Materials Science and Engineering A 393 (2005) 170–178

On the mechanical properties of alumina–epoxy composites with an

interpenetrating network structure
M.T. Tilbrook, R.J. Moon, M. Hoffman∗
School of Materials Science and Engineering, University of New South Wales, Sydney, NSW 2052, Australia

Received 17 June 2004; received in revised form 4 October 2004; accepted 4 October 2004

Abstract

Methods of predicting effective mechanical properties of composites with an interpenetrating network structure are currently not well
understood, particularly for cases in which the constituent materials have widely differing properties. Alumina–epoxy composites with an
interpenetrating composite structure have been produced via an infiltration process and the elastic properties were measured via the impulse
excitation technique. A strong dependence of properties on composition and processing was observed. Properties were compared with several
mixing law predictions made using compositional data obtained from microstructural analysis. The effective medium approximation (EMA)
was shown to predict properties adequately, whilst others models proved inappropriate.
© 2004 Elsevier B.V. All rights reserved.

Keywords: Interpenetrating network composite; Effective properties; Impulse excitation technique; Effective medium approximation

1. Introduction erated [1–5], whilst work on interpenetrating structures has

Composites potentially represent an advance in optimi-
sation of material performance, by offering a combination
of properties of several different constituent materials.
Underpinning composite design is an understanding of the
relation between mechanical properties of the constituent
materials and the effective mechanical properties of the
resulting composite material. For a particular composite,
this depends on the internal geometry and may be derived
theoretically [1–5], though experimental verification of
effective properties is crucial to accurately relate material
composition and properties.
Composites with an interpenetrating network structure are
of particular interest as they have been found to demonstrate
higher strength, toughness and wear resistance when com-
pared to other composite structures [6,7]. Traditionally, com-
posite materials tended to display a fibre/matrix, layered,
laminar or particle/matrix type structure. Correspondingly,
theoretical models based upon such geometries have prolif-
∗ Corresponding author. Tel.: +61 2 9385 4223; fax: +61 2 9385 5956.
E-mail address: mark.hoffman@unsw.edu.au (M. Hoffman).
been more limited [8].
This issue is of particular interest for functionally graded
materials (FGMs), which incorporate a spatial variation
of composition and properties [9–11], where experimental
verification of local mechanical properties is difficult.
Furthermore, FGMs often possess an interpenetrating
network structure [12]. Recent interest in graded materials
has motivated the development of semi-empirical models,
based on micromechanics and image analysis [13–15].
Several generic models may potentially be applica-
ble to interpenetrating network-structured composites, in-
cluding the isostress- and isostrain-approximation models,
Hashin–Shtrikman bounds, and the effective medium approx-
imation (EMA). In addition, several models have been pro-
posed for composites with two continuous phases with simple
regular internal geometries [8,16]. For a more rigorous treat-
ment of mixing laws, see the work of Torquato et al. [17],
Reiter et al. [11] or Suresh and Mortensen [10].
The simple isostress and isostrain relations provide a first
approximation of the bounds for effective Young’s modulus
[17]. These are:
E∗ = v1 E1 + v2 E2 (1)

0921-5093/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.msea.2004.10.004
 M.T. Tilbrook et al. / Materials Science and Engineering A 393 (2005) 170–178

