Available online at www.sciencedirect.

com

Acta Materialia 60 (2012) 1623–1632

www.elsevier.com/locate/actamat

Polycrystal model of the mechanical behavior of a Mo–TiC30 vol.%

metal–ceramic composite using a three-dimensional
microstructure map obtained by dual beam focused
ion beam scanning electron microscopy
D. Cédat a, O. Fandeur b, C. Rey a,⇑, D. Raabe c
a
Departement Mechanical Soils, Structures and Materials, Ecole Centrale Paris, F-92295 Châtenay-Malabry, France
b
CEA, DEN, DM2S, SEMT, LM2S, F-91191 Gif-sur-Yvette, France
c
Department Microstructure Physics and Metal Forming, Max Planck Institut für Eisenforschung, Düsseldorf, Germany

Received 16 July 2011; received in revised form 25 November 2011; accepted 27 November 2011

Abstract

The mechanical behavior of a Mo–TiC30 vol.% ceramic–metal composite was investigated over a wide temperature range (25–700 °C).
High-energy X-ray tomography was used to reveal percolation of the hard titanium carbide phase through the composite. Using a poly-
crystal approach for a two-phase material, finite-element simulations were performed on a real three-dimensional (3-D) aggregate of the
material. The 3-D microstructure, used as the starting configuration for the predictions, was obtained by serial sectioning in a dual beam
focused ion beam scanning electron microscope coupled to an electron backscattered diffraction system. The 3-D aggregate consists of a
molybdenum matrix and a percolating TiC skeleton. As for most body-centered cubic (bcc) metals, the molybdenum matrix phase is
characterized by a change in plasticity mechanism with temperature. We used a polycrystal model for bcc materials which was extended
to two phases (TiC and Mo). The model parameters of the matrix were determined from experiments on pure molydenum. For all tem-
peratures investigated the TiC particles were considered to be brittle. Gradual damage to the TiC particles was treated, based on an accu-
mulative failure law that is approximated by evolution of the apparent particle elastic stiffness. The model enabled us to determine the
evolution of the local mechanical fields with deformation and temperature. We showed that a 3-D aggregate representing the actual
microstructure of the composite is required to understand the local and global mechanical properties of the composite studied.
Ó 2011 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Metal–ceramic composite; Numerical simulation; Crystal plasticity; Polycrystal modeling; Damage

1. Introduction are needed at the grain scale, to obtain a better understand-

ing of damage initiation.
The aim of this study was to understand the mechanical In an earlier study Cédat et al. [4] simulated the local
behavior and damage evolution of a metal–ceramic com- mechanical fields using a three-dimensional (3-D) aggre-
posite (Mo–TiC30 vol.%) as a function of deformation and gate obtained from a random stack of layers of columnar
temperature in the range 800 °C up to 1000 °C. grains. These grains were characterized in terms of crystal
For compression tests performed at high temperature orientations, determined by electron backscattered diffrac-
damage observations cannot be monitored and simulations tion (EBSD). When using this aggregate the polycrystal
model failed to describe the mechanical behavior of the
composite. Different effects might be responsible for the
⇑ Corresponding author.
discrepancy between experiment and predictions, namely
E-mail address: [email protected] (C. Rey). microstructural size effects, higher complexity of the

1359-6454/$36.00 Ó 2011 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2011.11.055
1624 D. Cédat et al. / Acta Materialia 60 (2012) 1623–1632

