Study of moisture absorption in natural fiber plastic composites

Abstract

Moisture absorption of natural fiber plastic composites is one major concern in their outdoor applications. Traditionally diffusion theory is applied to understand the mechanism of moisture absorption; but it cannot address the relationship between the microscopic structure-infinite 3D-network and the moisture absorption. The purpose of this study is to introduce percolation theory into this field and conduct some preliminary work. First, two new concepts, accessible fiber ratio and diffusion-permeability coefficient, were defined; secondly, a percolation model was developed to estimate the critical accessible fiber ratio; finally, the moisture absorption and electrical conduction behavior of composites with different fiber loadings were investigated. At high fiber loading when accessible fiber ratio is high, the diffusion process is the dominant mechanism; while at low fiber loading close to and below percolation threshold, percolation is the dominant mechanism. The over-estimate of accessible fiber ratio led to discrepancies between the observed and model estimates.

Full text

Study of moisture absorption in natural fiber plastic composites
W. Wang, M. Sain
*
, P.A. Cooper
Centre for Biocomposites and Biomaterials Processing, Faculty of Forestry, University of Toronto, Canada
Received 20 July 2005; accepted 20 July 2005
Available online 26 September 2005
Abstract
Moisture absorption of natural fiber plastic composites is one major concern in their outdoor applications. Traditionally diffu-
sion theory is applied to understand the mechanism of moisture absorption; but it cannot address the relationship between the
microscopic structure-infinite 3D-network and the moisture absorption. The purpose of this study is to introduce percolation theory
into this field and conduct some preliminary work. First, two new concepts, accessible fiber ratio and diffusion-permeability coef-
ficient, were defined; secondly, a percolation model was developed to estimate the critical accessible fiber ratio; finally, the moisture
absorption and electrical conduction behavior of composites with different fiber loadings were investigated. At high fiber loading
when accessible fiber ratio is high, the diffusion process is the dominant mechanism; while at low fiber loading close to and below
percolation threshold, percolation is the dominant mechanism. The over-estimate of accessible fiber ratio led to discrepancies
between the observed and model estimates.
 2005 Elsevier Ltd. All rights reserved.
Keywords: Natural fiber reinforced composites; Diffusion; Percolation theory; Electrical conductivity
1. Introduction
Natural fiber plastic composites (NFPCs) have
gained increasing interest and bro ader application dur -
ing the past decade due to the desirable properties of
natural fibers [1]. Presently, the main application of
NFPCs is as construction materials [2,3], such as deck-
ing and railing products. However, natural fibers also
have an undesirable property, namely, hygroscopicity
because of their chemical constituents. The moisture
absorption by composites containing natural fibers has
several adverse effects on their properties and thus af-
fects their long-term performance. For example, in-
creased moisture decreases their mechanical properties
[4,5], provides the necessary condition for biodegrada-
tion, and changes their dimensions [6]. Numerous efforts
have been made to address this issue. Coupling agents,
compatibilizers or other chemical modifications are used
to improve the moisture resistance of composites [7–9].
However, moisture absorption of composites is still
one a major concern especially for their outdoor
application.
Moisture absorption increases with increasing fiber
loading. Thomas et al. [10] investigated the relationship
between the moisture absorption of pineapple-leaf fiber
reinforced low density polyethylene (LDPE) composites
and the fiber loadings (10%, 20%, and 30% by weight).
They found that the moisture absorption increased al-
most linearly with the fiber loading. Stark [4] found that
wood flour-pol ypropylene (PP) composites with 20 wt%
wood flour reached equilibrium after 1500 h in a water
bath and absorbed only 1.4% moisture while composites
with 40 wt% loading reached equilibrium after 1200 h
water submersion and absorbed approximately 9.0%
0266-3538/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compscitech.2005.07.027
*
Corresponding author. Present address: Department of Chemical
Engineering and Applied Chemistry, Earth Science Center, 33 Will-
cocks Street, University of Toronto, Toronto, Ont., Canada M5S 3B3.
Tel.: +1 416 946 3191; fax: +1 416 978 3834.
E-mail address: m.sain@utoronto.ca (M. Sain).
Composites Science and Technology 66 (2006) 379–386
COMPOSITES
SCIENCE AND
TECHNOLOGY
www.elsevier.com/locate/compscitech

