Sie sind auf Seite 1von 11

Available online at www.sciencedirect.

com

Acta Materialia 57 (2009) 97107


www.elsevier.com/locate/actamat

The inuence of the reinforcing particle shape and interface strength


on the fracture behavior of a metal matrix composite
V.A. Romanova a,*, R.R. Balokhonov a, S. Schmauder b
a

Institute of Strength Physics and Materials Science, Siberian Branch of the Russian Academy of Sciences, Department of Mechanics
of Heterogeneous Media, 634021 Tomsk, Russia
b
Institut fur Materialprufung, Werkstofkunde und Festigkeitslehre, Stuttgart University, 70569 Stuttgart, Germany
Received 6 November 2007; received in revised form 19 August 2008; accepted 28 August 2008
Available online 1 October 2008

Abstract
A numerical analysis of the reinforcing particle shape and interface strength eects on the deformation and fracture behavior of an
Al/Al2O3 composite is performed. Three-dimensional calculations are carried out for ve elasticbrittle particles embedded into the elasticplastic matrix, the reinforcing particle shape being varied from spherical to strongly irregular. It is shown that microstructural heterogeneity of the composite gives rise to a complex stressstrain state in the vicinity of particle boundaries and hence to near-interface
areas undergoing tensile deformation both in tension and compression. Within the strain range under study, compressive strength is not
achieved, either in compression or in tension, i.e., all cracks grow only under tensile stress. Particle fracture is found to occur by two
mechanisms: interface debonding and particle cracking. Individual and combined eects of the particle shape, interface strength, and
loading conditions on the fracture mechanisms are analyzed.
2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
Keywords: Finite-dierence modeling; Metal matrix composites; Fracture; Mesoscale stressstrain state

1. Introduction
Experimental and theoretical investigations bear witness
to a key role of the internal interfaces in the generation of
stress concentrations, localization of plastic strains, and
development of relaxation processes in materials subjected
to deformation. A special role of the microstructure is most
evident in composites characterized by distinct internal
interfaces. Knowledge of the deformation and fracture
mechanisms developing in these kinds of materials under
loading is of critical importance for estimation of the reliability, durability, and operating capacity of engineering
hardware and machine parts as well as for the optimization
of the materials composition and structure.
Particle-reinforced metal matrix composites (PRCs) are
an important class of composite materials. Due to high

Corresponding author. Tel.: +7 3822 286937; fax: +7 3822 492576.


E-mail address: varvara@ispms.tsc.ru (V.A. Romanova).

elastic moduli, strength, fatigue and wear resistance, these


lightweight materials have a number of advantages over
ordinary aluminum alloys [14]. It is common knowledge,
however, that on the one hand, reinforcing particles do
enhance the stiness, strength and wear resistance of metal
matrix composites and, on the other hand, reinforcements
can produce an adverse eect on the fracture toughness and
ductility of the materials (see, e.g., Refs. [1,5]).
Many factors responsible for the macroscopic properties
of the composites are discussed in the literature, including
load transfer between the matrix and reinforcements [6],
presence of precipitations at the matrix/particle interface
[2,7], mechanical characteristics of individual components
of the materials [813], residual stresses resulting from
the technological processing of a mismatch between the
thermal expansion coecients of the components
[12,14,15]. Important characteristics providing a key contribution to damage accumulation and fracture of the
materials are the reinforcing particle size, shape, volume
fraction, and spatial distribution [1,46,1013,1622].

1359-6454/$34.00 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2008.08.046

98

V.A. Romanova et al. / Acta Materialia 57 (2009) 97107

A large number of experimental and theoretical data on


the microstructural factors responsible for the deformation
and fracture behavior of the PRCs is available at present.
Composites with reinforcing particle clusters are reported
to possess a higher liability to fracture than those demonstrating a uniform distribution of the reinforcements
[15,16]. Larger particles tend to fail at lower external loads
[6,14]. This can be explained by the fact that a higher load
can be transferred to larger reinforcing particles from the
matrix than to smaller ones. An increase in the particle volume content enhances the composite strength and drastically decreases the fracture toughness [3]. Moreover, the
irregular shape of reinforcements can signicantly aect
the cracking process [1012,17,18,20]. As shown in Ref.
[17], the presence of coral-shaped Si particles in Al/Si alloys
inuences the crack evolution pattern in these materials.
Similar conclusions are drawn in Ref. [18] for sharp-cornered particles.
A detailed understanding of the macroscopic mechanical behavior of metal matrix composites requires knowledge of the quantitative characteristics of the stress and
strain distributions at the particle scale level. A thorough
study of the damage mechanisms at the meso level acquires
a special signicance, since gradual accumulation of irreversible deformation and damage at lower scales gives rise
to macroscopic failure of the structure [2325].
A correct prediction of the mesoscale stressstrain distributions and damage accumulation calls for the development of three-dimensional models accounting for real
microstructures of materials. Analytical methods based,
e.g., on the Eshelby analysis [26], cannot take into consideration the inclusion shape and clustering eects and can
hardly be extended to real composites with a complex
microstructure. While allowing for a proper estimation of
the stressstrain elds and crack-growth parameters for a
large number of materials, numerical simulations (see,
e.g., Refs. [10,11,13,17,20,2733]), for the most part, fail
to account for the irregular shape of reinforcements or
are limited to two-dimensional models. So are the combined numerical-analytical approaches involving, e.g.,
Laplace and fast Fourier transform algorithms [3436],
Ponte Castaneda formalism [35,36] etc. Recent computer
modeling eorts (e.g., Refs. [3741]) do incorporate irregularly shaped reinforcing particles in three-dimensional calculations, however. In Refs. [40,41], the deformation
behavior of an Al/Al2O3 composite with a computerdesigned microstructure was calculated numerically. The
particle shape and size were varied in a way similar to that
observed in the experiment [42] and a correlation between
the local characteristics of equivalent stresses and strains
and the macroscopic stressstrain curve was established.
It was shown that due to a strongly distorted matrix/particle interface, a complex three-dimensional stressstrain
state was formed in the vicinity of the interface even under
uniaxial macroscopic loading. Microstructure-based simulations of the deformation behavior of an aluminum alloy
reinforced with SiC particles were reported in Refs.