required constitutive laws of the phases involved, and the 2.1. Chemical composition
spatial distribution (connectivity and topology) of the
TiC brittle phase. The chemical compositions of the powder are given in
The current paper addresses this problem of properly Tables 1 and 2.
predicting the micromechanical behavior of complex
metal–ceramic composites. 2.2. X-ray tomography results
The investigations in Section 2 (using tomography)
reveal percolation of the TiC phase. A new representative Using high-energy X-ray tomography a 3-D investiga-
aggregate has been built to take into account such phase tion of the material was performed at the European Syn-
percolation in the computation of the local mechanical chrotron Radiation Facility (ESRF, Grenoble, France).
fields, using tomographic microstructure data obtained by This characterization technique proceeds by taking a
dual beam focused ion beam scanning electron microscopy sequence of recordings of a sample from different angles.
(FIB-SEM). Then an appropriate tomography reconstruction algorithm
The polycrystal model used, implemented in finite-ele- builds a 3-D map of the microstructure based on inversion
ment code, is described in Section 3. Damage is represented of the X-ray attenuation coefficient projection paths that
by a decrease in the TiC effective elastic modulus imple- penetrated the material upon synchrotron illumination.
mented in the model. Fig. 2 shows a 3-D image of the microstructure of the
Section 4 presents the meshing and the boundary condi- Mo–TiC material obtained with the highest spatial resolu-
tions applied to the 3-D EBSD aggregate. New model tion (0.28 lm) available on the ID19 beam line (X-ray
parameters for Mo are determined by an inverse method. energy 65 keV) of the ESRF equipment.
Section 5 deals with the results of the mechanical predic- Fig. 2 represents the set of all points in the carbide phase
tions for the composite. Section 6 draws conclusions. (purple) that can be linked together to the bottom and top
faces of the parallelepiped specimen (30  30  30 lm), by
2. Material and experimental procedures paths entirely contained in the carbide phase. The micro-
structure of the composite revealed a percolating skeleton
The material investigated was obtained by powder met- of carbides embedded in a molybdenum metallic matrix,
allurgy synthesis [2,3] (hot isostatic pressing (HIP) at with a high 3-D connectivity of the carbide phase.
1600 °C). No homogenization heat treatment was per-
formed after HIP processing. 2.3. 3-D aggregate mapping by 3-D EBSD
The composite microstructure was characterized in a
previous paper [4] using different chemical and physical To accurately compute the local mechanical fields a new
methods. According to Cédat et al. [1] the composite exhib- representative 3-D aggregate had to be designed, taking
ited three phases: molybdenum (Mo), titanium carbide into account the observed percolation of the TiC phase.
(TiC) and a (Mo,Ti) C phase, which is a face-centered cubic Our approach to 3-D EBSD was inspired by the works
(fcc) structure with lattice parameters close to those of TiC of Uchic et al. [5], Zaefferer et al. [6], Bastos et al. [7] and
(Fig. 1). The hard particles (TiC) showed a core/shell (or Konrad et al. [8], as well as by the 3-D texture measure-
core/rim) structure, with molybdenum as the binder phase. ments using synchrotron radiation published by Larson
The third phase, identified as TiC–Mo15 at.% had the same et al. [9] and Yang et al. [10]. The 3-D EBSD experiments
structure as TiC. In this paper the mechanical behavior of were conducted at the Max-Planck Institute (Düsseldorf),
the third phase is assumed to be equivalent to that of TiC. using a joint high-resolution field emission SEM/EBSD
set-up together with a FIB system in the form of a Zeiss
cross-beam 3-D crystal orientation microscope. Details
are given in Larson et al. [9] and Yang et al. [10].
The analytical method applied in this study involves
highly precise and fully automated serial sectioning by
Ga+ FIB and subsequent mapping of the crystallographic
texture in each of those layers using high-resolution EBSD.
For more details on this method, see Zaafarani et al. [11]
and Zaefferer et al. [12].
The spatial resolution of the 3-D orientation and phase
pixel information retrieved amounts to 50  50  50 nm.
The ion beam did not create noticeable damage, i.e. no seri-
ous deterioration of the EBSD pattern quality could be
observed after milling. This stability of the material when
Fig. 1. SEM picture showing the microstructure of the Mo–TiC30 vol.% exposed to an ion beam is attributed to its high melting
composite revealing three phases, molybdenum (light grey), TiC (black), point and the absence of phase transformations. The EBSD
and (Mo,Ti) C (grey) [1]. measurements were carried out sequentially in each layer
 D. Cédat et al. / Acta Materialia 60 (2012) 1623–1632 1625

Table 1
TiC composition (mass%).
C O N Ca Co W Ni Al Fe S
19.23 0.6126 0.0279 0.002 0.032 0.39 <4E4 0.0014 0.0061 0.0019

