Professional Documents
Culture Documents
Composite Homogenization Via The Equivalent Poly-Inclusion Approach - FERRARI
Composite Homogenization Via The Equivalent Poly-Inclusion Approach - FERRARI
Composite Homogenization Via The Equivalent Poly-Inclusion Approach - FERRARI
oo
Printed in Great Britain. 0 1993 Pergattton Press Ltd
MAURO FERRAIU
Department of Civil Engineering, and Department of Materials Science and Mineral Engineering,
University of California, Berkeley, CA 94720, U.S.A.
Abstract-For the efficient use and development of advanced materials it is necessary to develop
simple yet accurate analytic relationships linking their average physical response to as large a set
as possible of microstructural quantifiers. With this objective in mind, the theory of the effective
elastic response of biphase composite materials, with arbitrary second-phase geometry, concentra-
tion, and orientation distribution, is examined here. A formulation is presented, which employs
an equivalent poly-inclusion methodology, in the sense that its governing variable is the average
inclusion strain concentrator for an entirely uniform body undergoing a uniform eigenstrain in
a non-dilute family of morphologically identical internal regions (inclusions). Admissibility
criteria for approximate inclusion strain concentrators are formulated. The relation of the present
approach with the traditional one, based on the inhomogenity average strain concentrator, is
investigated, and some deficiencies of current methods based on the latter approach are discussed.
A family of fully admissible poly-inclusion concentrators is proposed, which differs by a single
scalar-valued function of the volume fraction-the so-called “interaction function”. A specific
interaction function is finally identified, which satisfies all admissibility criteria, is consistent with
the known analytic solutions for biphase composites, and reproduces literature data to within
experimental error.
1. INTRODUCTION
The “effective” or “equivalent” stiffness tensor C of a composite material is defined
through the relation
6 = C&O (1)
where so is a homogeneous deformation applied at the boundary and b is the resulting
average stress.
For a biphase composite, consisting of a matrix of stiffness Cm, containing a
volumetric concentration (Y of inhomogeneities, or fibers, of stiffness Cf, the effective
stiffness may be exactly expressed as
C = Cm + a((Cf - C*)A) (2)
where the angular brackets denote orientational averaging, weighted by a fiber orientation
probability density function (Ferrari and Johnson, 1989). In eqn (2), the (average) strain
concentration tensor A was introduced. This is defined by the equation:
where so is as in (l), a superscript f refers to the fiber phase, and overbars denote spatial
averaging.
Eshelby (1957) showed the strain concentrator for the limit case of a composite
consisting of a single anisotropic ellipsoidal inhomogeneity in an infinite matrix to be
T = [I + E(C*)-‘(C’ - Cm)]-’(4)
where I is the fourth rank symmetric identity tensor, and E is defined in eqn (8) below.
At non-dilute volumetric fiber concentrations, T is no longer an appropriate con-
centrator. The determination of the adequate concentrator A is essential, for the purpose
of homogenization, in view of the fact that, once the fiber orientation profile and the
tensor A are known, the homogenization problem is reduced to a mere computational
exercise, according to (2). However, this tensor is not generally available, in the case of
arbitrary fiber spacings and concentration levels. The need thus arises for introducing
estimates for A.
37
38 M. FERRARI
One such estimate, which has received considerable attention in recent literature, is
the method of Mori-Tanaka (Mori and Tanaka, 1973), which is based on the assumption
that
A = T[(l - ar)I + ar(T>]-‘. (5)
While the appropriateness of the Mori-Tanaka method, and all other approxi-
mations on A, must be assessed experimentally, on a case-to-case basis, some physical
considerations may guide the choice of the approximate expressions for the strain
concentrator A. In particular, the following are adopted here as necessary conditions for
the general validity of any homogenization scheme:
Requirement2: The effective stiffness C must be diagonally symmetric. This ensures
reversibility of any composite deformation in the linear elastic range.
Requirement2: In the power-series expansion of C in terms of (Y, the first-order term
must be ((C’ - Cm)T). Requirement 2 ensures the recovery of Eshelby’s exact solution at
dilute concentrations.
Requirement3: At the unitary fiber concentration limit, the effective stiffness
prediction must be independent of the matrix properties.
