
polyester would normally be present. Unlike voids created by air
bubbles during manufacture which are approximately circular, they
have an irregular shape (Fig. 3A) and are contoured by red ther-
mally damaged polyester. This type of void generally occurs within
polyester that has undergone a colour change from grey to deep
red. They are observed in the topmost boundary of the char region
at the pyrolysis front (Fig. 3B). It must be noted that there is a
volume of partially decomposed polyester observed in the char
region of samples which did not ignite during the cone tests.
3.3.2. Process of decomposition of bres
The damage of matrix may accelerate the decomposition and
degradation of bres which can be identied by discolouration of
the bre cell walls and the development of voids within the bre
cross-section. The process is illustrated chronologically in Fig. 4.
The cross-section of undamaged virgin bres within a polyester
matrix are characterised as a light grey colour (Fig. 4A). The cell
walls are well dened and there are minimal voids at the interface
between bre and matrix. Fibres that have experienced the initial
onset of thermal damage are characterised as a light red-orange,
the cell walls are a light red colour and the interface is still intact
at this stage (Fig. 4B). Further exposure and an increase in tem-
perature leads to a deeper discolouration of the bre, the primary
cell walls begin to turn black and there is an evidence of interfacial
separation due to bre shrinkage (Fig. 4C). The decomposition of
the secondary cell walls, which are lignin poor, occurs within the
char region (Fig. 4D). The lignin rich primary cell walls are deeply
blackened but intact. At the pyrolysis front of samples where
ignition occurred, bres are completely decomposed leaving a
residue of charred material and voids (Fig. 4E).
3.3.3. Role of the bre-matrix interface
The bre-matrix interface may play a vital role in the process of
the decomposition and degradation of NFCs. Transferring heat
through interface is directly related to the interface structure. A
compact interface may give rise to a better heat conduction while
the well control air pocket layer may serve as an insulation layer,
especially if the air within the pocket remains still. This mechanism
may continue to take an effect even after the burnt damage of bre
and polyester of the previous layers. After re structure of the bre-
matrix interface can be identied by a void bordering deeply
blackened bres and are observed in the middle zone of the char
Fig. 3. Decomposition of polyester: A Geometry of pyrolysis voids and B Pyrolysis front with evidence of thermally induced voids in polyester.
5
region and within the degradation region as discussed previously
(Fig. 4C). Indeed, this type of void may be different before and after
re, due to possible shrinkage of matrix and bre (Fig. 4F), i.e. this
type of void is not the same as a pre-existing void from manufac-
ture. The contours of the polyester and charred bres match sug-
gesting that they were conjoined at one point: voids between bre
and matrix which indicate a poor interfacial bond created during
manufacture do not share this attribute, as discussed above. The
contour of the polyester is slightly red which suggests that it has
undergone a higher degree of thermal damage than the polyester in
the immediate area (Fig. 4E). Given that the temperature in the
middle of the char region is expected to be too low for the pyrolysis
of polyester to occur, it is thought that these voids were created by
the process of thermal shrinkage associated with dehydration and
thermal decomposition of the bre. The red polyester is evidence of
thermal damage. It is possible that the bres, with a higher thermal
conductivity, increase the build up of heat in the immediate area,
hence resulting in thermal damage to the polyester that was pre-
viously in contact with the bre. This suggests that these voids may
represent possible degradation of interfacial bond with an increase
in temperature and a consequent potential degradation of me-
chanical properties. The degradation of interfacial bond is most
evident in the char region.
3.3.4. Decomposition induced delamination
The decomposition may be followed by a delamination between
lamina in the composite which can be identied by a large hori-
zontal void perpendicular to the direction of heat ow in this study
(Fig. 5A). This may be due to at least the shear stress generated by
the different deformation between matrix and bres or the gas
pressure. Micro-delamination can be identied by small voids and
appear to be the beginnings of a larger delamination. Delamination
normally occurs towards the lower part of the degradation region
but can occur higher and often coincides with divisions between
the non-woven bre mats (lamina). They can occupy the entire
exposed area of the composite but in general they occupy nite
regions. They can occur at multiple depths and in multiple locations
in the same composite. They often coincide with the presence of
existing voids which interconnect with each other in the case of
micro-delamination and interconnect with larger delamination.
The delamination voids are not pre-existing voids from manu-
facture. Each contour of the void region coincides with each
opposing contour as shown in Fig. 5B which suggests that the void
has opened up during thermal degradation; contours that were
previously conjoined have become dislocated. This could occur as a
result of thermal movement but is more likely due to the build up of
gases including steam which would result in a pressure spike in
6 M. Fan, A. Naughton / Composites Part B 88 (2016) 1e10
Fig. 4. Decomposition process of natural bres: A
that location [13]. This mechanism is believed to be associated with
the surface bulging observed during testing. Surface bulging was
linked to a build up of gases which ignited when the gas pocket
burst. The build up of gases within the composite could be exac-
erbated by the presence of voids and poor interfacial bond from
manufacture which would help propagate the process. The pres-
sure induced by the onset of a gas pocket would initiate the
dislocation between matrix and bres. This dislocation would
follow the path of least resistance and hence force a separation in
the relatively weak interface between two laminates (Fig. 5C).
Delamination can also appear in a fracture pattern. As described
above, the fracture delamination presents evidence of dislocation of
interfaces as shown in Fig. 5D. The fracture pattern appears where
there is a concentration of voids in the composite and a concen-
tration of poor interfacial bond. The mechanics of delamination are
similar to those described above. However, the concentration of
voids and an abundance of weak interfaces create an inter-
connected web of delamination which looks like a fracture.
The presence of delamination appears in some cases to protect
lower lying material from thermal degradation. It would appear that
B C D E.
the void has an insulating effect by inhibiting the thermal transfer
within the solid as shown in Fig. 5E. The insulating mechanism has
been well documented for FRPs by Gibson 2003 [11].
3.4. Correlation of decomposition and temperature
By combining the temperature results from the reduced cone
tests with the results from the analysis of the cross-section, it is
possible to determine the temperature at which the pyrolysis front
and the char front occur and the average temperature within each
region. For this paper, these temperatures will be referred to under
the umbrella term thermal decomposition temperature. The re-
sults are presented with respect to the experimental programs
outlined previously; exposure time, bre volume fraction and
depth. The microscopic samples are examined with respect to the
temperature through depth.
The temperature prole is dened as a function for temperature
with respect to depth (T(x)) which has been derived from the
maximum temperature recorded through the depth of samples at
the end of testing. This function has been used in conjunction with
 M. Fan, A. Naughton / Composites Part B 88 (2016) 1e10 7