Table 2
Molybdenum composition (mass%). are cylinders with a diameter D = 12 mm and cylinder
Mo O Fe K length L = 18 mm. The initial strain rate upon loading
99.98 0.0620 9 p.p.m. 29 p.p.m.
was about e_ = 5  104 s1.
The compression curves show a linear elastic stage, fol-
after FIB serial sectioning at a 300 nm step size between the lowed by a linear hardening stage, and, finally, a weak
abutting layers. The entire process of alternating FIB sec- strain hardening regime. The two strain hardening regimes
tioning and EBSD measurements was carried out for a change as a function of temperature and deformation. The
set of 22 layers to reconstruct the whole aggregate. Fig. 3 extension of the linear stage decreases with increasing total
shows SEM images of two different serial sections and deformation. Strain hardening decays with increasing
the entire 3-D microstructure reconstructed as a 3-D map. temperature.
Fracture surface observations reveal that at 25 °C the
2.4. Experimental analysis of the mechanical behavior of the two phases are characterized by a brittle behavior and that
composite and of the molybdenum phase at 300 °C the molybdenum matrix is ductile while the TiC
phase fails in a brittle manner. The mechanical properties
2.4.1. Composite of pure molybdenum have been previously studied by
The composite compression curves are given in Fig. 4 Cédat et al. [1] using two kinds of tensile tests, namely
for temperatures between 25 °C and 700 °C. The specimen strain rate change tests (5  104 and 5  102 s1) over

Fig. 2. 3-D representation of the phases obtained by X-ray tomography. (a) TiC; (b) Mo composite. Specimen size 30  30  30 lm.

Fig. 3. 30  30  6.6 lm aggregate obtained by 3-D electron orientation microscopy (3D EBSD). Top row, TiC and TiC–Mo15 at.% phases; bottom row,
whole aggregate. The color code is given by the crystallographic standard triangle. For simulations the loading direction is parallel to the ~
2 axis. (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
1626 D. Cédat et al. / Acta Materialia 60 (2012) 1623–1632

This polycrystal model was developed for bcc metals or

alloys taking into account temperature effects, as will be
outlined in more detail below. The two plasticity mecha-
nisms controlling plastic strain hardening of molybdenum
are double kink formation (low temperature) and interac-
tions between mobile and forest dislocations (high temper-
ature). In this paper we summarize the polycrystal model
extended to multi-phase composites. For more details see
Libert [18] and Libert et al. [14].

3.1. Single crystal constitutive law for the molybdenum

matrix
Fig. 4. Experimental composite compression behavior tests. r, tensile
stress; F, force; S(t), cross-section at each time interval; E22, strain.
For the bcc crystal structure of the molybdenum matrix
plastic glide occurs on the 24 slip systems {1 1 0}h1 1 1i and
{1 1 2}h1 1 1i. The chosen viscoplastic law can describe the
the temperature range 25–700 °C and temperature change competition between lattice friction and forest hardening.
tests. A transition in the polycrystal mechanical mecha- According to Kubin et al. [21], Rauch [22] and Tabourot
nisms was observed above 150 °C. For T > 300 °C the et al. [23] the slip rate on a given slip system (hereafter
viscoplastic flow weakly varied with temperature. These referred to by the index s) is given by:
results, confirmed by transmission electron microscopy  
DGðsseff Þ
(TEM) observations, suggested that molybdenum—like c_ s ¼ c_ 0 exp  sgnðss Þ ð1Þ
most body-centered cubic (bcc) metals—reveals different kT
plasticity properties at low and high temperatures. In bcc where DGðsseff Þ is the energy of activation for crystallo-
transition metals screw dislocations are less mobile than graphic slip of lattice dislocations, which is a function of
edge dislocations. Screw dislocation motion proceeds via the effective shear stress sseff . ss and c_ 0 are the resolved shear
thermally activated kink pair formation and expansion. stress on system s and the reference shear rate, respectively.
Optical microscopy (OM) and SEM observations [13] per- When T < T0 the shear rate c_ 0 depends on the geometry
formed on molybdenum surfaces as well as on fracture sur- of the segments of the screw dislocations. The energy of
faces after tensile testing confirmed these results. For activation is then associated with the double kink mecha-
T = 25 °C molybdenum rupture occurs by grain boundary nism and is given by a phenomenological description, as
decohesion. The fracture surfaces studied reveal features of proposed by Kocks et al. [24]:
brittle failure. For T = 200 °C grain boundary decohesion   s p q
s jseff j
and plastic glide is observed. For T = 300 °C deformation DGðseff Þ ¼ DG0 1  ð2Þ
sR
occurs by plastic glide and the fracture surface is character-
ized by ductile rupture features. where p and q are parameters describing the energy associ-
From these observations we concluded in our earlier ated with the Peierls barrier. In Eq. (2) sR is the effective
work that the plasticity mechanisms of this material change stress that is required to create a double kink at T = 0 K,
in the temperature range around 200 °C. One should note when the contribution of thermal activation is equal to
that this transition temperature T0 is different from the zero. This description assumes that the glide of dislocations
brittle to ductile transition temperature of molybdenum on a system s is activated when sseff is larger than sR.
(which is close to 25 °C).
3.2. Single crystal strain hardening law for the molybdenum
3. Polycrystal model matrix

