0c96052039e4483a6d000000
DESCRIPTION
kTRANSCRIPT
7/18/2019 0c96052039e4483a6d000000
http://slidepdf.com/reader/full/0c96052039e4483a6d000000 1/7
See discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/37466386
Thermal conductivity of Al-SiC compositeswith monomodal and bimodal particle size
distribution
ARTICLE in MATERIALS SCIENCE AND ENGINEERING A · MAY 2008
Impact Factor: 2.41 · DOI: 10.1016/j.msea.2007.07.026 · Source: OAI
CITATIONS
33
DOWNLOADS
460
VIEWS
166
5 AUTHORS, INCLUDING:
J. Narciso
University of Alicante
84 PUBLICATIONS 970 CITATIONS
SEE PROFILE
Ludger Weber
École Polytechnique Fédérale de Lausanne
78 PUBLICATIONS 902 CITATIONS
SEE PROFILE
A. Mortensen
École Polytechnique Fédérale de Lausanne
237 PUBLICATIONS 2,830 CITATIONS
SEE PROFILE
Enrique Louis
University of Alicante
287 PUBLICATIONS 3,328 CITATIONS
SEE PROFILE
Available from: Ludger Weber
Retrieved on: 08 July 2015
7/18/2019 0c96052039e4483a6d000000
http://slidepdf.com/reader/full/0c96052039e4483a6d000000 2/7
Materials Science and Engineering A 480 (2008) 483–488
Thermal conductivity of Al–SiC composites with monomodaland bimodal particle size distribution
J.M. Molina a,b,c,∗, J. Narciso b,d, L. Weber c, A. Mortensen c, E. Louis a,b
a Departamento de Fısica Aplicada, Universidad de Alicante, Apartado 99, E-03080 Alicante, Spainb Instituto Universitario de Materiales de Alicante (IUMA), Universidad de Alicante, Apartado 99, E-03080 Alicante, Spain
c ´ Ecole Polytechnique F´ ed´ erale de Lausanne, EPFL, Laboratory of Mechanical Metallurgy, CH-1015 Lausanne, Switzerland d Departamento de Quımica Inorg´ anica, Universidad de Alicante, Apartado 99, E-03080 Alicante, Spain
Received 22 January 2007; received in revised form 11 July 2007; accepted 16 July 2007
Abstract
The thermal conductivity of aluminum matrix composites having a high volume fraction of SiC particles is investigated by comparing data
for composites fabricated by infiltrating liquid aluminum into preforms made either from a single particle size, or by mixing and packing SiC
particles of two largely different average sizes (170 and 16m). For composites based on powders with a monomodal size distribution, the thermal
conductivity increases steadily from 151W/mK forparticlesof averagediameter8 mto216W/mKfor170m particles. For thebimodal particle
mixtures the thermal conductivity increases with increasing volume fraction of coarse particles and reaches a roughly constant value of 220 W/m K
for mixtures with 40 or more vol.% of coarse particles. It is shown that all present data can be accounted for by the differential effective medium
(DEM) scheme taking into account a finite interfacial thermal resistance.
© 2007 Elsevier B.V. All rights reserved.
Keywords: Thermal conductivity; Size-effect; Al–SiC; Bimodal mixtures; Modeling
1. Introduction
The use of high volume fraction metal matrix composites
(MMCs) in multifunctional electronic packaging is steadily
increasing [1–5]. Thekey requirements for this application are a
high thermal conductivity anda coefficient of thermal expansion
(CTE) similar to that of materials commonly used in micro-
electronic systems. Among the different composites that are
nowadays being considered for this application, Al/SiC com-
posites with a high volume fraction of particulate reinforcement
are a suitable and versatile option. Due to the large difference
in CTE between the Al matrix and the SiC particles, the overall
CTE of the composite depends strongly on the particulate con-tent. By adjusting the overall SiC content, mainly by varying the
ratio between coarseandsmall particles, theCTE canbe tailored
in a certain range.
∗ Corresponding author at: Departamento de Fısica Aplicada, Universidad de
Alicante, Apartado 99, E-03080 Alicante, Spain. Tel.: +34 96 590 3400x2055;
fax: +34 96 590 3464.
E-mail address: [email protected] (J.M. Molina).