Fig. 5. Decomposition induced delamination process.

the measurement of pyrolysis depth (dp), and char depth (dc) to
dene their associated temperatures: the pyrolysis temperature
(Tp) which occurs at the pyrolysis front and the char temperature
(Tc) which occurs at the char front.
The average temperature within each region through depth can
be evaluated by integrating between the region boundaries and
dividing by the region thickness. In this way, the average temper-
ature in the char region and the degradation region has been
analysed.
The temperature at the pyrolysis front is approximately constant
(z270e335  C) with respect to exposure time (Table 2). Char
temperature increases linearly with respect to exposure time
(Fig. 6). Similarly, the average temperature in the char region and
the degradation region increase linearly with exposure time.
The pyrolysis temperature for samples where ignition occurred
is constant (z354  C 13) with respect to Vf (Table 2). For samples
that did not ignite, the pyrolysis temperature is constant
(z282 12) with respect to Vf. There is a modest reduction in char
temperature with an increase in Vf for samples where ignition
occurred. The char temperature is constant (z204  C 7) for
samples that did not ignite. The average char temperature for
samples that ignited is constant (z271  C 13) with respect to Vf.
Similarly, the average char temperature for samples that did not
ignite is constant (z232  C 10). The temperature of the degra-
dation region decreases linearly with an increase in Vf for samples
where ignition occurred. There is a modest reduction in the average
temperature of the degradation region with an increase in Vf for
samples that did not ignite.
8 M. Fan, A. Naughton / Composites Part B 88 (2016) 1e10