Requirement4: The effective moduli associated with C must comply with the
variational bounds. Examples of such bounds are those obtained by Hashin and
Shtrikman (1963) for macroscopically isotropic composites with isotropic phases.
The viewpoint adopted here is that, in order to be qualified as admissible, an effective
medium theory must satisfy Requirements l-4 for any biphusecomposite,i.e. for any
material symmetry, morphology, volume fraction, and orientation distribution of the
phases. As will be apparent in the following, the most demanding of these restrictions is
associated with the orientation distribution of the embedded phase-a requirement that
has not only academic, but also a considerable technological significance in the context
of Functionally Graded Materials (FGMs), especially for their application in thermo-
chemically aggressive environments (Nadeau and Ferrari, 1993).
Elementary as they may be, the above-stated requirements suffice to prove the
general inadmissibility of the Mori-Tanaka method. In particular, this complies with
Requirements 1 and 3 if and only if the inhomogeneities are perfectly aligned, or consist
of isotropic material (Ferrari, 1991). For completeness, it is recalled that Voigt’s (1928)
assumption A = I obviously violates Requirements 2 and 4, while Eshelby’s (1957)
assumption A = T violates Requirements 3 and 4.
In an effort to simplify the search for a fully admissible effective medium theory,
in the present paper the effective medium theory is reformulated in the context of an
equivalent eigenstraining poly-inclusion problem. In this formulation, the poly-inclusion
strain concentrator & defined in (9), replaces A as the independent variable of the
homogenizing scheme. This affords a considerably simplified analytical statement of the
physical admissibility requirements.
On this basis, a family of fully admissible homogenization approaches is presented
below, that differ by a single scalar-valued function of the second-phase concentration-
the so-called interaction function, Comparison with exact analytical results and
experimental data is employed to conclude the paper with the identification of a specific
interaction function.
Here, sCis the strain perturbation due to 0. This procedurethen reducedthe single
inhomogeneityproblem to the previouslysolvedproblem of a largehomogeneousbody
with uniform non-zeroeigenstrainingin Szonly.
Eshelby (1957)proceededto determine that the strain in the inhomogeneityis
uniform, and givenby
gf = Ef = Tg” (7)
in duality with the fact that the strainin theequivalenteigenstraininginclusionis uniform,
and givenby
E” = E” = Eg* + go. (8)
For the generalpoly-inhomogeneitycase,relation (7) must be approximatedby (3).
Similarly, the averageinclusion strains for the caseof a nondilute concentrationof
inclusions,undergoingthe sameuniform eigenstrainE* and in the absenceof mechanical
loads, is not given by (8). In this context, the approximateexpression
g” = fig* (9)
is now introduced,in analogywith (3). By parallelingEshelby’soriginal developments,
the poly-inclusion and the poly-inhomogeneityproblemsare put in correspondence by
imposingthat the averageinhomogeneitystressand the averageinclusionstresscoincide.
Consequently,the inhomogeneityaveragestrain in the non-dilutecaseis
8’ = *go (10)
with correspondingapproximateeffectivestiffness
c = cm + a((Cf - Crn)f> (11)
wheref is obtainedfrom T by replacingE with a in (4).
Since no quantity in (16) depends on Cf, the holding of (16) for all composites is
equivalent to Requirement 3. Further specialization of (16) is possible, considering the
special case of perfect second-phase alignment. In this context, the orientational
averagings may be deleted, and (16) is equivalent to
fil = 0. (17)
For texture-independent a., Requirement 3 is thus equivalent to imposing (17) for all
composites.
The family of approximate poly-inclusion strain concentrators
ti = f(a)E + g(a)1 (18)
is now introduced. Here f( *) and g( *) are arbitrary functions of the fiber concentration,
and are referred to as “interaction functions”. The effective stiffness tensor, associated
with the general form (18) via eqn (1 l), may be expressed as
C = C” + cx<[(Cf - Cm)-’ + f(a)E(Cm)-’ + g(a)(C”)-‘I-‘). (19)
All d of the form (18) are in compliance with Requirement 1, since the contraction
E(C”)-‘, also known as the “polarization tensor”, is always symmetric (Walpole, 1981).