In order to determine the local mechanical fields in the dif- Strain hardening evolution as a function of temperature
ferent phases and especially the location and origin of dam- depends on interactions between mobile dislocations with
age in the TiC phase a previous model [14] was upgraded and lattice friction and/or forest dislocations. Lattice friction
numerical simulations were conducted. The crystal plasticity generates short-range stresses described by the effective stress
model used was developed within the framework of large sseff . The contributions described by ssint and s0 are shear stres-
transformations [15–20] (small elastic distortion but large ses and represent the interaction between mobile and latent
lattice rotation) and was implemented in the Abaqus finite- dislocations. Thus the yield stress required to activate the
element code, using a UMAT user subroutine. glide system s is equal to the sum of the three contributions:
The polycrystal is composed of different grains, each one ss ¼ s0 þ sseff þ ssint ð3Þ
having its own size and crystallographic orientation. Each Considering a segment of a screw dislocation pinned by
grain is considered as a single crystal characterized by a two obstacles submitted to lattice friction s0, Rauch [22]
constitutive law presented below. proposed the following expression of the internal stress:
 D. Cédat et al. / Acta Materialia 60 (2012) 1623–1632 1627

2P
ðlbÞ u¼1;24 a
su u
q According to Eq. (9) two dislocations of the same sys-
ssint ¼ ð4Þ tem s may undergo annihilation as soon as they are sepa-
ss  s0
su
rated by a spacing below gc. Similar dislocation-based
where a is a component of the dislocation interaction ma- constitutive laws for crystal plasticity finite-element models
trix as proposed by Franciosi [25], which is due to the inter- were applied, as suggested by Ma et al. [17,27].
action strength between the slip systems s and u (self and
latent hardening). Latent hardening experiments [25], per- 3.4. Damage law for the titanium carbide particles
formed on single crystals under tensile testing, revealed
that the asu parameter depends on the strain value: they in- Damage in the compound starts by initiation of cracks
crease up to 0.5–1% straining, then reaches asymptotic val- within the brittle carbide phase. A simple criterion that
ues. In this paper we use the asymptotic values. qu is the rests on the local stress in the brittle phase is introduced
dislocation density on the glide system u. Merging Eqs. in the polycrystal model. The failure strength of ceramics
(3) and (4) and solving the resulting equation one obtains: under compression being approximately 10–15 times larger
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffii
sseff 1 h s 2 X than the tensile strength [28], the critical stress leading to
s
s ¼ s0 þ þ ðseff Þ þ 4ðlbÞ2 u asu qu ð5Þ damage initiation is given by the approximation:
2 2
rcompressive K IC
The general expression of the hardening law (Eq. (5)) is rc ¼ rtensile  with rtensile ¼ pffiffiffiffiffiffi ð10Þ
appropriate to describe continuous evolution of the shear 15 pa
stress as a function of temperature: where KIC is the toughness factor and a the mean size of the
In the case of low temperature behavior (T < T0) plastic- crack.
ity is governed by the reduced mobility of screw disloca- The results of the simulations show that the carbide
tions and the double kink mechanism, hence sint is phase is mainly under tension when the composite is
negligible compared with seff. Eq. (5) leads to: mechanically loaded, while the molybdenum phase is under
2P compression. For the carbide phase the chosen damage cri-
s s
ðlbÞ u¼1;24 asu qu
s ¼ seff þ ð6Þ terion is formulated as a measure of the gradual accumula-
sueff
tion of damage. It is based on evolution of the elastic law:
In the case of high temperature (T > T0) plasticity
e e
e ðtÞ ¼ C
e 0 ½1  DðtÞ
mainly depends on forest hardening and seff is negligible C ð11Þ
compared with the two other terms. Eq. (5) leads to: with
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X
ss ¼ ss0 þ lb u¼1;24
asu qu ð7Þ b
DðtÞ ¼ ½1  expðaðt  t0 Þ ð12Þ
a
At low temperatures plastic deformation occurs accord-
ing to two non-equivalent slip system families h1 1 1i{1 1 0} where t  t0 ¼ e=_e, and e and e_ are the local strain and local
e
e 0 is the initial
and h1 1 1i{1 1 2}. strain rate at a given point, respectively. C
e
e
elastic modulus tensor and C ðtÞ is “the effective modulus”,