[38,39]. A serial sectioning technique was used to reconstruct a three-dimensional microstructure characterized
by real morphology and spatial distribution of the reinforcements. Subsequently, the stressstrain behavior of
representative real microstructures was simulated by the
nite-element method and compared with that predicted
analytically and numerically from simplied models
approximating the reinforcements by spheres or ellipsoids.
It was shown that the model incorporating the real microstructure of the composite could provide a more accurate
prediction of the macroscopic behavior of the material
under uniaxial loading. The same conclusion was made in
Ref. [37] for a SnAg alloy analyzed numerically in a similar manner.
Valuable as they are, Refs. [3741] cover only a few
aspects of the PRC deformation and disregard the fracture
phenomena. Many related problems call for further investigations, using three-dimensional microstructure-based
approaches. It is the goal of the present paper to perform
a numerical analysis of the particle shape and interface
strength eects on the composite deformation and cracking
at the mesoscale level. As an example, a metal matrix composite (MMC) made up of an Al(6061) matrix and Al2O3
reinforcements is examined in tension and compression.
The three-dimensional mechanical problem is solved by
the nite-dierence method [43,44]. To separate the size,
volume content, and spatial distribution eects inherent
in multi-particle models from those resulting from the reinforcing particle shape, we restrict the discussion to individual inclusions of dierent shapes embedded into the
aluminum matrix. Particular emphasis is given to an investigation of the interface strength eect on the fracture
mechanisms.
2. Mathematical formulation of the problem
2.1. System of equations
It is known that growing cracks generate release waves
which may aect nucleation of new cracks and evolution
of presenting cracks. To account for the crack propagation
dynamics, a three-dimensional mechanical problem is
solved in a dynamic formulation. The system of dierential
equations for a barotropic medium includes a continuity
equation
V_
 U i;i 0
V

an equation of motion
qU_ i rij;j

and expressions for stress tensor components


rij P dij S ij

where i = 1, 2, 3, j = 1, 2, 3, U i x_ i , Ui and xi are the velocity vector components and Cartesian coordinates, respectively, V = q0/q is a relative volume, q0 and q are the

V.A. Romanova et al. / Acta Materialia 57 (2009) 97107

reference and current densities, and dij is the Kronecker


delta. The dot denotes the time derivative.
The pressure is found from a linear barotropic equation
of the form
P Kekk
and deviatoric stresses are given as
1
S_ ij kS ij 2l_eij  e_ kk dij
3

where k > 0 under plastic deformation and k 0 at the


elastic loading stage, K and l are the bulk and shear moduli, respectively. The strain rate tensor is
1
e_ ij U i;j U j;i
2

The nite-dierence method developed in Ref. [43] and


extended to the three-dimensional case of a heterogeneous
material in Ref. [44] is adopted for a numerical solution of
Eqs. (1)(6).
2.2. Fracture model
The necessity of establishing an adequate local damage
criterion within the inclusion problem arose about three
decades ago (see, e.g., Ref. [45]). Subsequently, several
types of damage criteria were formulated, including
stress-based [46], strain-based [4749] and energy-based
ones [5052,33]. The latter approach is of particular importance in solving three-dimensional fracture problems, since
it is based on an invariant value and accounts for a combined eect of the stress tensor components.
In this work, a fracture criterion based on maximum
equivalent stress theory is adopted to simulate the damage
evolution in alumina particles. In its classical formulation
[50,51], this theory, widely used in engineering mechanics
and failure analysis for homogeneous ductile materials,
cannot provide an adequate description of brittle fracture.
For a heterogeneous material with explicitly introduced
microstructure, however, it is shown to yield correct results
both in tension and in compression [53]. For any loading
conditions, the microstructure heterogeneity causes local
mesoscale tensile areas to appear where fracture meets
the maximum equivalent stress criterion. The two-dimensional calculations [53] for an Al/Al2O3 composite subjected to tension or compression showed that the
majority of cracks at the mesoscale originated in tensile
or shear deformation regions even under compression. In
this paper, the fracture criterion based on a maximum
equivalent stress is extended to the three-dimensional case.
The criterion holds that fracture will occur, provided that
an equivalent stress req reaches a critical value r*, i.e.,
1 p
7
req p S ij S ij r
2
The experimental data [42] show that the compressive
strength of ceramic particles is generally several times
higher than the tensile strength. To account for the dier-