moisture. After the analysis, she concluded that the
wood flour is inhibited from ab sorbing moisture due
to encapsulation of the wood flour by the PP matr ix
and that the degree of encapsulation is greater for the
20% wood flour composite than that for the 40 % wood
flour composite.
Traditionally, the diffusion theory has been applied to
understand the mechanism of moisture absorption in
composites [11–14]. A model based on diffusion theory
was developed to understand the relationship between
the fiber loading/fiber orientation and apparent diffu-
sion coefficient [15].
As one theory regarding mass transfer, diffusion
theory is based on the assumption that the medium
structure is homogeneous and that the mass transfer is
only the result of the random molecular motion of the
diffusion agent whether it is in gaseous or liquid state
[16]. As the result of this molecular motion, the overall
trend is that diffusion agent moves in the direction from
higher to lower concentration as stated in FickÕ s Law
[17].
Diffusion theory does shed light on the mechanism of
moisture absorption and provides a framework for
quantifying the mass transfer; however, it fails to
address the structural complexity of composites. Since
natural fibers and polymer matrix exhibit diff erent prop-
erties in terms of moisture absorption, the fiber distribu-
tion in polymer matrix is a key to the overall moisture
absorption of composites. Hence, all the factors that
affect the fiber distribution will finally affect the moisture
absorption ability. These fact ors, includi ng fiber concen-
tration, size, shape, need to be addressed individually to
get a full understanding of the mechanism of moisture
absorption.
The problem raised above falls into the domain of
percolation theory which deals with the randomness of
the medium [18]. At the same time, the network model-
ing which takes into account the randomness of both
diffusion agent and medium significantly promotes the
development of percolation theory. Thus, to some
extent, the apparent diffusion coefficient is integrated
into percolation theory. Moreover, it provides further
understanding of composites which is lacking in diffu-
sion theory. One successful application of percolation
theory is the development of a percolation model con-
cerning controlled drug release system [19–25].
The present research is trying to introduce percola-
tion theory to the NFPC field, to better understand
the mechanism of moisture absorption in natural fiber
reinforced composites. To this end, the moisture absorp-
tion and electrical conductivity of composites with
different fiber loadings were investigated. Two new con-
cepts based on percolation theory were defined: accessi-
ble fiber ratio and diffusion-permeability coefficient. The
objectives are: (i) to develop a model to estimate critical
accessible fiber ratio; (ii) to understand the moisture
absorption and electrical cond uctivity behaviors using
estimated critical accessible fiber ratio and other perco-
lation concepts.
2. Theoretical approach
2.1. Diffu sion and Fick’s laws
Diffusion is the process by which matter is transported
from one part of a system to another as a result of ran-
dom molecular motions [16]. Fick first put diffusion on
a quantitative basis by adopting the mathematical equa-
tion of heat conduction after realizing the analogy be-
tween the heat conduction and the diffusion process
F ¼D
oC
ox
; ð1Þ
where F is the rate of transfer per unit area of section, C
is the concentration of diffusion substance, x is the space
coordinate measured normal to the section, and D is the
diffusion coefficient. Eq. (1) is referred to as FickÕ s first
law.
If the diffusion coefficient is constant, and if the con-
centration gradient is only along the x-axis, in other
words, if diffusion is one -dimensional, the fundamental
differential equation of diffusion in an isotropic medium
is derived by considering a rectangular whose sides are
parallel to the axes of coordinates as follows:
oC
ot
¼ D
o
2
C
ox
2
. ð2Þ
Eq. (2) is nor mally referred to as FickÕs second law.
Its solution for a plane sheet with uniform initial dis-
tribution and equal initial surface concentrations under
non-steady-state can be expressed by
M
t
M
1
¼ 1 
X
1
n¼0
8
ð2n þ 1Þ
2
p
2
exp½ðDð2n þ 1Þ
2
p
2t
=ð4l
2
ÞÞ;
ð3Þ
where M
t
denotes the total amount of diffusion sub-
stance entering the sheet at time t, M
1
denotes the cor-
responding quantity after infinite time, and l denotes the
half thickness of sheet.
At initial absorption stage, moisture absorption (M
t
)
increases linearly with
ffiffi
t
p
; hence Eq. (2) can be simpli-
fied to the following equation:
M
t
¼
2M
1
ffiffiffiffi
D
p
ffiffiffi
p
p
ffiffi
t
p
l
. ð4Þ
Thus, the average diffusion coefficient can be calculated
as follows:
D ¼
p
4
M
2
1
l
2
h
2
; ð5Þ
where h is the slope of M
t
vs.
ffiffi
t
p
plot.
380 W. Wang et al. / Composites Science and Technology 66 (2006) 379–386