3.3. Dislocation density evolution law which gradually decays as a function of the strain rate and
the strain. a and b are parameters that are arbitrarily cho-
The dislocation density evolution law is a generalization sen in order to create an abrupt decrease in the Young’s
of the relation proposed by Estrin and Mecking [26]. For modulus due to micro-crack formation. A positive critical
each slip system Eq. (8) describes the evolution of the 24 stress value rC is determined from (Eq. (10)). At each incre-
dislocation densities with the strain: ment of the simulation the criterion (Eq. (10)) is tested for
" pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P ffi # all elements of the TiC meshing. As soon as the criterion
s j_cs j 1 u–s q
u
s r P rc (r is the local equivalent stress) is verified within
q_ ¼ þ  gc ðT Þq ð8Þ
b Dgrain KðT Þ an element its effective modulus is decreased as given by
Eqs. (11) and (12). For D(t) = 1 (C(t) = 0) the element is
This expression is derived from the balance between disloca-
equivalent to a “hole”. Changes in the stiffness of some ele-
tion accumulation (Orowan’s relationship) andpdislocation
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
P u ments of the TiC phase contribute to that of the overall
annihilation which induces softening. KðT Þ= u–s q is
compound.
the dislocation mean free path, which increases with decreas-
ing temperature. As evolution of the dislocation density is
4. Boundary conditions, meshing and model parameters
weak at low temperature, we assume that K(T) increases with
decreasing temperature. The annihilation of dislocations is
The polycrystal model was implemented in Abaquse
controlled by the mean dislocation–dislocation spacing gc,
finite-element code using an UMAT subroutine. Most of
the temperature dependence of which is expressed here in
the molybdenum parameters are determined on the basis
terms of an Arrhenius law:
  of the experimental curves, using an inverse fitting method
Egc using coupling [18,20] between the Sidoloe and Abaquse
gc ðT Þ ¼ gc0 exp  ð9Þ
k B ðT Þ software packages. The initial density of the dislocations
1628 D. Cédat et al. / Acta Materialia 60 (2012) 1623–1632

before straining was determined by TEM. The parameters Table 4

used for molybdenum are given in Tables 3 and 4. Set of temperature-dependant parameters identified for molybdenum.
T DG(T) (eV) s0 (MPa) K gc (nm)
4.1. The aggregate 25 °C 0.35 125 440 20
150 °C 0.5 70 80 12
The polycrystalline aggregate is shown in Fig. 3. The 300 °C 0.97 70 50 26
500 °C 1.18 70 50 30
finite-element meshing is derived from the grid correspond- 700 °C 70 30 36
ing to the 3-D EBSD (0.3 lm step) measurements. It is
s0, shear stress due to lattice friction (Eq. (3)); DG(T) = DG(seff), slip sys-
built from 220,000 elements (C3D8R). Each element corre- tems energy of activation (Eq. (1)); K(T), material parameter associated
sponds to an original 0.3 lm square pixel of the material. with the dislocation mean free path (Eq. (8)); gc, dislocation–dislocation
The crystallographic microstructural properties of the annihilation distance (Eqs. (8) and (9)).
two phases are described in terms of the Euler angles for
each pixel of the two phases in the mesh, the TiC elastic
properties, and the molybdenum constitutive laws parame- Table 5
ters. Connectivity among the two phase sets and the perco- Parameters for titanium carbide (TiC) derived from the proposed model.
lation of TiC are thus automatically taken into account. E (GPa) m a b rc (MPa)
440 0.19 0.013 0.013 250
4.2. Boundary conditions E, Young’s modulus; m, Poisson coefficient; rc; rupture critical parameter
in tension.
The aggregate considered corresponds to one-eighth of
the whole computed structure. For symmetry reasons the
lower face of the crystalline aggregate, shown in Fig. 3, is
loaded in compression mode normal to direction ~ 2. The dis-
placement is u2 = 0 on the lower face. The other boundary
conditions are u3 = 0 on the face normal to the ~ 3 direction
and the parallel surface is free. The u2 applied displacement
corresponds to a strain rate of about 5  104 s1.
In addition to this set of parameters, this new investiga-
tion has identified temperature-dependent parameters by
an inverse method on the basis of experimental tensile
curves performed at 25–700 °C.
The titanium carbide elasticity is assumed to be isotro-
pic. Its features are given in Table 5.
According to our observations of the fracture surfaces
we assume perfect cohesion between TiC and the Mo
matrix.