Table 2
Average decomposition temperature (Heat ux: 25 kW/m2).a

Pyrolysis front (Tp) Chart front (Tc) Chart region (c) Degradation (d)

t (second) With respect to exposure time (t). Number of samples: 74; Fibre volume: 25e47%
300 293 (13.21) 180 (6.11) 209 (10.05) 150 (4.67)
450 272 (9.56) 180 (10.56) 208 (8.65) 154 (5.19)
600 288 (9.02) 201 (7.46) 231 (7.79) 172 (8.14)
750 274 (10.58) 198 (6.06) 226 (7.08) 176 (5.11)
900 305 (3.61) 234 (3.42) 262 (3.05) 211 (2.37)
1050 335 (2.99) 240 (4.17) 276 (2.90) 218 (3.21)
Mean 294 206 235 180
Vf With respect to bre volume fraction (Vf). Number of samples: 72; Exposure time: 600 s
Samples where ignition occurred
0.11 347 (7.78) 229 (2.62) 274 (4.38) 200 (2.50)
0.16 374 (8.56) 223 (5.38) 278 (6.47) 186 (5.38)
0.22 345 (17.97) 209 (2.87) 281 (11.03) 164 (3.66)
0.25 352 (4.83) 199 (2.10) 252 (3.17) 158 (0.63)
Mean 354 215 271 177
Samples where ignition did not occurred
0.25 282 (9.93) 202 (9.90) 228 (9,65) 178 (10.11)
0.4 294 (3.74) 212 (1.41) 243 (2.06) 173 (6.36)
0.47 269 (10.03) 198 (3.54) 225 (5.33) 168 (4.17)
Mean 282 204 232 173
h0 (mm) With respect to original thickness (h0). Number of samples: 24; Exposure time: 10 min; Fibre volume: 22%
6 298 (9.73) 271 (9.59) 280 (9.64) 265 (9.43)
9 325 (8.92) 246 (5.85) 294 (13.61) 207 (8.70)
12 330 (6.36) 257 (2.06) 282 (9.93) 210 (16.67)
15 299 (7.39) 190 (8.94) 215 (11.63) 149 (8.05)
18 267 (10.71) 136 (8.09) 173 (9.07) 111 (15.31)
Mean 304 220 249 188
a
Value in ( ) cov (%).

The pyrolysis temperature with respect to thickness for 22% Vf
samples is constant (z304  C 49) (Table 2). There is a reduction in
char front temperature and the average temperature of the char
region with an increase in thickness above 12 mm. The average
temperature of the degradation region decreases linearly with an
increase in thickness.
Although the observed temperature of thermal decomposition
reduces with Vf and increases with exposure time, as noted above,
the temperature of specic decomposition observed in the micro-
graphs should be constant regardless of the bre volume or thick-
ness of the composite. It is important to note that the temperature
of pyrolysis of polyester is approximately 360  C and the values for
the pyrolysis front are not to be construed as the temperature of
pyrolysis. Therefore, the temperature at the pyrolysis front may be
more indicative of the decomposition process of the composite as a
whole, rather than the decomposition of the constituent materials.
This is vitally important for understanding the decomposition of
NFC composites by virtue of unique nature of charring process of
natural bres and the complex interface structure between natural
bres and matrix.
The evidence of mass loss of polyester has been shown to occur
at the pyrolysis front. The temperature of the pyrolysis front has
been shown to be 354 13  C for samples where ignition occurred
and 297 9  C for samples where ignition did not occur. The
melting temperature of polyester is approximately 260  C which is
associated with a reduction in the molecular weight [14]. The
average temperature of the char region has been shown to be
271 13  C where ignition occurred and 232 10  C where ignition
did not occur in most cases. Mass loss at the pyrolysis front for
samples which did not ignite could occur due to the convective
transfer of volatile material; the low molecular weight of melting
polyester could permit such as mechanism. Mass could also be lost
by handling of samples and removal of the pyrolysis layer for
sample preparation. This could account for the lower than expected
temperature of pyrolysis (z300  C) for samples which did not
ignite. Another explanation could be that where ignition does not
occur, there is a greater volume of non-pyrolysed polyester residue
in the pyrolysis region as observed previously.