In previous contributions [6,7], high volume fraction com-
posites were produced by infiltration of liquid aluminum into
preforms made by mixing and packing SiC particles with aver-
age diameters of 170 and 16m. Themaximum particle volume
fraction, near 0.74, was attained for a mixture containing 67%
coarse particles. As shown in Ref. [7], these composites have
a reasonably low coefficient of thermal expansion, in the range
of 7.8–10.8ppm/K, and are hence suitable for application as
substrates in microelectronics. Best among these are compos-
ites with the maximum particle volume fraction, attained using
a bimodal particle size distribution.
The thermal conductivity of these composites is mainly gov-
erned by the conductivity of the individual phases, their volumefraction and shape, and also by the size of the inclusion phase
due to a finite metal/ceramic interface thermal resistance. There
are quite a few analytical models in the literature [8–15] that
can account for a finite interface thermal conductance, h (the
inverse of the interface thermal resistance). Many of these are
quite straightforward extensions of results obtained for compos-
ites with a matrix/inclusion topology having an ideal interface,
in which one replaces the intrinsic inclusion conductivity κinr by
an effective inclusionthermal conductivity, κeff r that incorporates
0921-5093/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.msea.2007.07.026
7/18/2019 0c96052039e4483a6d000000
http://slidepdf.com/reader/full/0c96052039e4483a6d000000 3/7
484 J.M. Molina et al. / Materials Science and Engineering A 480 (2008) 483–488
the effect of inclusion size. κeff r is defined knowing the radius a
of the equivalent sphere and the interface thermal conductance,
h, according to
κeff r =
κinr
1+ κinr /ha
or κeff r =
κinr
1+ 1/Bi=
κinr
1+ B, (1)
where Bi is the relevant Biot number characterizing the rein-forcement and B is its inverse as introduced by Clyne [16].
One of the most widely used models is that proposed by Has-
selman and Johnson [8]. This model has often been applied to
compositeswith a monomodalparticle sizedistribution [17–21].
For composites with a bimodal particle size distribution, the
application of the Hasselman and Johnson (HJ) model is not
straightforward, because of the need to choose a single charac-
teristic length scale. A way of tackling the problem is to define
the characteristic particle size via the overall specific surface
area of the preform, e.g. [22].
We propose in this contribution that for bimodal mixtures,
the two populations of particles with distinct sizes should be
treated as two different particles with distinct effective thermal
conductivities according to the respective Biot number of the
two sizes used. We develop the leading equations for this case
in the framework of the differential effective medium (DEM)
scheme, which has recently been shown to be more robust than
the Hasselman and Johnson equation [23]. We then predict how
the composite thermal conductivity should vary in a composite
with a bimodal particle size distribution and a variable ratio of
small-to-coarse particles. We confront this proposition to two
series of experiments. From a first series on composites made of
aluminum and SiC with a monomodal particle size distribution
but variable particle size, we determine the relevant parameters
of the system, i.e. κinr and h for the Al/SiC couple. The thermalconductivity of the matrix is inferred from measurements of its
electricalconductivity. The second series of composites features
a bimodal particle size distribution, with various ratios of small-
to-coarseparticles. Data from this secondseriesare then directly
confronted to the prediction of the DEM scheme for two distinct
inclusion phases.
2. Analysis
As has been shown previously, the differential effective
medium approach is well-suited to model conduction properties
of two-phase composites [23–25]. We limit our analysis to thechanges in the standard DEM theory needed to account for two
or several types of spherical inclusions having distinct effective
conductivities. The leading differential equation is
dκ
dV ∗p= κ
i
f i−κ− κeff
ri
κ− κeff
ri
S − κ
, (2)
with f i the fraction of type i of reinforcing phase in the blend
used to reinforce the composite (and, hence,i
f i = 1), and S
the polarization factor of a sphere, i.e. 1/3. V ∗p is the volume
fraction integration variable related to the effective final volume
fraction of reinforcement, V p, by V ∗
p = −ln(1− V p). We note
in passing that the extension to randomly oriented ellipsoidal
inclusions is not as simple as for the case of “perfectly bonded”
inclusions having no interfacial thermal barrier. We limit our-
selves to the observation that at relatively low phase contrast,
the aspect ratio of the inclusion is of minor importance. We
also note that a possible influence of particle to particle contact
points, which may ease heat conduction somewhat by creating
bridges between particles or slow conduction when partly infil-
trated, is not taken into account here other than by the choice of
mean-field model. Separation of variables leads to the integral
equation
κc κm
dκ
κi
f i−κ− κeff
ri
/κ− κeff
ri
S − κ
= − ln1− V p
.
