COMPOSITES I

Lecturer: Dr. S. Li Tel: 0161-200 3842

Office: P/C11 Fax: 0161-200 4537
Date: Sept 2002 Email: [email protected]

Micromechanics of Composites
In recognition of heterogeneity, i.e. the multiphase nature of composites, micromechanics is a
study to investigate the mechanical responses under loads, such as deformation and failure of
constituent phases of composites and their interactions, at a microscopic scale. The objective
of such study is to produce understanding, measures and properties of composites, which lead
to their useful interpretations at a macroscopic scale.

Heterogeneity

1) Due to the multiphase nature, i.e. reinforcements and matrix:

For most engineering applications, reinforcements and matrix can be smeared out
resulting in an effectively homogeneous equivalent but, for understanding materials’
behaviour, this level of heterogeneity is important. This is the subject for this part of
the course

2) Due to the laminated construction:

The heterogeneity from lamina to lamina must be considered. This will be addressed
appropriately in future.

Effective elastic properties

When a composite is employed by structural designers, it is usually treated as a material with

a set of properties, in particular, the elastic properties and strength properties. These
properties are averaged measures of the material’s behaviour in a macroscopic sense. They
do not normally represent the behaviour of any of the individual constituents. For this reason,
they are called effective properties of the composite. These properties can be obtained by
conducting standard material tests on this material but this can be sometimes expensive and
time consuming. An alternative way is to get some theoretical values for them using
analytical means. A reliable theory has to be tested by experiments but once established it
can save a lot of cost and time, bearing in mind that composites are designable to a great
extent and a slight change in constituents and volume fractions may result in a different
material. In other words, there is an inexhaustible number of tests even for a fixed set of
properties. This is the justification to explore micromechanics to achieve analytical solutions
for predicting effective properties of composites from those of the constituents and their
volume fractions.

In the approaches to be described below, some assumptions are always implied unless
otherwise specified, such as the linear elastic behaviour of constituents and perfect bonding
between reinforcements and matrix.

1
Micromechanics is also useful for investigating the microscopic behaviour of composites, in
particular, the initiation and development of damage and failure, although it is an area not as
well developed as that for effective elastic properties.

Rule of mixtures

When a jar of 1 litre water at 80°C is poured into a bucket containing 3 litres of water at 20°C,
one would expect the temperature in the bucket raises to

(1 litre × 60°C + 3 litres × 20°C) ÷ 4 litres = 30°C

This is the rule of mixtures. It applies perfectly well in the above example and many other
areas. It is intuitive to expect it to provide an estimate for the effective properties of
composites, so that

P = f i Pi + f m Pm

where P is any elastic properties of the composite, Pi and Pm are its counterpart of the
inclusion (reinforcements) and the matrix, respectively, and f i and f m the volume fraction of
the two phases, respectively. Superscript i will be replaced by p for particulates and f for
fibres. However, the rule of mixtures gives reasonable predictions only occasionally.

Mechanics of materials approach for UD fibre-reinforced composites

The mechanics of materials approach as given below is based on simple mechanical

deformation and force equilibrium considerations as sketched below.

Assume ε1 =const. and no mutual transverse constraint so

that uniaxial tensile stress state is maintained.

σ 1 A = E1ε 1 A = σ 1f A f + σ 1m A m = E f ε 1 A f + E m ε 1 A m

ε 2 b = −ν 12ε 1b = −ν 12f ε 1b f − ν 12m ε 1b m

Assume σ2 =const.

σ 2l σ 2l f σ 2l m
ε 2l = = +
E2 Ef Em

Assume τ =const.

τl τlf τ lm
γl= = +
G Gf Gm

2
The results can be summarised as

 f f P f + f mPm for E1 and ν 12

 −1
P= f f fm
  f + m  for E 2 and G12
 P P 

where P represents an elastic property. Obviously, the first expression reproduces the rule of
mixtures. They show a degree of agreement with experiments. There is no obvious solution
for ν23 in this approach.

Halpin-Tsai formulae for UD fibre-reinforced composites

The Halpin-Tsai formulae as given below is based on some analytical study with some semi-
empirical approximation. Assume P as an elastic properties

Pf
 f f P f + f mPm for E1 and ν 12 −1
 P m
P= 1 + ξηf f where η = f
 Pm for E 2 , G12 and ν 23 P
 1 − ηf f +ξ
Pm

and ξ is a fiddling factor to accommodate effects of fibre geometry and packing, etc.. In most
cases, it is a value close to unity.

The Halpin-Tsai formulae usually give a better agreement with experimental results than
those obtained from the mechanics of materials approach.

Voigt’s upper bound

Consider a volume Ω of composite under deformation described by a set of uniform strains ε.