2.2. Basic percolation concepts
Many properties of a macroscopic system are essen-
tially determined by the connectivity of the system ele-
ments. The properties of a system which emerge at the
onset of macroscopic connectivity within it are known
as percolation properties, e.g., conductivity.
2.2.1. Infinite cluster and percolation threshold p
c
The following square lattice (Fig. 1) is a simplified
representation of the dispersion of one component in
another component. Suppose that each square repre-
sents either one unit of natural fiber or one unit of plas-
tic, regardless of their differences in morphology.
Natural fibers are hygroscopic and ready to absorb
moisture while the plastics discussed here are highly
hydrophobic. The square is marked dark if occupied
by fiber; otherwise white if occupied by plastics.
Suppose that the fiber content here is defined as the
percentage of fiber units over total units (fiber and plas-
tic) and is designated as p. Then the probability (P) that
one individual square is occupied by fibers is equal to
the fiber content p.
At very low fiber content, individual fibers are com-
pletely encapsulated by plastic. At a little bit higher fiber
content, fiber clusters, which are defined by one fiber
square having at least one neighboring fiber square,
are formed. Fig. 1(a) shows the fiber distribution at
low fiber content. Fibers either exist individually or form
finite clusters. Since both individual fibers and finite fi-
ber clusters are still encapsulated by plastics, they are
un-accessible to moisture except the fibers exposed on
the approximate surface of the lattice. Hence, if only
the horizontal top surface of this lattice is open to mois-
ture while two lateral surfaces are sealed, water mole-
cules cannot penetrate the lattice through fibers to
reach the bottom of the lattice.
Fig. 1(b) exhibits another scenario of fiber distribu-
tion in the lattice. At high fiber content, all fiber squares
connect together and plastics fail to encapsulate these
fibers. Now if we repeat the same experiment, water
molecules easily find their ways to penetrate the lattice
and reach its bottom.
As shown in Fig. 1, the fiber clusters become large r
with increasing fiber content. Between the two above-
mentioned scenarios, there must be a fiber content at
which the fiber cluster first spans from one side of the
lattice to the opposite side. This spanning cluster is
called an ‘‘infinite cluster’’ ; the fiber content at this point
is called the ‘‘percolation threshold’’ or ‘‘critical fiber
content’’ and is designated as p
c
(%). Above this thresh-
old, the higher the fiber content, the more fibers connect
to this infinite cluster. This cluster serves as passages for
water molecules to travel through the lattice from one
side to the other.
2.2.2. P
1
P
1
denotes the probability that a fiber square
belongs to the infinite cluster. In other words, it denotes
the proportion of fibers belonging to the infinite cluster.
Within this infinite cluster, two types of fibers are differ-
entiated to facilitate the discussion of conductivity. As
shown in Fig. 2, fibers marked with ‘‘e’’ are the dangling
ends of infinite cluster; they absorb moisture but do not
serve as moisture passages and so do not contribute to
the conductiv ity of the lattice. Fibers marked with ‘‘l’’
form loops inside the infinite cluster and also do not in-
crease the conductivity of the lattice. They have exactly
the same effect as ÔeÕ fibers. ‘‘F’’ represents fibers which
form continuous passages for moisture and thus con-
tribute to the conductivity of the lattice.
If the fiber content is lower than p
c
, there is no infinite
cluster and P
1
is zero. Otherwise, its value increases
above p
c
as follows:
P
1
ðp  p
c
Þ
b
; ð6Þ
where p is the fiber content (%), p
c
is the percolation
threshold, and b is a critical static exponent. The value
of b is universal. It only depends on the dimension of
the lattice. For a three-dimensional lattice, b is approx-
imately 0.42 [26].
Fig. 1. Fiber distributions in plastic at different fiber contents: (a) low fiber content; (b) high fiber content.
W. Wang et al. / Composites Science and Technology 66 (2006) 379–386 381