−1
∗ v1 v2
E = + (2)
E1 E2
where E is the Young’s modulus, (*) represents the effective
composite property, v the volume fraction and subscripts 1
and 2 denote each phase. Replacing Young’s modulus, E, in
these equations with bulk and shear moduli, K and G, leads
to the Voigt and Reuss model expressions.
The Hashin–Shtrikman bounds [18] were used by
Torquato et al. to successfully predict the mechanical proper-
ties of interpenetrating boron carbide–aluminium composites
[17]. The effective medium approximation (EMA) [4] was
used by Hoffman et al. to predict mechanical properties of
aluminium–alumina-interpenetrating composites [19], al-
though the validity of this was not confirmed experimentally.
Tuchinskii developed a simplified unit-cell model for com-
posites with two continuous phases [16]. This was observed
by Jedamzik et al. [20] to fit experimental results for interpen-
etrating structured tungsten–copper composites. A similar
approach was applied by Feng et al. [21] to interpenetrating
composites with more than two continuous phases.
This paper details the investigation of effective mechan-
ical properties of alumina–epoxy composites with an inter-
penetrating network structure. Experimental results obtained
via the impulse excitation technique are presented and sev-
eral theoretical models are applied to this composite system.
The alumina–epoxy composite system has been adopted as a
model system due to the significantly differing elastic proper-
ties of the two phases. This enables a large stiffness variation
to be attained across the composition range.
The large difference in properties of the two constituents
leads to widely varying predictions from different mod-
els. Considering for instance, a composite of 50% alumina
(E = 400 GPa) and 50% epoxy (E = 4 GPa) [22], the predic-
tions from Eqs. (1) and (2) are 202 and 7.9 GPa, respectively.
These are clearly extremely disparate and a more precise pre-
diction is required. For a 50/50 composite of alumina and alu-
minium (E = 79 GPa), the predicted values, 240 and 132 GPa,
respectively, differ less drastically though still significantly.
An advantage of investigating an alumina–epoxy compos-
ite system is that the increased variation, between different
model predictions, enables improved identification of the ap-
propriate choice of model for composites with interpenetrat-
ing network structure.
2. Experimental techniques
2.1. Sample processing and preparation
Alumina–epoxy FGMs were produced by infiltration of
epoxy into an open porosity alumina preform. Samples were
produced in the composition range 5–50% epoxy, with an
interpenetrating network structure in which both phases are
continuous.
171
The processing was conducted similarly to the method de-
scribed by Cichocki et al. [23], which has been used to pro-
duce aluminium–alumina-interpenetrating composites [12].
Open-celled polyurethane (PU) foam (Bulpren S-31048, Eu-
rofoam, Troisdorf, Germany, 90 pores/in.) was used as an im-
print for the interpenetrating network structure. It was com-
pressed from an initial density of 2.5% to obtain the required
volume fractions of foam and air.
A colloidal suspension of alumina particles (99.99%
Al2 O3 , Taimicron TM-DAR, Taimei Chemicals Co. Ltd.,
Japan) was infiltrated into the PU foam under varying pres-
sure, cast and allowed to dry. The foam was then pyrolysed at
800 ◦ C, leaving a ceramic green-body (approximate dimen-
sions: 50 mm × 30 mm × 6 mm) with a network of intercon-
necting pores. The ceramic was sintered at 1500 ◦ C for 1 h
leading to ∼10% shrinkage in each dimension. Epoxy resin
(Epofix, Struers, Germany) was infiltrated into the ceramic
under varying pressure then cured at room temperature. Spec-
imens were then ground to size (50 mm × 30 mm × 5 mm)
with a 600 grit diamond-grinding wheel, and polished with
an automatic polisher and diamond paste down to a diamond
particle size of 1 ␮m in the final step. The plate specimens
were later cut into small beams (50 mm × 5 mm × 4 mm ap-
prox.) with a low speed diamond saw for further testing. This
process results in two continuous interpenetrating phases; the
interconnecting porosity of the foam perform is mimicked by
the alumina, which is a robust structure after firing, while the
complete penetration of the epoxy with no discernable poros-
ity in the composite confirms the complete interpenetration
of this phase.
Several samples of an alumina particulate/epoxy ma-
trix composite in the composition range 75–90% were
also produced, for comparison with the interpenetrating-
network composites. Alumina powder (Taimicron TM-DAR,
as above) was mixed with epoxy resin (Epofix) by stirring be-
fore curing. Due to the viscosity of the mixture, air bubbles
introduced during stirring were difficult to remove, which
lead to some porosity in these samples.
Monolithic alumina and epoxy specimens were also pro-
duced for comparison. Alumina specimens were produced
from a suspension of alumina particles (Taimicron TM-
DAR), which was slip-cast, dried and sintered at 1500 ◦ C
for 1 h. Epoxy specimens were produced from resin (Epofix)
cured at room temperature. Grinding and cutting was con-
ducted similarly to the composite specimens.
2.2. Microstructural analysis
Optical microscopy on the polished surfaces was used to
characterise the alumina and epoxy phases and to quantify
the porosity. Images were obtained using a Nikon 200 micro-
scope and digital camera, and then processed using Adobe
Photoshop and Image Processing ToolKit (Version 3.0) [24].
Examples of microstructures for a range of compositions,
for both the interpenetrating-network and powder compos-
ites, are shown in Fig. 1. The continuity of each phase in the
172 M.T. Tilbrook et al. / Materials Science and Engineering A 393 (2005) 170–178