(3)
In general, the integral on the left-hand side has no analytical
solution; however, Eq. (3) can be solved numerically for κc.
For the special case of a bimodal particle mixture, one could
alternatively proceed by first introducing the small particles in
thematrix yielding a new “effective” matrix, in which thecoarse
particles are introduced in a second step. Let V p be again the
overall volume fraction of inclusion phase and f s and f 1 be the
fraction of small and coarse particles, respectively. The fraction
V mps of small particles in the new “effective” matrix is given by
V mps=
f sV p
(1− f lV p) (4)
and its thermal conductivity κ∗m according to the DEM scheme
is
1− V mps
=
(κs/κm)− (κ∗m/κm)
(κs/κm)− 1
κ∗m
κm
−1/3
, (5)
with κs the effective thermal conductivity of the small particles.
The thermal conductivity of the composite is then given by
1− f lV p
=
(κl/κ∗m)− (κc/κ
∗m)
(κl/κ∗m)− 1
κc
κ∗m
−1/3
, (6)
where κl stands for the effective thermal conductivity of the
coarse particles. This second approach has the advantage of yielding an implicit analytical solution for κc whereas the first
approach requires numerical integration. For the low phase con-
trast case considered here, it turns out that both approaches yield
the same result, i.e. the numerical values of the predicted com-
posite conductivity differ by less than 0.01 W/m K. We note in
passing that the two-step procedure would also be a way around
the need for choosing a single characteristic length scale in
the Maxwell approximation-based Hasselman and Johnson for-
mula. For low phase contrast this gives indeedvirtually thesame
result as the two DEM-based solutionspresented above. At high
phase contrast, however, the use of a Maxwell mean field based
approach may lead to erroneous results [23,24].
7/18/2019 0c96052039e4483a6d000000
http://slidepdf.com/reader/full/0c96052039e4483a6d000000 4/7
J.M. Molina et al. / Materials Science and Engineering A 480 (2008) 483–488 485
Table 1
The main characteristics of monomodal composites investigated in this work: V p is the particle volume fraction, D is the average size ( D(4,3)) in m determined by
laser scattering, P0 is the threshold pressure for infiltration (kPa) taken from [6] or estimated (marked with *), TC is the thermal conductivity (W/mK), and EC is the
electrical conductivity (106 −1 m−1)
Mesh of SiC in MMC V p (−) D (m) P0 (kPa) TC (W/m K) EC (106 −1 m−1)
F100 0.58 167 278 221 5.9
F180 0.58 86.4 340* 209 5.97
F240 0.6 56.8 400* 203 5.81F320 0.59 37.1 500* 204 5.74
F400 0.58 23.4 600* 194 6.57
F500 0.55 16.9 778 193 8.14
F800 0.53 8.9 2170 154 8.62
3. Experimental procedures
3.1. Processing of composites
Composites were made from 99.99% pure aluminum from
Norsk Hydro, Grevenhorst,Germany, andgreen-SiC(>99.5%)
supplied by Navarro SiC S.A, Cuenca, Spain, in nominal meshsizes F100, 180, 240, 320, 400, 500, and800. Thecorresponding
mean diametersof theSiCparticulate aregiven in Table1. Parti-
cles arerounded(asclassified by thesupplier) with a (low)aspect
ratio of 1.9 as obtained by image analysis. To clean the powders
from milling residues, the powder was repeatedly washed in
pure ethanol, separating the fine part of the powder by sedimen-
tation of the coarse particles. Particle mixing was carried out
in ethanol by means of the method described in Ref. [6]. Parti-
cles were packed into quartz tubes 16.5mm of inner diameter by
repeatedly addinga small amountof powder that was compacted
by tapping with well-defined impacts. To this end a plunger was
placed on the preform and was hammered by a weight falling
freely over a distance of 35 mm. Vibrations were intermittently
applied. The number of applied strokes was increased roughly
linearly with the height of the already packed preform. Not all
compacts were prepared with the same procedure: in the case
of coarse particulates (mesh 180 and 100) only vibrations were
used to avoid particle breaking, while in bimodal mixtures only
strokes were applied to reduce segregation. The particle volume
fraction was calculated by measuring the preform weight and
dimensions (height and diameter).