5. Results and discussion

Fig. 5. Numerical and experimental compression curves for Mo,TiC and
Mo–TiC30 vol.% (T = 25 °C). r22, compression stress; e22, strain.
For pure molybdenum metal representing a polycrystal
matrix the simulation results delivered by the current poly-
crystal model have already been published [4]. The com- curves, up to the maximum stress. Compared with our first
pression test simulations were in good agreement with the simulations [1], carried out on crystalline aggregates (con-
experimental results. structed as superimpositions of random layers of columnar
grains), a significant improvement is observed in the cur-
5.1. Mo–TiC composite results for T = 25 °C rent full field 3-D predictions. This improvement is attrib-
uted to the more realistic 3-D mapping of the incipient
The numerical and experimental curves are compared in microstructure, where both percolation and damage to
Fig. 5. the TiC phase are taken into account.
At room temperature very good agreement is observed For each phase the computed average local stress hr22i
between the numerical (Fig. 5) and experimental (Fig. 4) versus the applied strain E22 are represented by dotted

Table 3
Set of parameters identified for molybdenum by Cédat et al. [1].
s0 auu = asu q0 gc0 Egc Dgrain sR c_ 0 DG0 p q
11 2 2 1
125 MPa 0.01 10 m 14 nm 2.17  10 eV 3 lm 498 MPa 10 1.1 eV 0.2 1.5
 D. Cédat et al. / Acta Materialia 60 (2012) 1623–1632 1629

lines. The molydenum curve presents a perfect plastic

behavior, whereas TiC shows a nonlinear behavior for
0.5% 6 E22 6 1.3%, followed by a maximum, then a
decrease. The perfect plastic behavior of the matrix corre-
sponds to hr22i = 590 MPa (close to the behavior of pure
metal Mo). The matrix yield stress is about 550 MPa and
the constant stress corresponding to the plateau is equal
to R22 = 620 MPa up to E22 = 0.15. The nonlinear TiC
curve corresponds to a decrease in the effective elastic mod-
ulus induced by damage.
The computed maps (Fig. 6) represent the evolution of
the distribution of local strain e22 in the grains of Mo
and TiC and of damage within the TiC particles. The local
strain (e22) distribution is homogeneous in TiC grains and
heterogeneous in large molybdenum grains. In TiC the
damage parameter D(t) (introduced in Eqs. (10)–(12))
increases with applied deformation and locally reaches 1
for an applied strain E22 ffi 10%. The value D = 1 corre-
sponds to rupture of the particles.
According to our numerical results we can propose the
following scheme for damage. During stage I (the elastic
stage) the TiC skeleton supports most of the deformation.
At the end of stage I and up to a compression strain rate of
3% (stage II) some bridges of TiC (normal to the compres-
Fig. 6. Local strain e22 and damage evolution under two applied strains
sion axes) are submitted to local tensile stress. As soon as E = E22 = 0.03 and E = E22 = 0.1 at room temperature (the ~ 2 axis
the critical stress criterion (Eq. (10)) is breached some corresponds to the compression axis). (a) Molybdenum phase; (b) TiC
bridges between TiC particles are damaged. During stage phase; (c) damage parameter.
II the damage parameter D increases, reaching 1 (rupture
of the bridge) for E22 ffi 10%. In stage III the composite
behaves as a matrix with large disconnected particles. A comparison between the curves for the composite
(Fig. 7a) and pure molydenum (Fig. 7b) reveals, for an
5.2. Composite results for high temperatures applied strain E22 = 0.09, a large increase in the macro-
scopic yield stress from 600 to 1450 MPa.
The numerical simulation and the corresponding exper- Fig. 7a shows, for E22 = 0.09, good agreement between
imental curves for high temperatures are given in Fig. 7. the numerical and experimental results. The difference DR22