Fig. 1. Microstructural images of alumina epoxy composite samples. Interpenetrating network structures: (A) 5% epoxy; (B) 15% epoxy; (C) 30% epoxy; (D)
45% epoxy. Alumina appears as the light phase, epoxy is the grey phase and the darker spots are pores. Powder composites: (E) 75% epoxy and (F) 90% epoxy.
Alumina appears as the grey phase, epoxy is the light phase and the darker spots are pores.

interpenetrating microstructures cannot be observed in two-
dimensional micrographs. Fig. 1(a–d) show the interpene-
trating network microstructures with the differences between
(a–b) and (c–d) attributable to the buckling of foam ligaments
during compression which resulted in increasing microstruc-
tural irregularity for higher foam compression ratios in the
resultant alumina phases (light regions). Fig. 1(e–f) shows
a different morphology with some alumina particle contact
(dark regions).
Fig. 2 displays an original greyscale microscope image
and the corresponding black and white image obtained via
a digital thresholding process, which was used for volume
fraction estimation. The samples tested varied in composi-
tion across the range 5–50% epoxy for the interpenetrating
structured composites. For samples containing pores, mea-
surement of porosity volume fraction was incorporated into
the compositional analysis, by a second thresholding pro-
cess. Although a small amount of porosity was observed in
most samples, this was generally <2% and at most 5%. Error
margins for composition measurements were estimated via
perturbation during the thresholding step and from variation
in compositions observed across single samples.
After cutting and polishing, the densities were determined
from mass and dimensional measurements. Composite densi-
ties were compared with those predicted using compositions
obtained from image analysis and measured densities of con-
stituent phases, as shown in Fig. 3. The observed disparity
between the two density measurements, particularly for more
 M.T. Tilbrook et al. / Materials Science and Engineering A 393 (2005) 170–178
Fig. 2. Microstructural image for 20% epoxy sample: (A) grey-scale image and (B) black and white image after thresholding.
Fig. 3. Comparison of density values determined from dimensional
measurements and those calculated from microstructural data, for
interpenetrating-network composites.
epoxy-rich samples, is most probably due to a higher volume
fraction of epoxy and pores near the surface where image
analysis was undertaken. Several samples were progressively
ground down, polished and microanalysed at several depths,
and slightly higher alumina volume fractions were observed
internally. For example, the density for the sample containing
∼40% epoxy was calculated, from surface microstructure, as
2.55, and from internal microstructure, as 2.7. Similarly, the
calculated densities for the sample containing ∼12% epoxy
were 3.5 at the surface and 3.65 internally.
It should be noted that the compositions predicted from
the initial PU foam compression ratios differed significantly
from those realised experimentally, as shown in Fig. 4. The
assumption that final epoxy volume fraction is equal to ini-
tial PU foam volume fraction appears to overly simplify the
process of slip-casting and drying. It is possible that infiltra-
tion of the PU foam with alumina slip is incomplete due to
trapped air at vertices of foam ligaments, or that shrinkage
during drying leads to increased porosity in the ceramic body.
173
Fig. 4. Comparison of theoretical alumina volume fractions calculated from
PU foam compression ratios with alumina volume fractions determined from
microstructural analysis.
2.3. Elastic property characterisation
For measuring elastic moduli, the impulse excitation tech-
nique (IET) provides an alternative to tensile or flexural load-
ing, either of which may damage brittle material samples.
This technique involves lightly striking a rectangular-shaped
specimen to stimulate vibration in its natural modes [25–27].
Vibrational frequencies were measured with a micro-
phone, amplifier and signal analyser (Analyzer 2000 V5+,
www.brownbear.de). Resonant frequencies were obtained for
flexural and torsional modes, from which Young’s and shear
moduli, respectively were obtained. Samples were tested
both as plates (50 mm × 30 mm × 5 mm approx.) and later as
beams (50 mm × 5 mm × 4 mm approx.), to investigate the
possibility of variation in properties across samples.
Values for Young’s modulus, E, were calculated from flex-
ural frequencies, fflex , using the relation:
mL3
E = 0.9465 T1 fflex
2
(3)
bt 3
174 M.T. Tilbrook et al. / Materials Science and Engineering A 393 (2005) 170–178