The composites were manufactured by gas pressure assisted
liquid metal infiltration. A metal ingot was placed on top of
the particle compact. In order to avoid reaction between melt
and preform prior to infiltration, an alumina fabric was placedbetween the preform and the metal ingot.
Thequartz crucible containing thepreform and themetal was
introduced in an induction-heated cold-wall infiltration appara-
tus. Homogeneous heating was ensured by means of a graphite
susceptorplacedaround thecrucible.Afterhavingreacheda vac-
uum of about 0.3 mbar in the infiltration vessel, heating was set
to approximately 250 ◦C to allow for slow desorption of humid-
ity and gas adsorbed on the particles while continuously pulling
the vacuum. Subsequently, heating was continued up to 750 ◦C
at a rate of approximately 5 ◦C/min. The maximum tempera-
ture was maintained for 30min followed by argon injection to
reach a set pressure in the range of 8–50 bars. More precisely,
the infiltration pressure was chosen for all preforms to be about
five times the respective threshold pressure reported in Table 1,
based on previously published data [6].
Porosity was estimated by calculating the difference between
the theoretical and the actual composite density measured by
Archimedes principle. The porosity level was below 1%, in all
composites,whichwaswithin uncertainty in thepreform densityestimated as described above.
3.2. Measurement of electrical and thermal conductivity
The electrical conductivity of the composites was measured
using anEddy-current apparatus. Their thermalconductivitywas
measured by means of a relative steady-state technique, in an
experimental set up designed and assembled at EPFL. In this, a
sample of diameter around 16mm and length 40mm is clamped
between a water-cooled block and a brass reference sample
which is in its turn connected to a thermally stabilized hot water
bath. Losses through radiation and convection are neglected;
thus, we take the heat flow through the reference and the sample
to be equal. The system was calibrated against pure aluminum
(99.99 wt.%) and pure copper (99.9998%), having a thermal
conductivity at 40 ◦C of 236 and 400 W/m K, respectively. The
temperaturegradient in thereferenceand in thesamplewas mea-
sured by three, respectively two thermocouples. Thus, linearity
of thetemperaturegradient in thereference could bemeasuredas
well; this was typically within±1%. Accounting also for uncer-
tainty on the geometry of the sample (diameter after grinding,
precise position of the thermocouples, thermal conductivity of
the reference) we evaluate the overall uncertainty of the mea-
sured thermal conductivities to within ±5% of the indicated
value.
4. Results
After infiltration the composite typically stuck to the surface
of the quartz tube, such that quartz pieces had to be ground
off. This indicates that there might have been some reaction
between the quartz crucible and aluminum. Metallographic cuts
performed on the remaining metal on top of the composite after
infiltration revealed indeed the presence of minor amounts of
silicon-rich phase in the matrix, cf. Fig. 1a. Since no aluminum
carbide was observed around the SiC particles by light opti-
cal microscopy, reaction with the crucible rather than reaction
7/18/2019 0c96052039e4483a6d000000
http://slidepdf.com/reader/full/0c96052039e4483a6d000000 5/7
486 J.M. Molina et al. / Materials Science and Engineering A 480 (2008) 483–488
Fig. 1. Optical micrographs of cross sections of different samples: remaining metal on top of the infiltrated preform after infiltration (a) and composites containing
67% (b), 100% (c) and 75% (d) of coarse particles.
with the SiC is most likely the origin of the observed Si in the
matrix.
Metallography shows a homogeneous structure for both the
monomodal and the bimodal composites, and also the effective-
ness of the sedimentation cleaning procedure, cf. Fig. 1b and c.
The only composite where inhomogeneity at the mesoscale was
observed is that with the 3:1 mixture of coarse to small particles.
In this composite, the quantity of small particles was not suffi-
cient to fill all the space left between touching coarse particles,
resulting in variable degree of filling of these spaces with small
particles, as shown in Fig. 1d.