99

ence in magnitude between the tensile and compressive


strength found in brittle materials, criterion (7) is given in
the form
 
rten if ekk > 0
8
req
rcom if ekk < 0
where rten and rcom are the values of the tensile and compressive strength. In other words, for ekk > 0, a local region
of the material will fail if the local equivalent stress req
equals the tensile strength, whereas for ekk < 0, the fracture
surface in stress space is limited by the compressive
strength. On further loading, for ekk > 0 the damaged regions are assumed to lose their strength so that both the
deviatoric stress (5) and pressure (4) reduce to zero. If the
local ekk is negative, the damaged material can oer resistance to a compressive load alone, i.e., the pressure is calculated by Eq. (4) and the stress deviator (5) is zero.
2.3. Specimen geometry, mechanical properties, and loading
conditions
Three-dimensional nite-dierence calculations were
performed for ve individual particles embedded into the
aluminum matrix, with the particle shape being varied
from spherical to strongly irregular, Fig. 1ae. In all models, the particles occupy 20% of the specimen volume. The
computational grid is 80  80  80 with a grid step of 1
lm. The load was applied to the specimen surfaces x1 = 0
and x1 = L1 (L1 is the current specimen length along the
axis of loading) moving parallel to the X1-axis either
toward one another in compression or in the opposite
direction in tension, as presented schematically in Fig. 1f.
The other four surfaces were not loaded.
For a quantitative comparison of the particle shapes, we
introduce a notion of relative surface roughness, S*, equal
to the ratio of the particle surface area to that of a sphere
of the same volume. Thus
S

S particle
S sphere

It is obvious that S  P 1, and the higher S*, the stronger


the surface roughness.
To design particles of dierent shapes, use was made of
a step-by-step packing (SSP) method put forward in Ref.
[44]. Based on the SSP-method, a procedure of design of
particles with controllable surface roughness was developed. First, 50 nuclei of the reinforcing phase were originally distributed over a cubic volume V0 (see Fig. 1f). At
each step in the processing time, the volumes surrounding
the nuclei were increased by preset values in accordance
with a spherical law. The procedure was repeated until
the volume fraction of the reinforcing phase reaches 20%
of the specimen. Due to a coalescence of the growing
spheres, each resulting model, Fig. 1ae, contained a single
monolithic particle of a complex shape. It was found that
the particle surface roughness S* depended on a relative

100

V.A. Romanova et al. / Acta Materialia 57 (2009) 97107

Fig. 1. Composite models (ae) and a schematic representation of the distribution of reinforcing phase nuclei in the SSP procedure (f). White and black
arrows indicate directions of load application in tension and compression, respectively.

volume of nuclei distribution, V 0 , in accordance with the


growth exponent law, Fig. 2
S  0:88 0:11 exp 0:12V 0

0

10

V0
Vs

Relative surface roughness S

where V  100%, Vs is the specimen volume. In the


discussion below the particles illustrated in Fig. 1ae are referred to as Pi, i = 15, in order of increasing of their S*
values.
According to the experimental data reported in Refs.
[7,42], Al2O3 particles possess elasticbrittle properties.
The elastic response is described by Eq. (5) with k 0.
The fracture criterion (8) is used to dene particle fracture.
As the crack nucleation in an Al(6061) alloy has not
been detected experimentally for a degree of strain of about
1.5% [42], the matrix is assumed to be elasticplastic. Its
elastic behavior is calculated by Eq. (5), whereas the plastic

2.0
1.8

Numerical data
Fitted curve (10)

1.6

P5

1.4

response meets the von Mises yield criterion [43]. To


account for the strain hardening of the matrix use is made
of a tting function of the Voce type as proposed in Ref.
[42]


. 
11
req r0 rmax  r0 1  exp epeq e0
q
Here epeq p23 epij epij is the equivalent plastic strain, rmax is
the saturation stress, r0 is the yield limit, and e0 is a model
constant controlling the strain hardening. The mechanical
properties of the matrix and particles and the tting constants derived in Refs. [42,53] are given in Table 1.
For the ve composite models, Fig. 1ae, a series of
numerical experiments was carried out where the interface
strength and loading conditions were varied in dierent
combinations. Three models of the interface behavior were
constructed for each of the ve specimens, with the interface strength rinterface related to the particle strength rAl2 O3
as 0.5, 1, and 1.5. The resulting fteen models diered by
the particle shape and interface properties were, in turn,
examined in conditions of tension and compression. Thus,
the results of thirty numerical experiments were brought
together to analyze the combined eects of the particle
shape, interface strength, and loading conditions on the
deformation and fracture behavior of the composites.
3. Numerical results and discussion

1.2
1.0

P1
0

P2

P3

P4
10

15
*

Relative volume of nuclei distribution V 0, %


Fig. 2. Relative surface roughness of the resulting particles vs. input data
of the SSP procedure.