The strain energy in this volume of composite is given as

1 T
2 Ω∫
U eff = ε C ε dΩ

where C is the effective stiffness matrix of the composite. Assuming the same strain state in
both the constituent phases of the composite according to Voigt, the strain energy can be
obtained as

1 T i 1
U Voigt = ∫
2 Ωi
ε C ε dΩ + ∫ ε T C m ε dΩ
2 Ωm

where Ωi and Ωm are the volumes of the inclusion (reinforcements) and the matrix,
respectively. Superscript i will be replaced by p for particulates and f for fibres.

3
If the effectiveness is established on the basis of energy equivalence, i.e.

U Voigt = U eff

one component of the effective stiffness matrix can be determined as an approximation from
the above equation for a given strain state.

For example, if a constant uniaxial strain state in direction 1 throughout the composite is
assumed, the sum of the strain energies from both phases is

U Voigt =
1
∫
2 Ωi
1 1
(
ε 1C11i ε 1 dΩ + ∫ ε 1C11m ε 1dΩ = ε 12 f i C11i + f m C11m
2 Ωm 2
)

If the properties of the composite can be described in an effective manner, the effective strain
energy can also be given as

1 1
U eff =
2Ω∫ ε 1C11ε 1 dΩ = ε 12 C11
2
one obtains

C11Voigt = f i C11i + f m C11m

It reproduces the rule of mixtures but for the components of the stiffness matrix of the
composite. To be rigorous, the effective engineering elastic properties, such as Young’s
moduli, Poisson ratios and shear moduli, should be obtained from the compliance matrix of
the composite S=C-1, e.g. E1=1/S11. In general, C11≠1/S11 and hence C11 in the above equation
cannot be simply replaced be E1. This can only be done as an approximation when the
difference in the Poisson ratios between the constituent phases of the composite is neglected.

Minimum total potential energy principle: Among all compatible sets of strains, one
minimises the total potential energy if the stresses obtained from this set of strains according
to the constitutive relationship of the material satisfy equilibrium conditions, and vice versa.

If it is loaded up in terms of displacements prescribed to the surface of the volume of

composite resulting in a uniform strain state, which are obviously compatible, the total
potential energy is the same as the strain energy. The effective properties obtained gives an
upper bound identically. As a uniform set of strains do not necessarily satisfy stress
equilibrium requirement, the energy corresponding to UVoigt is always greater or equal to the
effective one Ueff corresponding to an internally balanced stress distribution, i.e.

U eff ≤ U voigt hence C11 ≤ C11Voigt = f i C11i + f m C11m

To conclude, the Voigt prediction always gives an upper bound. How close the bound is to
the exact solution depends on how close the internal strain state is to a uniformly strained
state. This is why the rule of mixtures sometimes gives a reasonable result but sometime not.

4
Reuss’ lower bound

Consider a volume Ω of composite under deformation described by a set of uniform stresses

σ. The complementary strain energy in this volume of composite is given as

1
U ceff =
2Ω∫ σ T Sσ dΩ

where S is the effective compliance matrix of the composite. Assuming the same stress state
in both the constituent phases of the composite according to Reuss, the complementary strain
energy can be obtained as

1 1
U cRe uss = ∫
2 Ωi
σ T S iσ dΩ + ∫ σ T S mσ dΩ
2 Ωm

If the effectiveness is established on the basis of energy equivalence, i.e.

U cRe uss = U ceff

one component of the effective compliance matrix can be determined as an approximation

from the above equation for a given stress state.

For example, if a constant uniaxial stress state in direction 2 throughout the composite is
assumed, the sum of the complementary strain energies from both phases is

σ 2 S 22i σ 2 dΩ + ∫ σ 2 S 22m σ 2 dΩ = σ 22 ( f i S 22i + f m S 22m )

1 1 1
U cRe uss = ∫
2 Ωi 2 Ωm 2

If the properties of the composite can be described in an effective manner, the effective
complementary strain energy can also be given as

1 1
U ceff =
2Ω∫ σ 2 S 22σ 2 dΩ = σ 22 S 22
2
one obtains

1 fi fm
S 22Re uss = f i S 22
i
+ f m S 22m i.e. = i + m
E2 E2 E2

It reproduces part of result from the mechanics of materials approach. As this approach deals
with compliances, the effective engineering elastic properties can be obtained straightaway.

Minimum total complementary potential energy principle: Among all sets of stresses in
equilibrium, one minimises the total complementary potential energy if the strains obtained
from this set of stresses according to the constitutive relationship of the material are
compatible, and vice versa.