Table1 reportsexperimentaldata forparticle volumefraction,
average SiC particle diameter, and electrical and thermal con-
ductivity of the composites based on monomodal particle size
distributions used in this work. It is seen that all preforms have
similar volume fractions, with a slight tendency towards largervalues at larger particle size. The results for the particle volume
fractionincompactsofmixturesof100and 500mesh particlesin
differentrelative amountsare shownin Table2. Thevolume frac-
tion is maximal for 67% of coarse particles. Table 2 also gives
the experimental values of thermal and electrical conductivity
in composites based on bimodal mixtures.
5. Discussion
The microstructures shown in Fig. 1b–d indicate that the
present composite materials have indeed a regular and well-
defined microstructure; as such these can serve as model
materials suitable for confrontation with theory. We first ana-
lyze the data of the series based on monomodal particle size
distribution.
In order to reducethenumberof “free”parameters to account
for the experiments, the thermal conductivity of the matrix was
independently deduced from the electrical conductivity of the
composite. Indeed, knowing that in metal–insulator composites
the electrical conductivity scales directly with the matrix con-
ductivity, one can back-calculate the electrical conductivity of
the matrix using the differential effective medium scheme to
account for the non-dilute volume fraction of the electrically
insulating disperse phase. The DEM was indeed shown to be
well-suited for the prediction of composite electrical conduc-
tivity based on matrix conductivity at large volume fractions of
insulatingdispersephase[24]. In theDEMscheme,with insulat-
Table 2
The main characteristics of bimodal composites materials (obtained from par-
ticles F100 and F500 of Table 1): V p is the total particle volume fraction,
TC is the thermal conductivity (W/mK), and EC is the electrical conductivity
(106 −1 m−1)
Coarse (%) V p (−) TC (W/m K) EC (106 −1 m−1)
0 0.55 193 8.14
25 0.62 215 5.4
50 0.69 220 4
67 0.74 228 2.72
75 0.72 225 2.82
100 0.58 221 5.9
7/18/2019 0c96052039e4483a6d000000
http://slidepdf.com/reader/full/0c96052039e4483a6d000000 6/7
J.M. Molina et al. / Materials Science and Engineering A 480 (2008) 483–488 487
Fig. 2. Linear regression of log(σ c) vs. log(1−V p) of composite electrical con-ductivity in order to determine σ m and n to 2.84× 107 −1 m−1 and 1.70,
respectively.
ing inclusions, the composite conductivity, σ c, is simply linked
to the electrical conductivity of the matrix, σ m, by
σ c = σ m(1− V p)n, (7)
where n is a parameter reflecting the average particle shape, cf.
[24,25]. For typical angular particles produced by grinding, n
takes a value of 1.6–1.75. For the specific case of an aspect ratio
of 1.9 measured in 2D micrographs one expects n≈ 1.62 [25].
The experimental results for the electrical conductivity of both monomodal and bimodal composites have been fitted with
Eq. (7) using n and σ m as parameters. The fitting was car-
ried out by linear regression in a plot of log(1 −V p) versus
log(σ c), given in Fig. 2. The values of the fitted parameters are
n = 1.70 and σ m≈ 2.84× 107 −1 m−1, to be compared with
n = 1.62 expected from metallography and the conductivity of
pure aluminum, namely, 3.7× 107 −1 m−1 (had we taken the
Maxwellmean-fieldmodelinsteadof theDEMto back-calculate
the matrix electrical conductivity, the back-calculated matrix
conductivity would have been lower by a further 10–20%).
Compared to pure Al the electrical conductivity of thematrix
is thus approximately 20% too low. This can be rationalized by
the presence of Si in solid solution due to the reaction with thequartz crucible as mentioned above. Using data compiled by
Bass [26], this would correspond to a content of 1.0 at.% of Si
in solid solution in the Al matrix; this is somewhat high but not
implausible if one considers that some of the Si is present in
the form of precipitates. Using the Lorentz number for Al at
T = 300 K, namely, κ /(σ T )≈ 2.16× 10−8 W /K2 [27], we then
estimate thematrix thermal conductivity to beκm≈ 185 W/m K,
also about 20% below that of pure Al.