Fig. 3 shows fracture patterns in particles P1P5 in tension and compression. Long chains of the damaged regions
observed on the surface and in the bulk of the particles are
regarded as cracks. Occasionally the damaged regions
appear on the particle surface in the form of isolated
islands treated as voids. During loading some interfacial
voids coalesce into larger patches of the fractured material,

V.A. Romanova et al. / Acta Materialia 57 (2009) 97107

101

Table 1
Material constants and model parameters

Al(6061)
Al2O3

q0 (kg m3)

l (GPa)

K (GPa)

rten (MPa)

rcom (MPa)

r0 (MPa)

rmax (MPa)

e0

2700
3990

27.7
156.0

72.8
226.0

260.0

4000.0

105.0

170.0

0.048

which is addressed to as interface debonding. Other ones


give rise to the volume crack propagation inside the
particles.
In some cases it is dicult to dierentiate exactly the
debonding, void nucleation and volume cracking, since
these modes can transform into each other during loading.
Moreover, at an early stage of fracture it is hardly possible
to predict whether the interfacial voids will give rise to the
debonding or volume crack growth. For the sake of simplicity in the subsequent discussion, we will recognize two
basic fracture modes: interfacial fracture that implies failure of the regions belonging to the matrix/particle interface, and volume cracking that occurs inside the particles.
3.1. Fracture initiation
For all particles in tension and compression, rst cracks
appear near the interface due to stress concentration in the
near-boundary regions. Experimental and theoretical ndings (e.g., Refs. [23,42,53,54]) indicate that the stress concentration value is generally controlled by three factors:
the dierence in the mechanical properties between contacting materials, interface curvature, and loading conditions. At the prefracture stage of loading the dierence in
the mesoscale stressstrain elds between specimens
P1P5 is mostly due to the particle shape eect. The stronger the particle surface roughness, the higher the stress concentration in the vicinity of the interface and, hence, the
earlier the crack nucleation. Our calculations showed that
even small surface imperfections may give rise to a perceptible increase in local stresses. Quantitative estimates can be
given in terms of a stress concentration factor which can be
dened as
Kr

rmax
eq
hreq i

12

where rmax
eq and hreq i are the maximum (local) and average
(global) equivalent stresses. Calculations show that the
stress concentration factor depends on the surface roughness by a power law (triangles in Fig. 4). And the roughness
dependence of the macroscopic strain of damage initiation
can be tted by an exponential decay law (squares and circles in Fig. 4).
The interface points at which cracks nucleate are dependent on the particle shape and loading conditions. The
stressstrain analysis showed that: (i) the tensile regions
on the mesoscale are observed both in tension and compression, and (ii) these regions are in dierent positions.
According to the fracture criterion (8), the positions of
the crack appearance are dependent on the local values

of equivalent stresses and sign of the hydrostatic strain


ekk. Up to the point of crack initiation, the equivalent stress
elds in the same specimens in tension and compression are
identical, whereas the strain tensor components of the same
name are equal in their absolute values but opposite in
sign, so are the hydrostatic strains ekk. Since incipient
cracks in the tensile regions (ekk > 0) occur at much lower
stresses rten than in the regions experiencing compression,
cracks in tension and compression will appear at dierent
points. In tension, the regions of the highest equivalent
stresses undergo tensile deformation and the rst crack is
expected to originate there. In compression, the highest
equivalent stresses occur in compressive regions, but nevertheless the material strength is rst achieved in the tensile
regions. Fig. 5 demonstrates the evolution of maximum
equivalent stresses in tensile regions of specimens P1P5
in tension and compression. All calculations predict the
fracture initiation in the regions of hydrostatic tension,
while the compressive strength is not achieved at all. In tension fracture commences earlier in all specimens due to a
higher equivalent stress in the tensile regions.
While a systematic study of the ratio between the yield
stress and the fracture stress were not performed, some
conclusions on the plasticity eects in the matrix can be
drawn. It follows from the theory of elasticity that the
stress and strain patterns formed in an early loading stage
do not change in qualitative terms during the elastic loading, but grow proportionally as the external load is
increased. On further loading, plastic deformation in the
matrix would lead to the stress relaxation in particles,
which would postpone crack initiation in the material.
Our calculations show, however, that for all examined
models, Fig. 1ae, the matrix is unusable as a damping
material, since the plastic ow is but a trie ahead of the
rst crack nucleation both in tension and compression. In
the cases of irregularly shaped particles P3P5, the plastic
deformations in the matrix and the particle cracking commence nearly simultaneously. In specimens P1 and P2 crack
initiation in the particles is somewhat preceded by plastic
ow in the matrix, which inuences but a little the stress
concentration in near-boundary regions.
3.2. Evolution of fracture mechanisms in tension and
compression
Quantitative estimates will be given in the form of the
damage accumulation curves, Fig. 6, and the interface-tovolume fracture ratios, Fig. 7. In the latter case the curves
show how many areas of the fractured material belong to
the interfacial and internal regions (curves marked by solid

102

V.A. Romanova et al. / Acta Materialia 57 (2009) 97107

Fig. 3. Crack patterns in tension (ae) and compression (fj). Particle surfaces are marked by the computational mesh, internal sections are gray, damaged
regions are white-colored, and the matrix is transparent.