2.2.3. Percolation factor f
Flow properties, such as electrical conductivity, of
the lattice are of interest only above the percolation
threshold. Flow properties are related to the connectiv-
ity of the lattice. Lee et al. [27] defined percolation factor
(f) as a representation of connectivity as follows:
f ¼ ppR
2
; ð7Þ
where p is the fiber content (%) of composites and R is
the average radius of the fiber.
Suppose that the fiber content is higher than percola-
tion threshold and that the infinite cluster has already
formed. The low connectivity implies that the number
of moisture passages is low and that the passages are tor-
tuous and lengthy (Fig. 2(a)). Hence, it takes a long time
for mo isture molecules to travel through the lattice. The
conductivity of the lattice is also low. Fig. 2(b) shows a lat-
tice with higher connectivity. The higher number of flow
passages increases the conductivity. Moreover, the more
flow passages there are, the more direct (shorter) these
passages will be. This shortens the moisture travel time
from one side of composites to the other.
2.2.4. Apparent diffusion coefficient D
The diffusion problem in a heterogeneous composite
is mathematically equivalent to the electrical conduction
problem [26]. In the previous session, we have intro-
duced the concept describing the relationship between
the conductivity of a composite and physical parameters
of its components. Next, the conduction properties will
be understood from another perspective by looking into
the microscopic network formed by its components.
As discussed before, the infinite percolation cluster
has dangling ends and loops. These parts of the cluster
absorb the diffusing substance – moisture – but do not
contribute to the conduction property. In fact, accord-
ing to random walk theory of diffusion, these dangling
ends and loops act as ‘‘sinks’’ to slow down the trans-
port of moisture in the diffusing direction. By being ta-
ken into account this fact, together with other
percolation concepts, the apparent diffusion coefficient
D holds the following equation close to the percolation
threshold [28]:
D ¼ vD
0
ðp  p
c
Þ
l
; ð8Þ
where vD
0
represents a scali ng factor, and l the conduc-
tivity exponent, which is 2.0 for a three-dimensional lat-
tice. This value (2) will be adopted in present research.
3. Application of percolation theory in NFPCs
In NFPCs, the infinite fiber cluster is a 3D-network.
In order to facilitate the application of percolation the-
ory in NFPC system, the following assumptions are
essential before applic able new concepts can be de fined.
3.1. Assumptions
(1) Plastic is hydrophobic. Moisture can only pene-
trate into the composite through natural fibers.
(2) The composite is void free; the moisture absorp-
tion can be quantified by fiber saturation point
of the accessible fibers (FSP, assumed to be 30%,
dry mass basis).
(3) The samples are thin so moisture can penetrate
through the whole infinite 3D-network which is
present above percolation threshold during a spe-
cific time period; all the natural fibers belonging
to the infinite 3D-ne twork can reach FSP after
water submersion.
(4) Free fibers, which are not connected to the infinite
3D-network, are categorized into two types. Type
I fibers have no access to moisture as they are
dominantly entrapped in the bulk; Type II fibers
are on the proximate surface, and they absorb
moisture if not on the sealed surfaces.
3.2. Acce ssible fiber ratio r
The accessible fiber ratio r was theoretically derived
from the percolation concept P
1
, which denotes the
probability that a fiber square belongs to the infinite
cluster. According to this definition, the accessible fibers
only are those belonging to the infinite 3D-network in
Fig. 2. Flow passages in lattice with different fiber contents: (a) with one flow passage; (b) with two flow passages.
382 W. Wang et al. / Composites Science and Technology 66 (2006) 379–386

NFPCs. However, the application of this theoretical ra-
tio in small samples is limited, since the moisture
absorption by free fibers on the proximate surface is sig-
nificant at low fiber loading.
In the present study, accessible fibers include both the
fibers belonging to the infinite 3D-network and the fi-
bers on the proximate surface of composite. Ratio r
can be computed as follows:
r ¼
M
1
C