Fig. 5. Young’s modulus values. Experimental results for interpenetrating- Fig. 6. Young’s modulus values. Experimental results for interpenetrating-
network composite (
) and powder composite (䊉) samples, and theoretical network composite (
) and powder composite (䊉) samples, computational
predictions: isostress (A); isostrain (B); Hashin–Shtrikman upper (C) and predictions () from Wegner and Gibson [8], and theoretical predictions:
lower (D) bounds. Tuchinskii unit-cell model upper and lower bounds, and effective medium
approximation with ψ = 1 and 6.
where L is the specimen length, b the width, t the thickness,
m the mass and T1 the geometry constant determined from b
and t [27]. Similarly, the relationship used to calculate shear
modulus, G, from the torsional frequency, ftors , was:
Lm B
G=4 f2 (4)
bt 1 + A tors
where A and B are geometry constants determined from L, b
and t [27]. Poisson’s ratio was subsequently calculated as:
E
ν= −1 (5)
2G
Error margins in modulus values were estimated from er-
rors in dimensional measurements and resonant frequency
readings. Errors in Poisson’s ratio results were relatively high
due to build-up of errors from Young’s modulus and shear
modulus calculations. Fig. 7. Shear modulus values. Experimental results for interpenetrating-
network composite samples, and EMA predictions (ψ = 6).

3. Results and discussion

3.1. Experimental results

Experimental results for Young’s modulus are shown plot-

ted against composition in Figs. 5 and 6, in comparison with
a number of mixing law predictions, which are discussed
shortly. The results for shear modulus and Poisson’s ratio are
shown in Figs. 7 and 8, respectively. Table 1 shows values ob-