0.20

4.0

0.15

3.5

0.10

3.0

0.05

1.0

1.1

1.2

1.3

1.4

2.5

Stress concentration factor K

Damage initiation strain E, %

V.A. Romanova et al. / Acta Materialia 57 (2009) 97107

Relative surface roughness S

Max eq in hydrostatic tension regions, MPa

Fig. 4. Macroscopic strain of fracture initiation in tension (circles) and


compression (squares) and stress concentration factor at the prefracture
stage of loading (triangles) vs. the particle surface roughness.

250

ten

200
150

<eq>

100

P1 - squares
P2 - circles
P3 - up triangles
P4 - down triangles
P5 - diamonds

50
0
0.00

0.05

0.10

0.15

0.20

Engineering strain, %

Fig. 5. Evolution of maximum equivalent stresses in the regions of


hydrostatic tension vs. macroscopic strain. The curves obtained in tension
and compression are marked by solid and open symbols, respectively.

and open symbols, respectively, Fig. 7). Given along the Xaxis in Figs. 6 and 7 is the macroscopic strain corresponding to the specimen elongation in tension and to its contraction in compression


L1  L0 

  100%
13
E
L 
0

here L0 and L1 are the reference and current specimen


lengths along the axis of loading.
Scenarios of fracture evolution are strongly dependent
on the particle shape and loading conditions. The interfacial fracture in the form of debonding is a principal fracture
mode for particles P1 and P2 in tension (cf. fracture patterns in Fig. 3ab and corresponding curves in Fig. 7a).
As debonding encompasses most of the interface, the fracture rate decreases drastically, which corresponds to the
saturation portions of the damage accumulation curves
(marked by squares and circles in Fig. 6a). On further loading, the overall deformation process is localized in the

103

matrix, while the particles mostly separate from the matrix


and make a minor contribution to the load resistance of the
material.
A dierent fracture scenario is realized in tension in particles P3, P4 and P5 characterized by a more irregular
shape. The particle cracking and interfacial fracture
develop nearly simultaneously, with the latter occurring
in the form of debonding as well. Along with debonding,
the particles exhibit a few through-the-thickness cracks in
planes perpendicular to the axis of tension. All cracks originate from the interface and then propagate into the particle. The crack growth rate changes periodically. As the
crack propagates in the bulk, stress relaxation takes place
in the vicinity of the crack. This may slow down the crack
propagation to the point of cessation of the crack growth
for some time. In the meantime, other cracks continue to
develop in the volume or new cracks appear at the interface. This complex behavior is attributed to the non-homogeneous stressstrain state near the interface and in the
vicinity of the cracks as well as to the interaction of release
waves generated by the growing cracks. As one or more
cracks cover the entire sections of the particle, dividing it
into several parts, the damage accumulation progressively
approaches its saturation stage that corresponds to the plateau in the curves, Fig. 6a. Then, the particle virtually loses
its ability to resist the external load.
A basic tendency in tension is as the particle shape
becomes more irregular the debonding regions shift from
the sides to the central part of the particle surface (cf. damage patterns in Fig. 3ae) and the debonding contribution
decreases relative to that of volume cracking, Fig. 8.
In compression, Fig. 3fj, as soon as the rst damaged
regions are formed at the interface, they give rise to crack
propagation in the bulk. From the very onset of fracture
the volume cracking mechanism dominates that of debonding for all particles except of P1 where volume cracks
appear much later, Fig. 7b. In the spherical particle, at
the onset of fracture, there are no distinct surface imperfections to initiate a through-the-thickness crack. On further
compression, the material near the interface is progressively involved in the fracture process and a series of short
cracks is formed in the subsurface layer, Fig. 3f. These
cracks give rise to the additional stress concentration
regions in the subsurface layer and, as a result, a volume
crack appears, its edges being displaced relative to one
another.
It is notable that in all compressed specimens no complete debonding occurs, but the damaged regions localize
on the particle surface in the form of voids or isolated debonding patches, Fig. 3fj. As this takes place, the load is
transferred through intact surface regions into the reinforcement volume and the composite material retains its
ability to resist deformation for a comparatively longer
time than in tension conditions, where a continuous debonding process is at work (cf. curves in Fig. 6a and b).
It follows from the prefracture analysis that in compression the rst cracks appear near the interface in the regions

104

V.A. Romanova et al. / Acta Materialia 57 (2009) 97107

b
15

10

P1
P2
P3
P4
P5

0
0.1

0.2

0.3

25

Vol.% of damaged material

Vol.% of damaged material

0.4

20
15
P1
P2
P3
P4
P

10
5
0
0.0

Macroscopic strain E, %

0.5

1.0

1.5

Macroscopic strain E, %

Fig. 6. Damage accumulation curves for particles P1P5 in tension (a) and compression (b).

