JOURNAL OF MATERIALS SCIENCE 22 ( 1 9 8 7 ) 1733 1736

Computer model for the elastic properties of

short fibre and particulate filled polymers
YVES T E R M O N I A
Central Research and Development Experimental Station, El du Pont de Nemours Inc.,
Wilmington, Delaware 19898, USA

A finite difference type of approach is proposed for the study of the elastic properties of fibre
and particulate reinforced composites. The effects of the fibre modulus Ef and aspect ratio
(l/d) on the composite modulus are studied. In that respect, the ratio (Ef/Em)/(I/d), where Em
is the matrix modulus, is shown to be an important index of short fibre composites. Our results
f o r the dependence of the composite modulus on El, lid and volume fraction of the fibres are
compared to available experimental data on foams and particulate filled polymers. A good
agreement with experiment is found, which is much better than that obtained with the help of
the widely used semi-empirical Halpin-Tsai equation.

1. I n t r o d u c t i o n and concentrations of the constitutive components

There has been recently a growing interest in fibre [1, 2]. These relationships lead to similar predictions
reinforced composites due to their great versatility for fibre volume fractions less than 20 to 30%. At
and high performance. These materials often consist higher volume fraction loadings, the results diverge
of discontinuous stiff fibres embedded in a soft matrix and generally show p o o r agreement with experimental
with the fibre axes oriented in the direction of the data because, e.g. of the neglect of fibre-fibre interac-
applied load. The physical and mechanical properties tions. All these approaches, moreover, are only global
of these composites are, however, not well under- in nature and are unable to provide a detailed rep-
stood. The variation in fibre length as well as the resentation of the stress concentration near fibre ends.
interactions between the fibre ends and the matrix In a previous paper [3], we have developed a finite
are the key factors contributing to the complexity of difference type of approach for the study of the stress
the problem. The present paper addresses a basic transfer in fibre-reinforced composites. In that
question: the effective moduli of these composites and approach, the composite is represented by a regular
their dependence on volume fraction, stiffness and three-dimensional lattice whose nearest neighbour
aspect ratio of the fibres. nodal points are linked by bonds having different
Several constitutive relationships or combining elastic constants for the fibre and for the matrix. For
rules have been proposed for predicting the properties a given external strain, these nodes are relaxed
of reinforced composites in terms of the properties towards local mechanical equilibrium with their

N N @
@ @

Y @ @ @

>
(a} x (b} x

Figure 1 Two-dimensionalrepresentation of the lattice model in a plane paraIM (Fig. l a) and perpendicular (Fig. l b), respectively,to the
direction of loading (y-direction). The lattice comprises 120 nodes along the y-axis and 39 in the x- and z-directions. The parameters d and
I denote the fibre diameter and length, respectively.In the simulations, d was set equal to 3 to 7 lattice units. The spacing between fibres in
orthogonal planes does not have to be the same but, in a given plane, that spacing is kept constant. The bonds between nodes are assigned
different elastic constants for the matrix and for the fibre.

0022-2461/87 $03.00 + .12 © 1987 Chapman and Hall Ltd. 1 733

 20 l l r [

16

12 /
//,'7/
/ / I .'

~oj;" o/,,'"
/,,,,);.,.
1 I I I
0 0.2 0.4 0.6 0.8 1.C
h
Figure 3 Dependence of the composite modulus Ec (in units of Era)
on the volume fraction Vf of the fibres for a ratio Ef/E m = 20. The
curves are for different values of the fibre aspect ratio lid. The fibre
diameter d was set to values ranging from 3 lattice units (at
Vf = 0.2) to 9 units (at Vf = 0.7). We took vm = vr = 0.3. The
dashed line gives the Halpin Tsai [4] prediction.

mental data on the stiffness of foams and particulate-

reinforced composites. We find a much better agree-
ment with experiment than that obtained using the
Halpin-Tsai equation. Application of our model to a
detailed study of the stress concentration around the
fibre ends, and of the effect of fibre-matrix adhesion
on the composite modulus, will be presented in a
forthcoming publication.