As shown in Fig. 3, the thermal conductivity of the compos-
iteswith a monomodalparticlesizedistributioncanbeaccounted
for by a single pair of parameters κinr and h taking the values of
253 W/m K and 7.5× 107
W/m2
K, respectively. These values
Fig.3. Experimentalthermalconductivityof theSiC/Alcompositesproduced byinfiltration of aluminum into preforms of monomodal SiC particles of variable
average size compared to predicted values based on the differential effective
medium scheme including correction to find effective inclusion conductivities
according to Eq. (1). The values taken for the intrinsic thermal conductivity of
the inclusion, κinr , and the thermal interface conductivity, h, are 253 W/mK and
7.5 107 W/m2 K, respectively.
are in agreement with data found in the literature [19,22,28], yet
– at least for the interfacial thermal conductance – at the lower
end of the reported spectrum.
We cannow confront thepredictionsbasedon Eq.(3) withthe
experimental data gathered on the bimodal composites. As can
beseen in Fig. 4, the parametersetdetermined beforegivequali-
Fig. 4. Thermal conductivity of the SiC/Al composites produced by infiltra-
tion of aluminum into preforms of bimodal SiC mixtures (SiC-500/SiC-100) vs.
the percentage of coarse particles. The circles are the experimental results; the
dashed line is the calculation using the values of κinr and h derived to fit all com-
posites with monomodal particle size distribution, while the full line represents
the prediction based on the deterministic pair κinr and h that fits the conductivity
of monomodal composites with particle sizes used in the bimodal mixture.
7/18/2019 0c96052039e4483a6d000000
http://slidepdf.com/reader/full/0c96052039e4483a6d000000 7/7
488 J.M. Molina et al. / Materials Science and Engineering A 480 (2008) 483–488
tatively the experimentally observed evolution as the fraction of
coarseandsmall particlesvaries, yet theexperimentalvalues are
systematically larger thanthe calculated onesincluding thosefor
the twocomposites based on monomodalparticles. If we instead
use the – deterministic – values of κinr and h to fit the two com-
posites with monomodal particle size distribution of F100 and
F500, namely 259 W/m K and 1.05× 108 W/m2 K, respectively,
the predictions come in better agreement with the experimental
data, Fig.4. Thedifference between thetwosets of κinr and h may
be due to the fact that the smallest monomodal powder (F800)
was included into first above analysis: this particular composite
hasa surprisingly lowthermalconductivity(seeFig.3). Onemay
argue that for low particle sizes the silicon content in the matrix
might behighersincetherewas more SiC-surface fordissolution
available and, hence, the matrix conductivity might be lowered
further; however, the electrical conductivity measurements do
not confirm this assumption.
Withthedataathandwemightspeculateaboutwhattheupper
limit of achievable thermal conductivity of Al–SiC at room tem-
perature might be. We first note that the size of the employedparticles is presumably at its upper technical limit. Hence, an
increase in the particle effective conductivity might be difficult
to achieve. Furthermore, this parameter is limited to the intrin-
sic conductivity of technical SiC powder, i.e. 250–260W/m K.
Assuming therefore that the thermal conductivity of the SiC
filler cannot be further improved and that the interface ther-
mal conductance between Al and SiC is indeed of the order
108 W/m2 K (in agreement with values calculated using the
acoustic mismatch model [29]), the largest potential is obvi-
ously in improving the matrix conductivity to the value of pure
aluminum. This could be done by avoiding alloying with sili-
con both by using appropriate crucible or mold material and byreducing the contact time between melt and particles. Both can
be achieved by squeeze casting such composites. If we assume
that thematrix conductivity canbe maintained at thelevel ofpure
aluminum we may then expect the thermal conductivity of the
Al–SiC material to reach 250W/m K. For higher performance,
either better conducting matrices (e.g. copper or silver) or filler
materials (e.g. cBN, graphite or diamond) must be employed.
6. Conclusions
(1) The thermal conductivity of Al/SiC composites based on
bimodal powder mixtures canreasonably well be accounted
for by using the differential effective medium scheme inwhich thetwo particle populations are treated as twodistinct
types of inclusion.
(2) The finite interface thermal conductance at the parti-
cle/matrix interface is then taken into account by attributing
to theinclusionsofdifferentsizesa distincteffective thermal
conductivity.
(3) Based on this analysis the peak for thermal conductivity of
Al/SiC composites with bimodal particle sizes distributions
can be estimated as 250W/m K.