100

Debonding/cracking vol.%

Debonding/cracking vol.%

Debonding
P1
P2
P3
P4
P5
Cracking
P1
P2
P3
P4
P5

80
60
40
20
0
0.1

0.2

0.3

0.4

Macroscopic strain E, %

100
80
60
40
20
0
0.2

0.4

0.6

0.8

1.0

Macroscopic strain E, %

Fig. 7. Interface-to-volume fracture curves for particles P1P5 in tension (a) and compression (b).

Debonding/cracking vol. %

100

debonding
volume cracking
fitting curves

80
60
40
20
0
0.9

1.0

1.1

1.2

1.3

Relative surface roughness S

1.4
*

Fig. 8. The particle shape dependence of the debonding and volume


cracking contributions in tension.

undergoing tensile deformation. What is more, an analysis


of further fracture development shows that within the
strain range under study (up to 2%), the compressive
strength is achieved nowhere, i.e., all cracks grow only in
the regions of hydrostatic tension. At the prefracture com-

pression stage there are but a few regions of hydrostatic


tension for cracks to originate, all of these lying along
the interface. The incipient crack enhances the non-homogeneity of the stressstrain state in its neighborhood. On
further compression, the crack edges tend to separate or
to displace relative to one another, with additional tensile
regions formed near the crack tip. This is how the cracks
propagate in the bulk of the particle in compression.
Multiple volume cracks with dierent orientations relative to the compression axis act to divide particles into several parts. For particles P2P4, near straight lines of the
damaged material are observed in the sections parallel
and perpendicular to the loading axis, Fig. 4bd. This
implies that the cracks are predominantly at. Cracks in
particle P5 (Fig. 4e) exhibit more complex patterns. Round
crack contours seen in the specimen sections along with the
straight lines indicate that the whole parts of the particle
rotate relative to each other.
The results obtained for tension are in good qualitative
agreement with experiments [7,42] and with two-dimensional numerical simulations [53]. In compression, our
three-dimensional calculations provided a more detailed
description of the fracture behavior of the material than

V.A. Romanova et al. / Acta Materialia 57 (2009) 97107

the two-dimensional model [53]. In the plane strain simulations [53] cracks predominantly grew along the compression axis, whereas in three-dimensions the cracks orient
at dierent spatial angles to the load direction. This is
due to the fact that within the two-dimensional-approximations certain components of the stress and strain tensors
are assumed to be negligible. Our calculations showed,
however, that all stresses and strains on the mesoscale level
are dierent from zero and make a perceptible contribution
to the deformation and fracture behavior. Similar conclusions were done, e.g., in Ref. [10]. The authors stated that
the numerical analysis of stressstrain elds in composites
must be three-dimensional, as a two-dimensional approximation fails to describe experimental data.
3.3. Interface strength eects
To understand to what extent the fracture mechanisms
are aected by the interface properties, for each of the ve
specimens we carried out calculations where the interface
strength was two times lower and 1.5 times higher than that
of the particle.
The calculations show that the strength of a thin interface layer can essentially aect the macroscopic strain of
crack initiation, Fig. 9. This is particularly evident in the
case of spherical particles (squares in Fig. 9). For instance,
specimen P1 with the interface strength 1.5 times higher
than that of the particle demonstrates a roughly twofold
increase in the fracture resistance as compared to the case
of equal strength in the bulk and at the interface. This phenomenon cannot be explained in the framework of a simplied theory where the mechanical properties of a
composite material can be dened as the sum of contributions from all components in proportion to their volume
percent. Indeed, the volume fraction of the high- or lowstrength interface layer is negligible as compared to the
total volume of the composite, but it can produce a pro-

0.15

Particles
P1
P2
P3
P4
P5

0.10

0.05
0.4

4. Summary
Taking an Al/Al2O3 composite as an example, a numerical analysis was performed to investigate the eects of the
reinforcing particle shape and interface strength on the
deformation and fracture behavior of the material in tension and compression. Three-dimensional nite-dierence