FSP
; ð9Þ
where M
1
denotes the maximum moisture absorption
of sample, C denotes the fiber loading in composites
(percentage by weight, the name ‘‘fiber loading’’ is used
to differentiated the name ‘‘fiber content’’ in previous
section). In this research, the FSP of natural fibers is as-
sumed to be 30%.
The percolation threshold of r is called the critical
accessible fiber ratio, designated as r
c
. It is the ratio at
which the infinite 3D-network formed for the first time.
3.3. Diffu sion-permeability coefficient DP
From the proceeding introduction, it can be concluded
that the moisture absorption behavior of NFPCs depends
on both diffusion coefficient and percolation network. In
this context, we define a new property of composite from
diffusion coefficient and designate it as diffusion-perme-
ability coefficient (DP). DP is defined as follows:
DP ¼
h
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2

C  r
p
; ð10Þ
where h is the slope of Fickian M
t
vs.
ffiffi
t
p
plot, C is the
fiber loading of composite, and r is the accessible fiber
ratio. The unit for DP is s
0.5
.
3.4. Modeling
DP is a derivative of diffusion coefficient D correlated
through slope h. By combining Eqs. (8) and (10), we have
DP ¼
h
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2

C  r
p
¼ kðr  r
c
Þ; ð11Þ
where k is a constant.
By investigating the moisture absorption behavior of
composites with different fiber loadings near the thres h-
old, DP values can be calculated; then, the critical acces-
sible fiber ration r
c
can be obtained through the linear
regression by plotting DP vs. r.
4. Experimental
4.1. Mater ials
Rice hulls (40 mesh size), HDPE, and maleic anhy-
dride polyethylene (MAPE) were commercial products.
4.2. Sampl e preparation
HDPE and rice hulls were compounded at 185 Cand
3200 rpm. 2% MAPE was added to improve the compat-
ibility of hydrophobic polymer and hydrophilic rice hulls
and thus improve the interfacial bonding. Discharged
dumps were hot pressed in a Wabash press at 185 C
and 30 ton for 5 min, and cooled for 20 min. The dimen-
sions of the mould are 12 cm · 12 cm · 2.4 cm. Five
sheets of different rice hull loadings, i.e., 40%, 50%,
55%, 60%, 65% by weight, were prepared.
Four blocks, 2 cm · 2 cm, were cut from each sheet.
All the samples were dried in an oven at 53 C for
24 h and then cooled down in desiccators. The edges
of these samples were sealed by unsaturated polyester.
After being cured, their dry weights were measured
before water immersion.
4.3. Experimental procedure
Samples were immersed in distilled water at room
temperature, i.e., 23 C. After specific time intervals,
samples were removed from water, their surface mois-
ture was removed by tissue paper, and their weights
(±0.0001 g) and electrical conductivities (to be intro-
duced in Section 4.5) were measured immediately.
4.4. Moisture content (M
t
)
Moisture absorption was determined by the weight
gain relative to the dry weight of the samples. The mois-
ture content of a sample was computed as follows:
M
t
¼
W
t
 W
0
W
0
 100%; ð12Þ
where W
0
and W
t
denote the dry weight of the sample
and the weight at any specific time t, respectively.
Equilibrium moisture absorption of samples, M
1
,
was assumed to be reached when the daily weight gain
of samples was less than 0.01%.
4.5. Calculation of electrical conductivity S
t
Initially the dry NFPC has no measurable electrical
conductivity. After water submersion, natural fibers
belonging to an infinite cluster absorb moisture and thus
impart electrical conductivity to composites even though
it is very low. This conductivity depends more on the
number of conduction passages rather than the ratio
of fibers belonging to the infinite 3D-network (Fig. 2).
Since the electrical resistance of samples was extre -
mely high and so could not be measured at the earlier
stage of moisture absorption, electrical current A
t
in
mA was measured whenever the weight measurement
took place. The electrical current was measured by dig-
ital multi-meter to the accuracy of 0.1 mA.
W. Wang et al. / Composites Science and Technology 66 (2006) 379–386 383