Fig. 7. Experimental and numerical stress–strain curves on a 3-D numerical model and a real polycrystalline aggregate of the composite for five
temperatures. Experimental and numerical stress–strain curves on molybdenum are given in the inset.
1630 D. Cédat et al. / Acta Materialia 60 (2012) 1623–1632

Fig. 8. Distribution of the uniaxial stress r = r22 for E22 = 0.09. (a) TiC phase; (b) molydenum matrix; (c) pure molydenum.

between the macroscopic experimental stress is less than ning of the plastic stage arise from the material parameters
10 MPa. For small strains the agreement is only qualitative. determined for pure molybdenum. For pure molybdenum
For the composite the small differences between the exper- the simulated yield stresses are lower than the experimental
imental results and simulation observed at the very begin- ones (Fig. 7b). The material parameters for pure molydenum
 D. Cédat et al. / Acta Materialia 60 (2012) 1623–1632 1631

(Tables 3 and 4) were identified, for E22 P 0:2%, in simplified 6. Conclusions

stress–strain curves. Accurate identification would need a
higher number of material parameters. In addition, the dam- Understanding the mechanical behavior of a two-phase
age may be considered much too simple. In stage II the aver- Mo–TiC30 vol.% metal–ceramic composite over a wide tem-
age computed effective elastic modulus of the TiC bridges perature range requires simultaneous experimental and
decreases less than is actually observed. numerical studies. The first difficulty is to work on a realis-
The computed distributions of internal stress r22 for TiC tic representation of the material, then to take into account
(rTiC Mo
22 ) and Mo (r22C ) are given in Fig. 8a and b. In order to damage to the brittle phase and its effect on the ductile
understand the composite phase behavior the distribution phase. The composite studied presents small grains and a
rMo
22 for a pure molydenum aggregate is reported in Fig. 8c. brittle phase with percolation, the crystalline orientation
Whatever the temperature, the TiC behavior being of which is not easily observable. Mechanical polishing
assumed to be elastic or brittle, the distribution rTiC
22 is inde- and EBSD measurements cannot be used in this case and
pendent of the matrix behavior. In contrast, in the Mo aggregates built from random layers of grains lead to incor-
matrix the distribution rMo22C depends on both temperature rect results. Consequently, a dual beam FIB-SEM was used
and TiC behavior. to characterize and reconstruct a real 3-D polycrystalline
two-phase Mo–TiC30 vol.% metal–ceramic composite aggre-
5.2.1. Molydenum matrix gate layer by layer. Implementation of damage to the brit-
A comparison of the distributions rMo
22C (Mo matrix) and tle phase on the scale of the grain is required and provides
Mo
r22 (pure Mo) for E22 = 0.09 raises several important good results for the microscopic stress–strain fields when
points. applied to this 3-D aggregate. The TiC damage criterion
is described in terms of cumulative decay in the elastic stiff-
 The rMo22C distributions show high background noise ness of this phase as a function of straining. Very good
and wider distributions than rMo 22 . It should be noted agreement with the experimental curves is observed at
that some points in the Mo matrix are actually in 25 °C. The stage III behavior is accurately described for
traction. At these points the stress values can reach any temperature.
500 MPa and can lead to damage to the matrix. The proposed 3-D aggregate and the corresponding
 For low temperatures the rMo 22C distribution curves two-phase polycrystal model can be used to forecast the
present an asymmetry shape which disappears with behavior of any related metal–ceramic or metal–metal
increasing temperature. Our computation reveals composite at different temperatures, just by changes in
that at 25 °C, 150 °C and 300 °C molydenum grains the matrix and particles parameters.
close to inter-phase boundaries show very high inter-
nal stresses (rMo
22C 6 500 MPa). Acknowledgements
 The average values hrMo Mo
22C i and hr22 i are very close
and show the same evolution with temperature. The microstructure of the starting composite was char-
These values correspond to the macroscopic stress acterized at the Max Planck Institute Düsseldorf. The
of pure molydenum compression curves. authors are very grateful to Prof. S. Zaefferer for his kind
and helpful discussions. The Mo–TiC composite studied
was elaborated at the Comissariat à l’Energie Atomique.
5.2.2. TiC carbide The authors thank Drs M. Le Flem and J.L. Bechade
Although for E22 = 0.09 some TiC bridges are broken for giving us the material and for financial help with
(D(t) = 1), a small part (15%) of the TiC phase is still sub- the tomography experiments. The authors are grateful
mitted to local tensile stresses greater than the critical stress to Prof. J.H. Schmitt (ECP) for stimulating and helpful
(rc P 250 MPa). Damage evolution then slows down in discussions.
stage III. The TiC skeleton presents a number of micro-
cracks and the composite is reduced to a Mo matrix with References
large TiC inclusions. As a result the composite hardening
slope plateaus, before becoming negative. At high temper- [1] Cedat D, Libert M, Le Flem M, Fandeur O, Rey C, Clavel M, et al.
atures evolution of the damage parameter with applied Int J Refract Met Hard Mater 2009;27:267–73.
[2] Ettmayer P, Kolaska H. In: Bildstein H, Ortner HM, editors. 12th
deformation is slowed down compared with low tempera- Plansee seminar; 1989. p. 771–801.
tures, leading to an increase of the size of the plateau. [3] Le Flem M, Allemend A, Urvoy S, Cédat D, Rey C. J Nucl Mater
Damage is represented by decay of the effective elastic 2008;380:85–92.
modulus in some elements of the TiC mesh as soon as [4] Cédat D, Rey C, Clavel M, Schmitt JH, Le Flem M, Allemand A. J
Nucl Mater 2009;385:533–7.
the critical stress criterion is satisfied. The square elements
[5] Uchic MD, Groeber M, Dimiduk DM, Simmons JP. Scripta Mater
of the mesh are more representative of small holes than 2006;55:23–8.
actual thin micro-cracks. The increasing number of “holes” [6] Zaefferer S. Mater Sci Forum 2005;495-497:3–12.
in TiC is sufficient to explain the weak negative slope of the [7] Bastos A, Zaefferer S, Raabe D. J Micros 2008;230:487–98.
simulated stress–strain curves for large strains. [8] Konrad J, Zaefferer S, Raabe S. Acta Mater 2006;54:1369–80.
1632 D. Cédat et al. / Acta Materialia 60 (2012) 1623–1632