2. M o d e l
We start by describing our three-dimensional simple
cubic model for a unidirectional multi-fibre composite.
A two-dimensional representation (Figs l a and b) is
given of the lattice model in a plane parallel (Fig. 1a)
y
1.0 I I I I

~E' 0.8 /
X

Figure 2 Two-dimensional representation of the deformation ~- 0.6 A

scheme in the x - y plane for a fibre reinforced composite witb
lid = 2, Er/E m = 20. The fibre volume fraction Vf was set equal tc
0.3. The x and y scales have been distorted so that the stres, ~ -i (El / Ern) / (l/d)
distribution can be easily seen. ~E 0.4

neighbours, by a systematic sequence of operations, 10 I

which steadily reduces the net residual force acting on
each node.
That approach [3], applied to single-fibre reinforced O' I

composites, is now extended to multi-fibre composites. 0 0.2 0.4 0.6 0.8 1.0
The dependence of the composite modulus on the vf
aspect ratio and volume fraction of the fibres is Figure 4 Composition dependence of the modulus E c (suitably
normalized) for different values of the ratio (Ef/E,,)/(l/d). Values
studied and the predictions are compared to those of
for the pairs (El~Era, l/d) are as follows: (o) (20, 10); (D) (6.7, 3.3);
the widely used semi-empirical Halpin-Tsai equation (zx) (60, 30); ( e ) (20, 2); ( I ) (60, 6); (A) (I0, 1). We took
[411 Our results are also compared to available experi- vm = vf = 0.3.

1734
 i i f
force acting on each node. For simplicity, we assume
5 t that the displacements of the nodes along the co-
/ ordinates axes are mutually independent and we focus
on motions along the y-axis. Since the shearing mode

/
along the y-axis predominates when it comes to load
transfer between the mamx and the fibres, lateral
o motions in the x- and z-directions should be of
secondary importance for the present study [3-5]. The
3
validity of our assumption has also been discussed

~,
/ I_
//

a
O

l .... ~ I
[3].
Fig. 2 gives a two-dimensional representation of
the deformation scheme in the x - y plane for a fibre
reinforced composite with l / d = 2, Er/E m = 20. The
fibre volume fraction Vfwas set equal to 0.3. The x and
0 (3.2 0.4 0.6 0.8 1,0 y scales have been distorted so that the stress distri-
vf bution can be easily seen. The figure shows a high
Figure 5 Composition dependence of Ec/Em for glass spheres in degree of shearing in the matrix between nearest
epoxy. O: experiments [7]; ( - O - ) present model. Values for the neighbour fibres along the transverse x-axis, indicative
parameters are as follows: EllEn, = 25, vf = 0.23, vm = 0.39. The of the importance of the shearing mode in the load
dashed line indicates the Halpin-Tsai prediction.
transfer process between adjacent fibres.

and perpendicular (Fig. lb), respectively, to the direc- 3. Results and discussion
tion of loading (y-direction). The lattice comprises The dependence of the composite modulus Ec (in units
120 nodes along the y-axis and 39 in the x- and z-direc- of E~) on the volume fraction Vf.of the fibre for a ratio
tions. The parameters d and l denote the fibre diameter Et./Em ~- 20 is shown in Fig. 3. The curves are for
and length, respectively. In the simulations, d was set different values of the fibre aspect ratio l/d. For con-
equal to 3 to 7 lattice units. The spacing between fibres tinuous fibres (lid = oo), the dependence is linear and
in orthogonat planes does not have to be the same but, Eo reaches its maximum value given by the so-called
in a given plane, that spacing is kept constant. The Voigt average E~ = EfVf + E~(1 - V0. For discon-
bonds between nodes are assigned different elastic tinuous fibres, the curves show a negative deviation
constants for the matrix and for the fibre. The Young from the ideal continuous case and the modulus
and shear moduli of the matrix and of the fibre are decreases with a decrease in the aspect ratio. The
denoted by (E~, Gm) and (El, Gf), respectively. For results in Fig. 3 have also been compared to the
simplicity, we assume both components to be isotropic predictions of the widely used semi-empirical Halpin-
so that G m = Em/2(1 + Vm) and Gf = Ef/2(1 + vr), Tsai equation [4]. Our values are seen to be lower for
where v denotes a Poisson ratio. Vf < 0.5 and higher for Vr > 0.5, the discrepancy
The composite lattice is strained by a constant between the two sets of predictions being larger for
amount (1%) along the y-axis. In order to estimate the lower values of the aspect ratio.
relative displacements of the fibre and the matrix, each The results presented above thus indicate that, for a
node is relaxed towards local mechanical equilibrium given value of Vf, the composite modulus increases
with its neighbours. This is done with the help of a with an increase in the fibre aspect ratio l/d and in the
sequence of fast computer algorithms, described pre- fibre modulus Ef (for given Era). It seems therefore
viously [3], which steadily reduces the net residual natural to inquire whether these two fibre character-
istics can be unified in any way. Fig. 4 depicts the
composition dependence of E~ (suitably normalized) for
0 // two different values of the ratio (Ef/E~)/(l/d). Although

), I
/ Ef/E~ values within a given ratio vary by a factor as
large as 10, all our results for a given (Er/Em)/(l/d) fall
on the same curve. That ratio [6] thus seems to be an
]/' important index of fibre reinforced composites: the