0.25

0.20

found eect on the fracture behavior of the whole of the


material.
A plausible explanation can be given on the basis of the
fracture mechanisms and stress concentration eects. As
discussed in Section 3.2, the spherical particle demonstrates
predominantly interfacial fracture in the form of debonding or void nucleation. In the bulk of the particle, on the
other hand, the stress concentration eects are of minor
importance and cannot cause crack nucleation. Due to
strengthening of the thin interface layer, where the stresses
are maximum, the debonding mechanism is blocked up to a
macroscopic strain of 0.24% in tension and 0.37% in compression. For larger deformations, several individual voids
appear at the interface, resulting in a stress increasing in the
subsurface layer, which gives rise to crack nucleation in the
subsurface regions.
In the meantime, the interface modication of irregularly shaped particles has but a minor eect on the onset
of fracture initiation. This is due to the volume cracking
which is a dominating fracture mechanism in these specimens is at work. Even small surface imperfections give rise
to the stress concentration regions in the subsurface layer.
This makes it possible for cracks to appear there even
though a little later or earlier in comparison to the case
of a non-modied interface. It is very likely, however, that
the strengthening of a thicker interface layer would eectively block crack nucleation in the subsurface of irregularly shaped reinforcements and, thus, additionally
postpone fracture of the material.

Damage initiation strain, %

Damage initiation strain , %

0.6

0.8

1.0

1.2

1.4

interface/ Al O
2

1.6

105

0.4

0.3

0.2

Particles
P1
P2
P3
P4
P5

0.1

0.4

0.6

0.8

1.0

1.2

1.4

1.6

interface/ Al O
2

Fig. 9. Macroscopic strain of damage initiation in particles P1P5 in tension (a) and compression (b) vs. the relative interface strength.

106

V.A. Romanova et al. / Acta Materialia 57 (2009) 97107

calculations were carried out for ve elasticbrittle particles


embedded into the elasticplastic matrix, the particle shape
being varied from spherical to strongly irregular. To
describe the damage accumulation in the particles, a maximum equivalent stress criterion used originally in twodimensional calculations was extended to the three-dimensional case. The calculations showed that the fracture criterion combined with the explicit introduction of material
microstructure provided an adequate description of the
fracture behavior of the material both in tension and compression. The results obtained from the simulations suggested the following conclusions:
(1) Due to the structural heterogeneity, a complex stress
strain state is realized in the vicinity of the matrix/
particle interface and regions undergoing tensile
deformation are formed near the interface both in
tension and compression. Within the strain range
under study, there are no regions in the material
where compressive strength is achieved either in compression or in tension. In other words, no matter
what the loading conditions may be, the cracks are
seen in the tensile regions alone. Because the positions of the tensile regions are dierent in tension
and compression, fracture in these loading conditions
goes by dierent scenarios.
(2) The reinforcing particle shape essentially aects the
macroscopic strain of fracture initiation. The stronger the interface roughness, the earlier the onset of
fracture. Even slight imperfections of the particle surface can give rise to a signicant increase in the stress
concentration and hence to earlier commencement of
fracture.
(3) Two basic fracture mechanisms are operative in the
particles: interface debonding and particle cracking.
The ratio between these depends on the particle
shape, loading conditions and interface strength. In
spherical particles interface debonding dominates
particle cracking, whereas irregularly shaped reinforcements exhibit a system of through-the-thickness
cracks in the bulk. Debonding in compression manifests itself as individual patches regarded as voids.
These regions are crack nucleation sites. On further
loading the cracks propagate into the bulk of the
particle.
(4) By increasing or decreasing the strength of a thin
interface layer the onset of fracture initiation can be
aected. This is especially pronounced in the case of
an ideal spherical particle where the debonding is a
dominating fracture mechanism. Due to strengthening of a thin interface layer, the debonding mechanism
is blocked whereas in the bulk there is no stress concentration regions where crack would appear.
Summarizing, we should say that examined in this paper
are the simplied models which disregard many phenomena related to PRC fracture, such as multiple interaction