Fig. 3 shows the set-up of electrical current measure-
ment. Each of two identical thin steel plates was tightly
attached to the surfaces of rubber stoppers. One end of a
wire was welded to the steel plate and the other end
passed through the middle hole of the stopper and then
connected to the circuit. The upper stopper was fixed to
a stand, while the lowe r one was attached to an adjust-
able platform. After the sample was put on the top of
the steel plate on the lower stopper, the height of plat-
form was adjusted until the sample touched the upper
steel plate at sufficient pressur e.
The output of the adapter is 13.6 V DC. The fixed
resistance R
0
of this set-up is 27 X. The electrical con-
ductivity S
t
can be computed by following equation:
S
t
¼
A
t
V  A

t
R
0
¼
A
t
13.6  27A
t
. ð13Þ
5. Results and discussions
5.1. Moisture absorption and calculations
Fig. 4 shows the moisture absorption as a function of
ffiffi
t
p
. The values of diffusion coefficient and accessible fi-
ber ratio for each composite were calculated and are
shown in Table 1. It should be noted that the composites
with 40%, 50%, 55%, and 60% fiber loadings had not
showed the level-off which signals the equilibrium mois-
ture absorption. However, their electrical conductivities
(Fig. 5) already reached maximum, which suggest that
the moisture absorption was near to or already reached
equilibrium. Hence, the most recent moisture absorp-
tions of composites were adopted as M
1
.
In addition, in order to vilify the assumption 1,
HDPE films were made and submersed in the distilled
water at room temperature. After one year, no weight
gain was detected. This observation suggests that mois-
ture only penetrate into composites through rice hulls.
5.2. Estimate of critical accessible fiber ratio r
c
The critical accessible fiber ratio r
c
can be assessed by
plotting DP against r and extrapolating to DP = 0 as
shown in Fig. 6. The value of r
c
is 0.44.
For the composite with 65% fiber loading, the acces-
sible fiber ratio was 0.91. At this fiber loading , fibers
were highly connected. HDPE failed to encapsulate
these fibers and led as high as 91% fibers accessible to
moisture. As discus sed above, higher connectivity re-
sults in shorter flow passages and higher conductivity.
In this case, the flow passages are almost all vertical ly
running through the one side of sample to the opposite
side, rather than the tortuous ones shown in Fig. 2.
Hence, this composite reached equilibrium significantly
Fig. 3. Experimental set-up of electrical current measurement.
0
5
10
15
20
0
600 1200
SQRT(t)(Sec)
Mt (%)
40% 50%
55% 60%
65%
Fig. 4. Moisture absorption behavior of composites with different fiber
loadings.
Table 1
Values of slope of moisture absorption curve (h), diffusion coefficient
(D), diffusion-permeability coefficient (DP), and accessible fiber ratio
(r) for composites with different fiber loading
Fiber
loading (%)
h (%/s
0.5
) D (cm
2
/s · 10
9
) DP (/s
0.5
· 10
4
) r
40 0.0070 3.96 1.67 0.62
50 0.0117 4.63 2.45 0.77
55 0.0124 4.50 2.11 0.75
60 0.0142 4.67 2.18 0.78
65 0.0241 8.42 3.84 0.91
0
10
20
30
0 600 1200
SQRT(t)(sec)
Conductivity(S*10
-3
)
Conductivity(S*10
-3
)
0
1
2
3
65%-First axis
40%-Second axis
50%-Second axis
55%-Second axis
60%-Second axis
Fig. 5. Electrical conductivity of composites.
384 W. Wang et al. / Composites Science and Technology 66 (2006) 379–386