Table 1
Experimentally obtained properties for alumina and epoxy
Property Alumina Epoxy
E (GPa) 390 3.4
Poisson’s ratio, ␯ 0.2 0.35
K (GPa) 280 3.8
Fig. 8. Poisson’s ratio values. Experimental results for interpenetrating-
G (GPa) 155 1.3
network composite samples, and EMA predictions (ψ = 6).
 M.T. Tilbrook et al. / Materials Science and Engineering A 393 (2005) 170–178
Fig. 9. Variation of effective Young’s modulus with measured density. Ex-
perimental results for interpenetrating-network composite and powder com-
posite samples, and EMA predictions (ψ = 6).
tained for the monolithic alumina and epoxy samples. These
are in agreement with results published elsewhere [22].
Fig. 9 shows a correlation between Young’s modulus and
density. It can also be seen that there is some variation in prop-
erties between beam samples cut from the same plate. This
was attributed to compositional variation across the plates,
which was identifiable in the variation of measured density.
These results are of particular importance from a process-
ing point-of-view and highlight the importance of ensuring
sample homogeneity.
The possibility was considered that samples might display
anisotropic properties, as a result of the production process. If
samples were anisotropic, their flexural rigidities in different
planes would differ, which would lead to differing Young’s
modulus results for in-plane and out-of-plane measurements.
As shown in Fig. 9, these results did not differ significantly.
This was corroborated by microanalysis of samples in planes
perpendicular (Fig. 2), and parallel (Fig. 10), to the direction
of foam-compression. No anisotropy was observed.
Results for the powder–matrix composite samples are in-
cluded in Figs. 5, 6 and 9. Compositions were estimated from
Fig. 10. Representative microstructural images, on planes parallel to the direction of foam compaction for (A) 25% epoxy sample and (B) 10% epoxy specimen.
175
density measurements and microstructural analysis, however,
this was difficult due to the disordered structure and residual
porosity, therefore, values are only approximate. The prop-
erties are apparently dominated by the epoxy matrix phase,
although the presence of alumina particles does increase stiff-
ness significantly.
3.2. Theoretical modelling predictions
The effective composite property models men-
tioned earlier have been applied to this alumina–epoxy
interpenetrating-structured composite system. The proper-
ties of alumina and epoxy given in Table 1 were used in all
calculations.
The isostress and isostrain bounds are shown in Fig. 5 as
curves A and B. Although these bounds do contain the experi-
mental results, they differ widely, prohibiting their use for ac-
curate predictions. The Hashin–Shtrikmann bounds [18] are
shown in Fig. 5 as curves C and D. The results also lie within
these bounds; however, they provide only a slight improve-
ment on the isostress and isostrain bounds. The inadequacy of
the Hashin–Shtrikman model is probably attributable to the
disparity in constituent properties, resulting in irregularities
in stress transfer at interfaces and stress distribution between
phases.
The bounds predicted by Tuchinskii’s unit-cell model [16]
are displayed in Fig. 6. Values predicted by this model are too
high. This may be attributed to the assumption of a simple
regular cubic structure, which as shown in the microstructural
images, was not the case for these composites.
3.3. Comparison with computational model
Wegner and Gibson modelled interpenetrating network
structured two-phase composites using finite element anal-
ysis [8]. They assumed one phase was comprised of slightly
overlapping spheres in a regular stacked array, whilst the other
phase filled the interstitial volume. This compares well to
the microstructures shown in Fig. 1 for lower epoxy volume
fractions. They predicted a sharp drop-off in stiffness with
increasing interstitial phase volume fraction, which agrees
176 M.T. Tilbrook et al. / Materials Science and Engineering A 393 (2005) 170–178
with the experimental results as shown in Fig. 6. This is good
although such a computational model is time-consuming to
produce and limited in applicability.
3.4. Effective medium approximation
= 1).
For a material with a random grain structure or a matrix-
inclusion topology with contacting inclusions, the effective
medium approximation method may be appropriate [4]. This
technique, based on Hill’s self-consistent model [28], treats
the composite material as being comprised of ellipsoidal in-
clusions of each phase, surrounded by an effective medium
with the same properties as the total composite [29,30]. As
well as composite elastic properties, this concept has been ap-
plied to effective electrical [31] and transport [32] properties
and to microcracked solids [33].
The derivation, which is outlined here and given in full
by Kreher and Pompe [4], utilises the second-order tensor
formulation for stress and strain. These are related by the
fourth-order compliance tensor:
σ- = C- ε-
Tensors are denoted by an underline. Considering a single el-
lipsoidal inclusion with compliance, C- , in surrounding mate-
rial with effective compliance, C- ∗ , and a prescribed far-field
strain, the local strain field is determined via the results of
Eshelby [34].
Applying this self-consistently to each phase of the com-
posite results in a weighted mean for the effective compliance
tensor, C- ∗ , involving the compliance, C- (n) , and the inclusion
shape-dependent strain relation tensor, T- (n) , for each phase:
(C- (n) − C- ∗ )T- (n) = 0
where  denotes the mean over all phases. This expands to
yield expressions that implicitly define the effective bulk and
shear moduli, K* and G* :
(Kn − K∗ )TK = 0, (Gn − G∗ )TG = 0
(n) (n)
(i) (i)
The strain relation tensor terms, TK and TG , are deter-
mined from a tensor A - which is defined in terms of Eshelby’s
inclusion shape tensor P- and the shape functions f(ψ), where
ψ is the shape parameter describing the ellipsoid aspect ratio
assumed for each phase (prolate for ψ < 1, spherical for ψ = 1,
and oblate for ψ > 1). These tensors and functions are given
in the appendix for completeness. For spherical inclusions,
the strain relation tensors simplify considerably leading to
the simplified equations:
   