4. Overall appraisal of decomposition mechanisms

The observed decomposition of hemp bres can be described

Fig. 6. Char front temperature with respect to exposure time. 25 kW/m2.
with reference to the position within the damaged cross-section
and the temperature at that position. The temperature of decom-
position of the bres is dictated by the decomposition of the con-
stituent material: cellulose (64%), hemicellulose (14%), lignin (5%),
pectin (5%) and other extraneous material. Technical hemp bre
bundles are composed of elementary bres bonded together by
lignin and pectin [15]. The elementary bres are mainly composed
of microbrils which form different cell wall layers. The inner part
of the secondary cell wall is mostly composed of cellulose and
hemicellulose. The outer part of the secondary cell wall has high
 M. Fan, A. Naughton / Composites Part B 88 (2016) 1e10
levels of lignin and the primary cell wall is composed of lignin and
pectin [16]. The measured temperatures of decomposition in this
study correspond with the known decomposition temperatures of
hemp bres. The onset of decomposition of hemp bres occurs at
150  C [17] which is associated with the decomposition of hemi-
cellulose up to 180  C [16]. Lignin in hemp bres begins to
decompose at 160  C and continues to decompose up to 400  C [18].
Cellulose decomposition occurs between 200 and 300  C with rapid
mass loss occurring between 300 and 360  C [17]. The decompo-
sition of the lignin in the primary cell wall could result in the
degradation of the bre matrix interface as observed in Fig. 4C. The
decomposition of lignin and hemicellulose in the secondary cell
wall of elementary bres results in the voids observed in Fig. 4D.
The complete decomposition of constituent materials results in the
voids observed at the pyrolysis front in Fig. 4E. The process is
itemised with respect to the measured temperature of the damage
prole below:
Degradation region 150e180  C
 Hemicellulose in the secondary cell wall begins to decompose.
 Dehydration of chemically bound moisture occurs with a
resulting mass loss of 10%.
 Lignin in the primary cell wall begins to decompose and will
continue to decompose until z400  C.
C
Char region 200e300
 Cellulose in the secondary cell wall begins to decompose
resulting in the formation of volatiles, non-combustible gases,
tar and some char.
 Lignin continues to decompose.
Pyrolysis front 300e360  C
 Rapid mass loss occurs.
 Cellulose decomposition results in the formation of carbona-
ceous residues.
 Hemicellulose has completely decomposed.
 Lignin is decomposed.
Hemicellulose forms covalent bonds with lignin and hydrogen
bonds are formed between xylan1 and cellulose [16]. The decom-
position of hemicellulose leads to the disintegration of the bres
into cellulose microbrils resulting in the degradation of mechan-
ical properties of the bre bundle [16]. It is expected that the
reduction in mechanical properties of the bre bundles will
signicantly affect the mechanical properties of the composite.
Therefore, there is a reduction in mechanical properties at rela-
tively low temperatures beyond the char front which is not
accounted for in present models for residual mechanical strength of
re damaged FRPs. This process needs to be considered in further
research.
Contrary to GFRPs the damage prole of NFCs in this study is
dominated by the decomposition of the bres. For GFRPs, the
mechanisms for thermal decomposition are usually dictated by the
polyester due to the higher glass transition temperature of glass
bres. Hence the damage prole of GFRPs is relatively simple when
compared to NFCs, being composed of three damage regions; a char
region, a slim region of thermally damaged polyester and a region
of virgin material [19]. Moreover, the char built-up of natural bres
could complicate the mechanisms of further re exposure. Conse-
quently, it is necessary to consider that the thermal decomposition
1
Xylan is a type of polysaccharide found in plant cell walls and some algae.
9
kinetics of NFCs may be completely different to GFRPs and tradi-
tional models for thermal decomposition of composites need to be
altered to account for the inclusion of natural bres.
It has been shown that the temperature of the char front in-
creases with exposure time. This suggests that with increased
exposure there is a change in the thermal properties of the com-
posite. The change in physical composition such as the reduction in
density due to thermal decomposition would result in a subsequent
change in the thermal properties. With increased duration of high
temperature the charred bres produce monomers that require a
higher temperature before decomposition can continue [17].
However, the increased temperature of decomposition could be a
result of temperature congestion caused by the insulation of the
sample holder. Congestion is evidenced in the effect of thickness on
the temperature of decomposition. It has been noted that the