Acknowledgements
The authors acknowledge partial financial support from the
Spanish “Ministerio de Educacion y Ciencia” (grant MAT2004-
03139) the “Generalitat Valenciana” (grants GRUPOS03/092
and GRUPOS03/212) and the Universidad de Alicante. Work of
LW and AM was funded by Laboratory core funding at EPFL.
References
[1] G. Lefranc, H.P. Degischer, K.H. Sommer, G. Mitic, In: T. Masard, A.
Vautrin (Eds.), ICCM-12,Woodhead Publishing Limited,Paris,1999(elec-
tronic support).
[2] C. Zweben, JOM 44 (7) (1992) 15–23.
[3] C. Zweben, JOM 50 (6) (1998) 47–51.
[4] D.D.L.Chung, C. Zweben, in: A. Kelly, C. Zweben (Eds.), Comprehensive
Composite Materials, vol. 6, Elsevier Science Ltd., 2000, p. 701.[5] C. Zweben, Power El. Tech. 32 (2) (2006) 40–47.
[6] J.M. Molina, R.A. Saravanan, R. Arpon, C. Garcıa-Cordovilla, E. Louis, J.
Narciso, Acta Mater. 50 (2002) 247–257.
[7] R. Arpon, J.M. Molina, R.A. Saravanan, C. Garcıa-Cordovilla, E. Louis, J.
Narciso, Acta Mater. 51 (2003) 3145–3156.
[8] D.P.H. Hasselman, L.F. Johnson, J. Compos. Mater. 21 (1987) 508–515.
[9] Y. Benveniste, T. Miloh, Int. J. Eng. Sci. 24 (1986) 1537–1552.
[10] F.C. Chen, C.L. Choy, K. Young, J. Phys. D: Appl. Phys. 9 (1976) 571–
586.
[11] A.J. Markworth, J. Mater. Sci. Lett. 12 (1993) 1487–1489.
[12] T.D. Fadale, M. Taya, J. Mater. Sci. Lett. 10 (1989) 682–684.
[13] H. Cheng, S. Torquato, Proc. R. Soc. Lond. A 453 (1997) 145–161.
[14] M.L. Dunn, M. Taya, J. Appl. Phys. 73 (1993) 1711–1722.
[15] R. Lipton, B. Vernescu, Proc. R. Soc. Lond. A 452 (1996) 329–358.
[16] T.W. Clyne, in: A. Kelly, C. Zweben (Eds.), Comprehensive CompositeMaterials, vol. 3, Elsevier Science Ltd., 2000, p. 447.
[17] D.P.H. Hasselman, J. Am. Ceram. Soc. 85 (2002) 1643–1645.
[18] D.P.H. Hasselman, K.Y. Donaldson, A.L. Geiger, J. Am. Ceram. Soc. 75
(1992) 3137–3140.
[19] D.P.H. Hasselman, K.Y. Donaldson, J.R. Thomas, J. Compos. Mater. 27
(1993) 637–644.
[20] A.G. Every, Y. Tzou, D.P.H. Hasselmann, R. Raj, Acta Met. Mater. 40
(1992) 123–129.
[21] A.L. Geiger, D.P.H. Hasselman, P. Welch, Acta Mater. 45 (1997)
3911–3914.
[22] J.M. Molina, E. Pinero, J. Narciso, C. Garcia-Cordovilla, E. Louis, Curr.
Opin. Solid State Mater. 9 (2005) 202–210.
[23] R. Tavangar, J.M. Molina, L. Weber, Scripta Mater. 56 (2007) 357–360.
[24] L. Weber, J. Dorn, A. Mortensen, Acta Mater. 51 (2003) 3199–3211.
[25] L. Weber, C. Fischer, A. Mortensen, Acta Mater. 51 (2003) 495–505.[26] J. Bass, in: J. Bass, K.H. Fischer (Eds.), Resistivity in Dilute Alloys,
Landolt-Bornstein New Series III/15a, Metals: Electronic Transport Phe-
nomena, Springer Verlag, Berlin, 1982, p. 166.
[27] N.W. Ashcroft, N.D. Mermin, Solid State Physics, Saunders College Pub-
lishing, 1976, pp. 21.
[28] C.W. Nan, X.P. Li, R. Birringer, J. Am. Ceram. Soc. 83 (2000) 848–854.
[29] E.T. Swartz, R.O. Pohl, Rev. Mod. Phys. 61 (1989) 605–668.