Kn − K∗ Gn − G∗
=0 =0 (9)
Kn + K̃∗ Gn + G̃∗
with the terms in the denominators given by:
4 ∗ ∗ 1 9K∗ + 8G∗
K̃∗ = G G̃ = G∗ ∗ (10)
3 6 K + 2G∗
By solving these self-consistently for K* and G* using
MathCad, E* and ν* may be determined for spherical inclu-
sions (ψ = 1), as shown in Fig. 6. An iterative scheme was
used to calculate composite properties for non-spherical in-
clusions (ψ
While it is possible to define different values of ψ for each
phase, resulting in different shape tensors, this was not pur-
sued in the current work, due to the uncertainty in selecting
shape factor values. Rather, a single shape factor was used
for both phases. As seen in Fig. 6, the closest fit was ob-
tained with the EMA by altering the inclusion shape param-
eter, ψ, to a value of 6. This is not physically unreasonable
as microstructural images indicated a fairly irregular inter-
nal geometry, particularly for the epoxy phase. Also, from
considering the results in Fig. 5, the Poisson’s ratio of the
alumina was assumed to be 0.2, which is lower than values
typically quoted in the literature [22]. This prediction is in-
cluded in Figs. 6–9, and in each of these cases shows good
agreement with measured results.
The success of the EMA in simulating elastic proper-
ties for this type of composite structure may be attributed
(6) to its lack of inherent assumption regarding internal geom-
etry. Rather, inclusions are assumed to be surrounded by an
effective medium. In the composites investigated, any partic-
ular microstructural element of alumina or epoxy would be
in contact with both alumina and epoxy though not in any
regular configuration, and thus, is most accurately modelled
via the effective medium concept. This is in agreement with
the conclusions of Christensen [35] and Munro [36].
It is possible that observed deviations from the predictions
of the EMA could be explained by the observed variation in
microstructural morphology across the composition range.
(7) Incorporating such effects into a model would be difficult,
however, and of questionable benefit, given the reasonably
good fit obtained using the EMA in the formulation presented.
(8) 4. Conclusions
The following conclusions were made:
1. The impulse excitation technique is a suitable method for
measuring mechanical properties of ceramic–epoxy com-
posites.
2. The effective medium approximation provided the only
accurate prediction of elastic properties of two-phase
composites with an interpenetrating network structure.
3. The Hashin–Shtrikman bounds and Tuchinskii model did
not accurately predict elastic properties. This was at-
tributed to widely differing constituent properties and ir-
regularity in internal geometry.
4. Slight compositional variation resulted in significant vari-
ation in mechanical properties, which highlighted the im-
portance of careful processing.
 M.T. Tilbrook et al. / Materials Science and Engineering A 393 (2005) 170–178 177

Appendix A where δ(α,β) is the Kronecker delta, equal to one for α = β,

and zero otherwise.The total magnitude are given by:
The full set of equations used in modelling composite
2 (n) (n) (n) (n) (n)
properties via the effective medium approximation technique (A(n) ) = A33 (A11 + A12 ) − 2A13 A31
is included here, for completeness. A more detailed deriva-
tion is given in [4] which follows the work of Eshelby [34] Inverting A (n)
- gives the corresponding strain relation tensor,
and others [28,32]. The shape functions, f1 and f2 , are defined T- :
(n)

in a piecewise manner, in terms of shape parameter, ψ: (n) −1

 T- (n) = (A )