5
If it is loaded up in terms of tractions prescribed to the surface of the volume of composite
resulting in a uniform stress state, which obviously satisfies the equilibrium conditions, the
total complementary potential energy is the same as the complementary strain energy. The
effective properties obtained give a lower bound identically. As a uniform set of stresses do
not necessarily satisfy strain compatibility requirements, the energy corresponding to UcReuss is
always less or equal to the effective one Uceff corresponding to an internally compatible strain
distribution, i.e.
−1
 fi fm
U eff
c ≤U Re uss
c hence E2 ≥ E Re uss
2 =  i + m 
 E2 E2 

To conclude, the Reuss prediction always gives a lower bound. How close the bound is to the
exact solution depends on how close the internal stress state is to a uniform stress state.

Usually, the upper and lower bounds as presented above are substantially away from the exact
results except a few exceptional cases. Better predictions have always been an area of
research. As a result, many improved alternatives have been established. They are based on
different considerations. Some typical results are presented below but details of derivations
will be omitted. Interested readers are encouraged to explore in the literature.

Better bounds have been obtained by many others, among whom Hashin and Shtrikman
(1963) is one of the theories most frequently referred to in the literature. Specifically for UD
composites, Hashin-Rosen bounds (1964) tend to give good results but in much more
complicated expressions.

Apart from various bound theories, there have been many established approaches, e.g.
Eshelby’s equivalent inclusion method (1957) and Budiansky’s self-consistent method
(1965). Many assumed ellipsoidal inclusions and hence are relevant to particulate or short
fibre reinforced composites.

To a great extent, an accurate prediction of the effective material properties relies on an

accurate estimate of the microscopic distributions of stresses and strains in the composite. In
any theory, these distributions can only be obtained as accurate as to the assumptions of the
theory adopted, e.g. elastic constituents, perfect bonding and regular packing, etc.. As a
versatile means of stress and strain analysis, finite elements are readily applicable for
micromechanical analysis. As a result, an FE approach for predicting effective properties of
composites is open to users as one of the options, usually not the first one.

6
Examples:

The constituents’ properties of UD composite (E-glass / epoxy) are given as

E f = 74.0 GPa , and < f = 0.20,

and E m = 3.35 GPa , and < m = 0.35.

Predicted results are compared with experimental data in the following table.

60% VF Halpin-Tsai (ξ=1) Hashin & Rosen F.E. Exp. data

E1 45.75 (GPa) 45.85 (GPa) 45.76 (GPa) 45.6 (GPa)
{45.77}
E2 11.47 (GPa) 13.11(GPa) + 11.80 (GPa) 16.2 (GPa)
9.740 (GPa) - {14.91/9.854}
G12 4.318 (GPa) 4.318 (GPa) 4.339 (GPa) 5.83 (GPa)
{4.577}
ν12 0.2600 0.2517 0.2515 0.278
{0.2497}
ν23 0.2516 0.3346 - 0.4020 0.400
0.5055 + {0.2649/0.6182}
G23 4.582 (GPa) * 4.910(GPa) + 4.208 (GPa) 6.338 (GPa) *
3.235 (GPa) - {3.254/5.894}
NOTE :
1. * values are worked out assuming transverse isotropy.
2. ± upper and lower bounds, respectively.
3. {A/B} values from a square unit cell, A is obtained in 0°/90° direction and B in 45° direction.

The constituents’ properties of a glass particulate reinforced epoxy composite with 30% of
particulate volume fraction are given as

E p = 76.0 GPa , and < p = 0.230,

and E m = 3.01 GPa , and < m = 0.394.

Predicted results are compared with experimental data in the following table.

30% VF E1 (GPa) E2 (GPa) G23 (GPa) G12 (GPa) <12

SC# 6.822 1.839 0.3354
Finite #
BCC 5.335 2.150 0.3736
Element
Prediction FCC# 5.150 2.052 0.3729
CPH# 5.810 5.681 2.067 2.003 0.3601
Reuss’lower bound 4.228 1.596 0.325
Voigt’s upper bound 24.91 9.26 0.345
(Rule of mixtures)
Exp. data (30.25%VF) 6.01 2.19 0.375
NOTE :
# Simple Cubic, Body Centred Cubic, Face Centred Cubic and Close Packed Hexagonal packing of spherical particles, respectively

7
Failure Mechanisms in Composites

Composites are all multiphased. The mechanisms of failure in one phase are usually different
from those of another. Further complications arise from the existence of interfaces between
different phases. These considerations apply to UD fibre reinforced composites, short fibre
reinforced composites as well as particulate reinforced composites, although the actual
mechanisms in each of them could be rather different. Only UD fibre reinforced composites
will be discussed below. Interested readers are encouraged to acquire information on other
types of composites through vast volume of publications in the literature.

For unidirectionally fibre reinforced composites, there are several typical modes of failure.
Their microscopic descriptions are provided below. Both the fibre and matrix can be brittle as
well as ductile depending on the constituents selected.