[9] Larson BC, Yang W, Ice GE, Bdai JD, Tischler JZ. Nature [17] Erieau P, Rey C. Int J Plast 2004;20:1763–88.
2002;425:887–90. [18] Ma A, Roters F, Raabe D. Acta Materiala 2006;54:2169–79.
[10] Yang W, Larson BC, Pharr GM, Ice GE, Tischler JZ. MRS Symp [19] Libert M. PhD thesis. Ecole Centrale Paris, France; 2007.
Proc 2003;779:W5.34.1. [20] Roters F, Eisenlohr P, Hantcherli L, Tjahjanto D, Bieler TR, Raabe
[11] Zaafarani N, Raabe D, Singh RN, Roters F, Zaefferer S. Acta Mater D. Acta Mater 2010;58:1152–211.
2006;54:1863–76. [21] Kubin LP, Louchet F, Vesely D. Philos Mag A 1978;38:205–21.
[12] Zaefferer S, Wright SI, Raabe D. Metal Mater Trans A [22] Rauch E. Key Eng Mater 1994;97/98:371–6.
2008;39A:374–89. [23] Tabourot L, Fivel M, Rauch E. Mater Sci Eng A 1997;234/
[13] Cédat D. PhD thesis, Ecole Centrale Paris, France; 2008. 235:639–42.
[14] Libert M, Rey C, Vincent L, Marini B. Int J Solids Struct [24] Kocks UF, Argon AS, Ashby MF. Prog Mater Sci 1976:19.
2011;48:2196–208. [25] Franciosi P. Acta Metall 1983;31:1331–42.
[15] Peirce D, Asaro RJ, Needleman A. Acta Metall 1983;31:1951–76. [26] Estrin Y, Mecking H. Acta Metall 1984;32:57–70.
[16] Teodosiu C, Raphanel JL, Tabourot L. Large plastic deformation. [27] Ma A, Roters F, Raabe D. Acta Mater 2006;54:2181–94.
Proceedings of the international seminar MECAMAT ’91; 1991. p. [28] Sadowski T, Samborski S. Comput Mater Sci 2003;28:512–7.
153–68.