of reinforcements, presence of precipitations and intermetallic phases at the matrix/particle interface, clusterization
eects etc. Even so, the numerical results suggest us to conclude that through a modication of the particle shape and
interface properties it is possible to aect the fracture
mechanisms on the mesoscale level and, hence, the fracture
response of the material.
Acknowledgments
Financial support of the Deutsche Forschungsgemeinschaft (436RUS 17/18/07) and Siberian Branch of the Russian Academy of Sciences (project 3.6.2.3) is gratefully
acknowledged.
References
[1] Chawla N, Chawla KK. Metal matrix composites. New
York: Springer; 2006.
[2] Ceschini L, Minak G, Morri A. Compos Sci Technol 2006;66:333.
[3] Kiser MT, Zok FW, Wilkinson DS. Acta Mater 1996;44:3465.
[4] Llorca J, Gonzalez C. J Mech Phys Solids 1998;48:128.
[5] Lloyd DJ. Int Mater Rev 1994;39:1.
[6] Davidson DL. Metall Trans 1991;18A:2115.
[7] Tursun G, Weber U, Soppa E, Schmauder S. Comput Mater Sci
2006;37:119.
[8] Babout L, Brechet Y, Maire E, Fougeres R. Acta Mater
2004;52:4517.
[9] Llorca J, Poza P. Mater Sci Eng 1994;A185:25.
[10] Christman T, Needleman A, Suresh S. Acta Metall 1989;37:3029.
[11] Llorca J, Needleman A, Suresh S. Acta Metall Mater 1991;39:2317.
[12] Dunand D, Mortensen A. Mater Sci Eng A 1991;144:179.
[13] Miserez A, Rossoll A, Mortensen A. Eng Fract Mech 2004;71:
2385.
[14] Broeckmann C, Pyzalla-Schieck A. Comput Mater Sci 1996;5:32.
[15] Povirk GL, Needleman A, Nutt SR. Mater Sci Eng A 1991;132:31.
[16] Segurado J, Gonzalez C, Lorca JL. Acta Mater 2003;51:2355.
[17] Ayyar A, Chawla N. Compos Sci Tech 2006;66:1980.
[18] Lebyodkin M, Deschamps A, Brechet Y. Mater Sci Eng A
1997;234:481.
[19] Qin Shuyi, Chen Changrong, Zhang Guoding, Wang Wenlong, Wang
Zhongguang. Mater Sci Eng A 1999;272:363.
[20] Shen Y-L, Finot M, Needleman A, Suresh S. Acta Metall Mater
1995;43:1701.
[21] Kouzeli M, Weber L, San Marchi C, Mortensen A. Acta Mater
2001;49:3699.
[22] San Marchi C, Fahe Cao, Kouzeli M, Mortensen A. Mater Sci Eng A
2002;337:202.
[23] Panin VE, editor. Physical mesomechanics of heterogeneous media
and computer-aided design of materials. Cambridge: Cambridge
International Science Publishing; 1998.
[24] Sih G, editor. Role of mechanics for development of science and
technology. Proceedings of mesomechanics-2000, Xian, China; 2000.
[25] Carpinteri A, Mai Y-W, Ritchie R, editors. Advances in fracture
research. Springer; 2006.
[26] Kanninen MF, Popelar CH. Advanced fracture mechanics. Oxford: Oxford University Press; 1985.
[27] Tvergaard V, Needleman A. Int J Solids Struct 2006;43:6165.
[28] Needleman A. Ultramicroscopy 1992;40:203.
[29] Maire E, Wilkinson DS, Embury JD, Fougeres R. Acta Mater
1997;45:5261.
[30] Ghosh S, Li M, Moorthy S, Lee K. Mater Sci Eng A 1998;249:62.
[31] Raabe D, Roters F, Barlat F, Long-Qing Chen, editors. Continuum
scale simulation of engineering materials. Wiley-VCH Verlag GmbH
& Co. KgaA; 2004.

V.A. Romanova et al. / Acta Materialia 57 (2009) 97107


[32] Chandra N, Li H, Shet C, Ghonem H. Int J Solids Struct
2002;39:2827.
[33] Chingshen Li, Ellyin F. Int J Solids Struct 1998;35:637.
[34] Lahellec N, Suquet P. Int J Solids Struct 2007;44:507.
[35] Idiart MI, Moulinec H, Ponte Castaneda P, Suquet PJ. Mech Phys
Solids 2006;54:1029.
[36] Moulinec H, Suquet P. Phys B Condens Matter 2003;338:58.
[37] Sidhu RS, Chawla N. Scripta Mater 2006;54:1627.
[38] Chawla N, Ganesh VV, Wunsch B. Scripta Mater 2004;51:161.
[39] Chawla N, Sidhu RS, Ganesh VV. Acta Mater 2006;54:1541.
[40] Romanova V, Soppa E, Schmauder S, Balokhonov R. Comput Mech
2005;36:475.
[41] Romanova V, Balokhonov R, Soppa E, Schmauder S. Comput Mater
Sci 2007;39:274.
[42] Soppa E, Schmauder S, Fischer G, Brollo J, Weber U. Comput Mater
Sci 2003;28:574.
[43] Wilkins M. Computer simulation of dynamic phenomena. Springer;
1999.

107

[44] Romanova V, Balokhonov R, Makarov P, Schmauder S, Soppa E.


Comput Mater Sci 2003;28:518.
[45] Argon AS, Im J, Safoglu R. Metall Trans 1975;6A:825.
[46] Needleman A, Rice JR. In: Koistinen DP, Wang NM, editors.
Mechanics of sheet metal forming. New York: Plenum Press; 1978.
[47] Gurland J, Plateau J. Trans ASM 1963;56:442.
[48] Ravichandran G, Liu CT. Int J Solids Struct 1995;32:979.
[49] Ellyin F, Xia Z. J Eng Mater Technol 1993;115:411.
[50] Kachanov L. Basics of fracture mechanics. Moscow: Science; 1974 [in
Russian].
[51] Cherepanov G. Mechanics of brittle fracture. New York: McGraw
Hill; 1979.
[52] Lemaitre J. J Eng Mater Technol 1985;107:83.
[53] Balokhonov R, Romanova V, Schmauder S. Comput Mater Sci
2006;37:110.
[54] Panin AV, Klimenov VA, Abramovskaya NL, Son AA. Phys
Mesomech 2000;3:83.

Das könnte Ihnen auch gefallen