Tensile fracture of single fibres

The tensile fracture of four typical fibres are shown in the figure below. A flat fracture
surface indicates a brittle behaviour while a point-ended fracture results from a necking
process which is a typical plastic behaviour. Although the ductility of matrix material can be
obtained from bulk properties of the material, it can be affected to a certain extent by the
existence of different phase, fibres, embedded in it.

Failure of composites due to tension along fibre direction

The sequence of microscopic failure in a composite can be one of the following

8
1) matrix cracks/yields followed by fibre fracture/yield
2) fibre fractures/yields followed by matrix cracking/yield.

These two possibilities have been depicted in the sketch below, where brittle behaviour is
assumed for matrix and fibre respectively and, as a result, they crack instead of yield in a
plastic manner.

Failure due to longitudinal tension is by large the most important mode of failure as fibre
fracture usually indicates the failure of the composite. When the composite is used in a load
bearing structure, it is often followed by an imminent, catastrophic failure of the structure. As
it happens, there are usually two types of behaviour, one involves fibre pull out and one does
not, as shown in the figures below for the fracture surfaces. This depends on the relative
strength of the interface to that of the fibres.

9
Compression along fibres

Failure of UD composites under compression along fibres is a structural problem but at a

microscopic scale. Usually, fibres tend to buckle in a manner similar to beam on foundation.
The photoelastic pictures below taken for single fibres embedded in matrix show the sign of
the sinusoidal buckling mode while the deflections are still too small to observe. As the
buckling progresses, especially in presence of small misalignment, the deflection in the fibres
tends to localise. This triggers a global failure mechanism commonly referred to as kinking
band. An example of such failure is illustrated in the figure below. A close examination of
the fracture surface reveals the typical bending fracture characteristics of fibres at the kinking
band.

10
Tension or compression transverse to fibres

Failure of composites under transverse tension and compression is a matrix dominated

behaviour. As the composite is isotropic in the plane transverse to the fibres, the failure
mechanism is similar to conventional materials in this plane, the weakest perspective of the
composite. If the matrix is ductile, the composite yields and if the matrix is brittle, the
composite tends to crack perpendicular to the load under tension but at 45° to the load under
compression.

Tension perpendicular to the crack

Shear along fibres

Shear along the fibres is one of the most peculiar aspects of UD composites. Usually, before
failure takes place, the composite has already shown a high degree of nonlinear behaviour.
The nonlinearity results from a mixture of many factors, e.g., plasticity, viscosity and
microcracking. A typical shear stress strain curve is shown in the figure below where unload
hysteresis and permanent strain are observed.

11
 70

60

50

Shear stress (MPa)

40

30

20

10

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Shear strain (%)

Upon fracture failure, the fracture surfaces often show hackles as a most characteristic feature
as illustrated in the photograph below..

Shear transverse to fibres

If the transverse shear stress is represented tensile and compressive stresses at 45°, one often
find the behaviour of the composite resembles that under transverse tension or compression
but rotated by an angle of 45°.

Interfacial failure

In most of the failure mechanisms mentioned above, failure of interfaces between the fibres
and matrix gets involved at some stage to a certain extent. Interfaces are usually of different

12
strength from either constituent. Coatings are sometimes employed to alter the interface
strength to suit particular applications. The existence of interfaces complicates the failure
mechanisms to a great extent by providing alternative ways of failure/damage propagation.
The sketch below gives an example of this kind. Modelling interfacial cracks has been a
difficult task even for idealised situations at a macroscopic scale.

Concluding Remarks

The developments of micromechanics of composites have resulted in reasonable success in

prediction of elastic properties from those of their constituents. These properties are
relatively insensitive to defects in the composite. For relatively defect-sensitive properties,
such as strengths, while micromechanics has produced a degree of insight into the behaviour
of these materials, the success so far is far less impressive. Great efforts are still being made
in this respect in research of composites.

13
Tutorial Questions:

1) A part of the outcome from the mechanics of materials approach reproduces the result
of Voigt’s bound while the other part does the Reuss’ bound. Using the mechanics of
materials approach and from the viewpoint of the minimum total potential and
complementary potential energy principles, assess the accuracy of all the predictions
from Voigt and Reuss’ bounds and explain why agreements and disagreements arise.

2) Most effective properties of a composite fall between their counterparts of the

constituents, which would suggest a weighted rule of mixtures produce good results.
However, ν23 of UD composites seems to behave out of the step as described above, as
can be seen through the data shown in the examples in the notes. Describe the
mechanism which is responsible for that.

3) Select any three failure mechanisms as describe in the notes, compare them with their
counterparts in homogenous, isotropic materials of similar ductility. Comment on the
similarities and dissimilarities.

14