 -
 −ψ 2 1 1
 cosh−1 − , 0 < ψ < 1 (prolate ellipsoids)
 2(1 − ψ2 )3/2 ψ 2(ψ2 − 1)

1
f1 (ψ) =  , ψ = 1 (spheres)
3


 ψ2
 −1 1 1
 2|1 − ψ2 |3/2 cos − , 1 < ψ (oblate ellipsoids)
ψ 2(ψ2 − 1)

 3f1 (ψ) − 1 The strain relation coefficients are then calculated as:
 , ψ
= 1
f2 (ψ) =  1 − ψ2 (n) 1 (n) (n) (n) (n) (n)
 0.2, TK = [A11 + 2A33 + A12 − 2(A13 + A31 )]
ψ=1 3(A(n) )
2

The shape polarisation tensor P- is given in condensed Voigt

notation as [34]: (n) 2 (n) (n) (n) (n) (n)
TG = 2
[A11 + 0.5A33 + A12 + A13 + A31 )]
3K∗ + G∗ 15(A(n) )
P11 =P22 = ((13 − 16ν∗ )f1 (ψ) − 3f2 (ψ))  
8G∗ (3K∗ + 4G∗ ) 1 2 1
+ (n) (n)
+ (n)
3K∗ + G∗
5 A11 − A12 A44
∗ ∗
P33 = (1 − 2ν − (1 − 4ν )f1 (ψ) − f2 (ψ))
G∗ (3K∗ + 4G∗ )
These calculations were done iteratively via MathCad for a
number of compositions. Initial guess values, K(0) and G(0) ,
−3K∗ − G∗
P12 = P21 = (f1 (ψ) + f2 (ψ)) were determined from Eq. (8), then the following iteration
8G∗ (3K∗ + 4G∗ ) scheme was utilised to obtain stable solutions:
−3K∗ − G∗ (n)
(Kn − K(i) )TK (K(i−1) , G(i−1) ) = 0
P13 =P31 =P23 = P32 = (f1 (ψ) − f2 (ψ))
2G∗ (3K∗ + 4G∗ )
(n)
(Gn − G(i) )TG (K(i−1) , G(i−1) ) = 0
3K∗ + G∗
P44 = P55 = where the strain relation coefficients were calculated using
2G∗ (3K∗ + 4G∗ )
the effective property values from the previous iteration.
×(1 − ν∗ − (2 − ν∗ )f1 (ψ) + f2 (ψ))