temperature beyond the pyrolysis front decreases with an increase
in thickness. This suggests that the heat transfer properties change
with thickness. The most sensible conclusion of this effect is that
the low conductivity of the boundary material has a greater inu-
ence on the sample material the closer it is to the heat source.
5. Conclusions
The behaviour and damage prole of NFCs under re exposure
have been determined, and their comprehensive mechanisms of
thermal decomposition have been illustrated. NFCs under re
exposure rstly underwent complex processes from the initial light
discolouration to the blackening and in some cases blistering. The
former was closely combined with the emission of a light white
smoke and the latter were caused by a pressure build up in a gas
pocket associated with the thermal decomposition of polyester and
hemp bre. Auto-ignition was observed for the materials with be-
tween 11 and 25% Vf, but did not occur with Vf higher than 25%. The
damage prole of NFCs included a combust region, in addition to a
char region, a degradation region and a region of virgin material.
The decomposition of NFCs started from the polyester due to its
rst direct contact with re, followed by the bres which was
accelerated by the decomposition and degradation of matrix. Fire-
matrix interface played a vital role in the process of the decom-
position and degradation of NFCs. The decomposition could also
induce the delamination between lamina in the NFC composites. It
has been shown that the process of thermal decomposition of NFCs
could be dictated by the decomposition and charring of natural
bres. It is therefore necessary to update and develop existing
thermal decomposition models to account for the effect of natural
bres within polymer composites. The degradation of the interface
is expected to result in diminished mechanical properties. Further
research should focus on a mechanical model that can account for
the efcacy of interfacial adhesion after thermal treatment.
References
[1] Mouritz AP, Mathys Z, Gibson AG. Heat release of polymer composites in re.
Compos Part A 2006;37:1040e54.
[2] Littimer BY, Quellette J. Properties of composite materials for thermal analysis
involving res. Compos Part A 2006;37:1068e81.
[3] Becker C, Gabbardo AD, Wypych F, Amico SC. Mechanical and ame-retardant
properties of epoxy/Mg-AILDH composites. Compos Part A 2011;42:196e202.
[4] Correia JR, Bramco FA, Ferreira JG. The effect of different passive re protec-
tion systems on the re reaction properties of GFRP-pultruded proles for
civil construction. Compos Part A 2010;41:441e52.
[5] Beyler CL, Hirschler MM. Thermal decomposition of polymers. In: SFPE
handbook of re protection engineering. Quincy: NFPA; 2002.
[6] Kim H, Yang H, Kim H, Lee B, Hwang T. Thermal properties of agro-our-lled
biodegradable polymer bio-composites. J Therm Anal Calorim 2005 08;81(2):
299e306.
[7] Rudnik E. Thermal properties of biocomposites. J Therm Anal Calorim 2007
05;88(2):495e8.
10 M. Fan, A. Naughton / Composites Part B 88 (2016) 1e10
[8] Dorez G, Taguet A, Ferry L, Lopez-Cuesta JM. Thermal and re behavior of
natural bers/PBS biocomposites. Polym Degrad Stab 2013 1;98(1):87e95.
[9] Hapuarachchi TD, Ren G, Fan M, Hogg PJ, Peijs T. Fire retardancy of natural
bre reinforced sheet moulding compound. Appl Compos Mater 2007 07/
01;14(4):251e64.
[10] Naughton A, Fan M, Bregulla J. Fire resistance characterisation of hemp bre
reinfroced polymer composite for use in the construction industry. Compos
Part B 2014;60:546e54.
[11] Gibson AG, Wright PNH, Wu Y, Mouritz AP, Mathys Z, Gardiner CP. Modelling
residual mechanical properties of polymer composites after re. Plastics
Rubber Compos 2003 02;32(2):81e90.
[12] Mouritz AP, Mathys Z. Mechanical properties of re-damaged glass-reinforced
phenolic composites. Fire Mater 2000;24(2):67e75.
[13] Burchill PJ, Mathys Z, Gardiner CP. An analysis of the burning of polyester and
vinylester bre glass composites. Fire Mater 2005;29(4):249e64.
[14] Rudin A. The elements of polymer science and engineering. 2nd ed. London:
Academic Press; 1999.
[15] Fan M. Characterisation and performance of elementary hemp bre: factors
inuencing tensile strength. Bio Resour 2010;5(4):2307.
[16] Thygesen A. Properties of hemp bre polymer composites e an optimisation
of bre properties using novel debration methods and bre characterisation.
Denmark: Royal Veterinary and Agricultural University; 2006.
[17] Ming G, Qui-ju D. Studies on the thermal degradation of cellulosic bers
treated with ame retardants. Chin J Process Eng 2006;6(2).
[18] Chapple S, Anandjiwala R. Flammability of natural ber-reinforced compos-
ites and strategies for re retardancy: a review. J Thermoplast Compos Mater
2010 November 01;23(6):871e93.
[19] Mouritz AP, Feih S, Mathys Z, Gibson AG. Mechanical property degradation of
naval composite materials. Fire Technol 2011;47(4):913e39.