earlier than other composites with lower fiber loadings
(Fig. 4).
Fig. 7 shows the increment of elect rical conductivity
with the increasing moisture content of composite wi th
65% fiber loading. The composite started showing con-
ductivity after it absorbed approximately 50% of maxi-
mum moisture. After this, conductivity increased
quickly with further moisture absorption. The pattern
of the increment of electrical conductivity suggests a dif-
fusion pro cess of moisture absorption.
Composites with 50%, 55%, and 60% fiber loading
exhibited similar absorption behaviors and all eventu-
ally had measurable electrical conductivities (Fig. 5).
Their accessible fiber ratios were almost identical
(0.76). This observation suggests that the 3D-network
inside composites did not change much within this range
of fiber loading and that the percolation threshold may
lie just below 50% fiber loading.
However, this observation deviated significantly from
the model-estimated value 0.44. One reason causing this
discrepancy may be the over-estimate of accessible fiber
ratio. As made clear in the definition, accessible fibers in-
clude both the fibers belonging to the infinite 3D-network
and the fibers on the proximate surface of composite. This
inclusion of proximate surface fibers more greatly affects
the total accessible ratio of the composite with lower fiber
loading. For composite with low fiber loading, since the
fiber connectivity is low and moisture would not go far
into the composite, the proximate surface fibers consist
of a higher proportion of total accessible ratio.
Fig. 4 shows that all the curves (except one with 65%
fiber load ing) underwent slight change of slopes. If the
above proposal is valid, then the reason causing different
degrees of slope change is explicit. For composite with
65% fiber loading, the accessible fiber ratio is as high
as 0.91; compared to this high ratio, those proximate
surface fibers which do not belong to the infinite 3D-net-
work con sist of a minor proportion and are negligible.
Hence, there was no observable slope change during
the initial absorption period.
For composite with 40% fiber loading, no electrical
conductivity was detected. This impl ies that this fiber
loading was lower than percolation threshold and that
there was no infinite 3D-network. Here, the observed
accessible fiber ratio (0.62) was even higher than its critical
value, inconsi stent with the implication from the observa-
tion of electrical conductivity. Again, the over-estimate of
accessible fiber ratio can be applied to explain this dis-
agreement. At low fiber loading, the fiber connectivity is
low, so the proximate surface fibers not belongi ng to infi-
nite network consist of a significant proportion of the
accessible fibers. This effect decreases with increasing fiber
loading. The higher effect on composite with 40% fiber
loading resulted in greater slope change compared to
composites with 50%, 55%, and 60% fiber loadings.
It is believed that, in NFPCs, natural fibers are encap-
sulated by hydrophobic plast ics. From this research, it
can be concluded that the plastics do encapsulate the
natural fibers and prevent the moisture of composites
at low fiber loading. However, at high fiber loading,
e.g., 65%, this function decreases. Precautions are
needed to prevent moisture absorption when high fiber
loading is applied.
6. Conclusions
Through the proceeding discussion, this study led to
the following conclusions:
1. At high fiber loading when fibers are highly con-
nected, the diffusion process is the dominant mecha-
nism; while at low fiber loading close to and below
the percolation threshold, the formation of a contin-
uous network is key and hence percolation is the
dominant mechanism.
2. The model can be used to estimate the threshold
value which can be in turn used to explain moisture
absorption and electrical conduction behavior. How-
ever, it is still on the preliminary stage; further work
is needed to improve the accuracy of its prediction.
0
5
10
15
20
0 600 1200
SQRT(t)(sec)
M
t
(%)
0
10
20
30
Electrical conductivity
(S*10
-3
)
Moisture content Electrical conductivity
Fig. 7. The moisture absorption and electrical conductivity for
composite with 65% natural fiber.
y = 0.0765x - 0.0341
R
2
= 0.8612
0
0.01
0.02
0.03
0.04
0 0.5
1
Accessible fiber ratio r
DP (/sec
0.5
)
Fig. 6. Diffusion-permeability coefficient vs. accessible fiber ratio plot.
W. Wang et al. / Composites Science and Technology 66 (2006) 379–386 385