P66 = 0.5(P11 − P12 ) References

The summative terms are given as: [1] Z. Hashin, J. Appl. Mech. 50 (1983) 481.
3 [2] G.P. Sendeckyj, Elastic behaviour of composites, in: G.P. Sendeckyj
P1 = P2 = f1 (ψ), (Ed.), Mechanics of Composite Materials, vol. 2, Composite Mate-
3K + 4G∗
∗
rials, Academic Press, 1974, p. 45.
3 [3] K.K.U. Stellbrink, Micromechanics of Composites: Composite Prop-
P3 = (1 − 2f1 (ψ)) erties of Fibre and Matrix Constituents, Hanser, 1996, p. 12.
3K∗ + 4G∗ [4] W. Kreher, W. Pompe, Internal Stresses in Heterogeneous Solids,
Akademie Verlag, 1989.
Tensor, P- (n) , for each phase is determined from the terms of
[5] S. Torquato, Appl. Mech. Rev. 44 (1991) 37;
the components of P- and the summative terms: S. Torquato, Int. J. Sol. Str. 35 (1998) 2385.
[6] H. Prielipp, M. Knechtel, N. Claussen, S.K. Streiffer, H. Müllejans,
Aαβ = δ(α, β) + 2(Gn − G∗ )Pαβ
(n)
M. Rühle, J. Rödel, Mater. Sci. Eng. A 197 (1995) 19.
[7] M. Sternitzke, M. Knetchel, M. Hoffman, E. Broszeit, J. Rödel, J.
+[Kn − K∗ − 23 (Gn − G∗ )]Pα , α, β = 1, 2, 3 Am. Ceram. Soc. 79 (1) (1996) 121.
[8] L.D. Wegner, L.J. Gibson, Int. J. Mech. Sci. 42 (2000) 925.
[9] A. Neubrand, J. Rödel, Zeit. f. Met. 88 (1997) 358.
A44 = 0.5 + 2(Gn − G∗ )P44
(n)
[10] S. Suresh, A. Mortensen, Int. Mater. Rev. 42 (3) (1997) 85.
178 M.T. Tilbrook et al. / Materials Science and Engineering A 393 (2005) 170–178
[11] T. Reiter, G.J. Dvorak, V. Tvergaard, J. Mech. Phys. Solids 45 (8)
(1997) 1281.
[12] T.J. Chung, A. Neubrand, J. Rödel, T. Fett, Ceram. Trans. 114 (2001)
789.
[13] J.R. Cho, D.Y. Ha, Key, Eng. Mater. 183–187 (2000) 373.
[14] P.-C. Zhai, Q.-J. Zhang, R.-Z. Yuan, Mater. Sci. Forum 308–311
(1999) 995.
[15] M. Dong, P. Leßle, U. Weber, S. Schmauder, Mater. Sci. Forum
308–311 (1999) 1000.
[16] L.I. Tuchinskii, Porosh Metall 7 (247) (1983) 85 (translated in: Pow-
der Metallurgy and Metal Ceramics, Plenum, 1983).
[17] S. Torquato, C.L.Y. Yeong, M.D. Rintoul, D.L. Milius, I.A. Aksay,
J. Am. Ceram. Soc. 82 (5) (1999) 1263.
[18] Z. Hashin, S. Shtrikman, J. Mech. Phys. Solids 11 (1963) 127.
[19] M. Hoffman, S. Skirl, W. Pompe, J. Rödel, Acta Mater. 47 (2) (1999)
565.
[20] R. Jedamzik, A. Neubrand, J. Rödel, Mater. Sci. Forum 308–311
(2000) 782.
[21] X.-Q. Feng, Z. Tian, Y.-H. Liu, S.-W. Yu, Appl. Comp. Mats. 11
(2004) 33.
[22] M.F. Ashby, D.R.H. Jones, second ed., Engineering Materials 1: An
Introduction to their Properties and Applications, vol. 2, Butterworth
Heinemann, 1996.
[23] F.R. Cichocki, K.P. Trumble, J. Rödel, J. Am. Ceram. Soc. 81 (6)
(1998) 1661.
[24] J.C. Russ, Image Analysis Handbook, third ed., CRC Press, Boca
Raton, FL, 1998.
[25] ASTM Standard C1259-98. Standard Test Method for Dynamic
Young’s Modulus, Shear Modulus and Poisson’s Ratio for Advanced
Ceramics by Impulse Excitation of Vibration.
[26] K. Heritage, C. Frisby, A. Wolfenden, Rev. Sci. Instr. 59 (6) (1998)
973.
[27] R.R. Atri, K.S. Ravichandran, S.K. Jha, Mater. Sci. Eng. A 271
(1999) 150.
[28] R. Hill, J. Mech. Phys. Solids 13 (1965) 213.
[29] B. Budiansky, J. Mech. Phys. Solids 13 (1965) 223.
[30] L.J. Walpole, J. Mech. Phys. Solids 17 (1969) 235.
[31] D. Stroud, Superlattices Microstruct. 23 (3–4) (1998) 567.
[32] J.R. Willis, J. Mech. Phys. Solids 25 (1977) 185.
[33] Y. Huang, K.X. Hu, A. Chandra, J. Mech. Phys. Solids 42 (8) (1994)
1273.
[34] J.D. Eshelby, Proc. Roy. Soc. A 349 (1957) 376.
[35] R.M. Christensen, J. Mech. Phys. Solids 38 (3) (1990) 379.
[36] R.J. Munro, J. Am. Ceram. Soc. 84 (5) (2001) 1190.