/
higher the ratio, the higher the efficiency of load trans-
fer between the matrix and the fibre.
The predictions of our model to experimental data
on several particulate (lid = 1) filled polymers are com-
pared in Figs 5 to 7: glass spheres in thermosetting resins
2i [7, 8] and natural silica particles in epoxy [9], Our results
are in good quantitative agreement with experiment.
The Halpin-Tsai equation [4], on the other hand, is
1 ~ . 1 J. I _ , _ I , seen to systematically underestimate the composite
0 0.2 0.4 0.6 0.8 ' --1.0
modulus at high volume fractions of the filler.
h
The results presented above were for filler particles
Figure 6 As Fig. 5, but for glass fibres in polyester resir~. Experi-
mental data [8]. Ef/E~ = 42, v~ = 0.21, v~ ~ 0.44.
which are stiffer than the surrounding matrix. We now
turn to application to materials, like certain polyblends
1735
 I I I I I I I I
1.0 - , , , , , , , ---r---7%ti

,,p
• #
0.8 ,,y%

~o 06 - / / /JtU •
/ /-
~4

~" 0.4 s S
::x<
st
I "Jr t .
0.2 ."
t"/ ~/~AI
, -

~S
SS~ A
~&
] " I ! 1 I I I I I I 0 I I I , I I__
0 0.2 0.4 0.6 0.8 .0 0 0.2 0.4 0.6 0.8 1.0
vf density rotio

Figure 8 Dependence o f the modulus on the density ratio for

Figure 7 As Fig. 5, but for silica particles in epoxy. Experimental thermoplastic foams. ( - O -) present model. The other symbols are
data [9]. Ef/Em = 35, vf = 0.25, vm = 0.4. for experimental data: (1:3) [10]; (O) [11]; (A) [12]; (zx) [13]. The
dashed line indicates the Halpin-Tsai prediction.

and polymeric foams, in which the continuous phase is

more rigid than the included phase. Fig. 8 compares our References

results for the modulus to experimental data on several h S, McGEE and R. L. M c C U L L O U G H , Polymer Comp. 2
(1981) 149.
thermoplastic foams [10-13]. We took lid = 1 for the
2. A. A. BERLIN, S . A . VOLFSON, N . S . ENIKOLO-
included phase. The results here follow the often conjec- PIAN and S. S. NEGMATOV, "Principles o f Polymer
tured [1, 13] density-squared relationship. Again, the Composites" (Springer Verlag, Berlin, 1986).
agreement with experiment is excellent and far better 3. Y. TERMONIA, J. Mater. Sci. 22 (1987) 504.
than that obtained with the help of the widely used 4, J. C. HALPIN and J. L. KARDOS, Pol. Eng. Sci. 16
(1976) 344.
Halpin-Tsai equation [4]. We have also studied the
5, R. A. LARDER and C. W, BEADLE, J. Compos. Mater.
dependence of our results of Fig. 8 on the distribution 10 (1976) 21.
and size of the included phase. As far as we could 6. H. F U K U D A and K. KAWATA, Fibre Sci. Technol. 7
determine, the modulus of the foam seems to be a (1974) 207.
unique function of the average density ratio. 7. J. C, SMITH, J. Res. Nat. Bur. Stand. 80A (1976) 45.
8. T. G. RICHARD, J. Comp. Mat. 9 (1975) 108.
To conclude, we have presented a finite-difference 9. O. ISHAI and L. J, COHEN, [nt. J. Mech. Sci. 9 (1967)
type of approach for the study of the elastic properties 539.
of short fibre and particulate-filled polymers. Our 10. S. Y. HOBBS, J. Cell. Plast. 12 (1976) 258.
results are in better quantitative agreement with experi- 11. S. BAXTER and T. T. JONES, Plast. Polym. 40 (1972)
ment than those obtained with the help of the semi- 69.
12. D. R. MOORE, H. K, COUZENS and M . J . IRE-
empirical Halpin-Tsai equation. Another major advan- MONGER, J. Cell. Plast. 10 (1974) 135.
tage over previous approaches [1, 4] is that the present 13. R. C. P R O G E L H O F and J. L. THRONE, Poly. Eng. Sci.
model is microscopic in nature and allows a study of the 19 (1979) 493,
effect of the fibre size distribution and of the fibre- 14. Y. TERMONIA, P. M E A K I N and P. SMITH, Macro-
matrix adhesion [3] on the composite modulus. The molecules 18 (1985) 2246.
15. Idem, ibid. 19 (1986) 154.
model also permits an easy evaluation of the stress
distribution inside the material and therefore lends itself
to a detailed study of the fracture process itself [14, 15].
These studies will be the object of a forthcoming Received 20 June
publication. and accepted 9 September 1986

1736