Acknowledgments
We are grateful to Natural Science and Engineering
Research Council of Canada (NSERC) for funding the
present study. We are also grateful to Composite Build-
ing Products IntÕl. Inc.
References
[1] Bledzki AK, Gassan J. Composites reinforced with cellulose
based fibers. Prog Polym Sci 1999;24:221–74.
[2] Clemons C. Wood–plastics composites in the United States: the
interfacing of two industries. J Forest Prod 2002;52(6):10–8.
[3] Morton J, Quarmley J, Rossi L. in: Natural and woodfiber
composites in North America and Europe, 7th international
conference on woodfiber–plastic composites.
[4] Stark N. Influence of moisture absorption on mechanical prop-
erties of wood flour–polypropylene composites. J Thermoplast
Compos Mater 2001;14:421–32.
[5] Marcovich NE, Reboredo MM, Aranguren MI. Dependence of
the mechanical properties of woodflour–polymer composites on
the moisture content. J Appl Sci 1998;68:2069–76.
[6] Rangaraj SV, Smith L. Effects of moisture on the durability of a
wood/thermoplastic composite. J Thermoplast Compos Mater
2000;13(March):140–61.
[7] Lu JZ, Wu Q, McNann HS. Chemical coupling in wood fiber and
polymer composites: a review of coupling substances and treat-
ment. Wood Fiber Sci 2000;32(1):88–104.
[8] Nabi Saheb D, Jog JP. Natural fiber polymer composites: a
review. Adv Polym Technol 1999;18(4):351–63.
[9] Rozman HD, Lee MH, Kumar RN, Abusanag A, Mohd. Ishak
ZA. The effect of chemical modification of rice husk with glycidyl
methacrylate on the mechanical and physical properties of rice
husk–polystyrene composites. J Wood Chem Technol 2000;20(1):
93–109.
[10] George J, Bhagawan SS, Thomas S. Effects of environment on the
properties of low-density polyethylene composites reinforced with
pineapple-leaf fiber. Compos Sci Technol 1998;58:1471–85.
[11] Sreekala MS, Kumaran MG, Thomas S. Water sorption in oil
palm fiber reinforced phenol formaldehyde composites. Compos
Part A 2002;33:763–77.
[12] Pothan LA, Thomas S. Effect of hybridization and chemical
modification on the water-absorption behavior of banana fiber-
reinforced polyester composites. J Appl Polym Sci 2004;91:
3856–65.
[13] Bao L-R, Yee AF. Effect of temperature on moisture absorption
in a bismaleimide resin and its carbon fiber composites. Polymer
2002;43:3987–97.
[14] Bao L-R, Yee AF, Lee CY-C. Moisture absorption and hygro-
thermal aging in a bismaleimide resin. Polymer 2001;42:7333–7.
[15] Marcovich NE, Reboredo MM, Aranguren MI. Moisture diffusion
in polyester–woodflour composites. Polymer 1999;40:7313–20.
[16] Sahimi M. Applications of percolation theory. London: Taylor
& Francis; 1994.
[17] Crank J. The mathematics of diffusion. Oxford: Clarendon
Press; 1975.
[18] Berkowitz Brian, Ewing Robert P. Percolation theory and
network modeling applications in soil physics. Surv Geophys
1998;19:23–72.
[19] Bonny JD, Leuenberger H. Matrix type controlled release systems
II. Percolation effects in non-swellable matrices. Pharmaceut Acta
Helv 1993;68:25–33.
[20] Leuenberger H, Bonny JD, Kolb M. Percolation effects in matrix-
type controlled drug release systems. Int J Pharmaceut
1995;115:217–24.
[21] Caraballo I, Millian M, Rabasco AM. Relationship between drug
percolation threshold and particel size in matrix tablets. Pharma-
ceut Res 1996;13(3):387–90.
[22] Millian M, Caraballo I, Rabasco AM. The role of the drug/
excipient particle size ration in the percolation model for tablets.
Pharmaceut Res 1998;15(2):216–20.
[23] Melgoza LM, Caraballo I, Alvarez-Fuentes J, Millian M,
Rabasco AM. Study of morphine hydrochloride percolation
threshold in Eudragit RS-PM matrices. Int J Pharmaceut
1998;170:169–77.
[24] Caraballo I, Melgoza LM, Alvarez-Fuentes J, Soriano MC,
Rabasco AM. Design of controlled release inert matrices of
naltrexone hydrochloride based on percolation concepts. Int J
Pharmaceut 1999;181:23–30.
[25] Melgoza LM, Rabasco AM, Sandoval H, Caraballo I. Estimation
of the percolation thresholds in dextromethorphan hydrobromide
matrices. Eur J Pharmaceut Sci 2001;12:453–9.
[26] Bunde A, Dieterich W. Percolation in composites. J Electroceram
2000;5(2):81–92.
[27] Lin B, Lee C. Percolation and dispersion of mass transport in
saturated fracture networks. Water Resour Manage 1998;12:
409–32.
[28] Stauffer D. Introduction to percolation theory. London: Taylor
& Francis; 1992.
386 W. Wang et al. / Composites Science and Technology 66 (2006) 379–386