The limit conditions, Requirements 2 and 3-eqn (17)-are satisfied if and only if
lim f(o) = 1, lim f(o) = 0, lim g(o) = 0, lim g(o) = 0. (20)
IX-0 a--r1 Or-0 a-1
Fig. 1. Domain of admissibility of the functionf for g = 0. f(e) and g(e) ar defined in eqn (18).
f= 5(3x + 4)/6(x + 2) for x s 4/3, f= (3x + 4)/3x for x 2 4/3.
From (32) it follows that Voigt’s assumption f(s) = 0 is not admissible for any
concentration levels, while the dilute approximationf(.) E 1 does not violate the bounds
if and only if
9x + 8
x I 4/3
cYs 5(3x + 4)
(34)
4
a=3x+4 x 2 4/3.
l.OF
\* Experimenlaipointe l
\
\
\ Lowerli-Sbound _---
0.6 - ‘\ f(a) - 1 + a
\
0 ‘. UpperH-sbound
\ f(a)=cfmetx(l-a)““”
\
0.6 - \
l ‘, \
f(a)
.A,
0.4 -
‘. \
. .. \
*... \
\
-*.. \
. ..* \
0.2 - \
--.*
.... ’ \
-**. \
‘-..\
I I I , . .y
0.0 0.2 0.4 0.6 0.6 1.0
Fiber volume fraction (a)
Fig. 3. Interaction function versus volume fraction (from Hasselman and Fulrath, 196513).
,a-
\ LowerH-Sbound ____
0.6 - ’ f(a) - 1 + a
A\
U~~erH-Sbovt-mi ,.....
0’ \ f(a)=canstx(l -a)
\
0.6 - \
f(a) 0 ‘\
\
*a ‘\
0.4 - -\\
“Y- \
-... \
\
. . .. \
-... \
0.2 - \
. .. . \
. . .. \
.... \
. . .. \
I I I
0.0 0.2 0.4 0.6 0.6 1.0
Fiber volume fracfion (a)
Fig. 4. Interaction function versus volume fraction (from Hasselman and Fulrath, 1965a).
M. FERFARI
1.2 .
r
1.0X-
\
\ mrl-pdntl l
\
\
\ Lcwwf+-sbcund ----
0.6 - \‘\ f(a) - 1 + a
\
\
\
\ UwH-S- .....
\\ f(a)=amIx(l -aI
fta) 0.6 L \
\
\
*. \
**. *.. .
-. a. .‘\
0.4 -
OA- *-....**.
0.
*. L
*.
..
*. ‘,*
*. \
0.2- *. \
-- -. \
**._ ‘.
..
*.I,
I I I
0.0 02 0.4 0.6 0.6 1.0
Fiber volume fraction (a)
Fig. 5. Interaction function versus volume fraction (from Nishimatsa and Gurland, 1960).
6~.~
Fig. 6. C,, versus volume fraction (from Dean and Turner, 1973).
A similar procedure was employed ibidem to determine the interaction function for
a cobalt/tungsten carbide composite, with experimental data obtained from Nishimatsu
and Gurland (1960). The results of the comparison are given in Fig. 5, and again are in
agreement with (21). Finally, eqn (13) sub assumption (21) was directly employed to
compute the transverse Young’s moduli of two different unidirectional carbon-epoxy
composites. The computed values are shown in Fig. 6 to coincide within experimental
error with the measurements of Dean and Turner (1973).
Acknowledgements-The author thankfully acknowledges support from the National Science Foundation,
Award # ECS 9023714.
REFERENCES
Dean, G. D. and Turner, P. (1973). The elastic properties of carbon fibers and their composites. Composites,
July 1973, 174-180.
Eshelby, J. D. (1957). The determination of the elastic field of an ellipsoidal inclusion, and related problems.
Proc. R. Sot. A 241,276-303.
Ferrari, M. (1991). Asymmetry and the high concentration limit of the Mori-Tanaka effective medium theory.
Mech. Muter. 11, 251-256.
Ferrari, M., Imam. A. and Kuong, L. (1993). Recent developments in homogenization theory: on the interaction
function of the equivalent poly-inclusion formalism. In Composite Materials Technology 1993 (Edited by
D. Hui, 0. Ochoa and T. J. Kozik), Vol. PD-53, pp. 237-241. ASME, New York.
Ferrari, M. and Johnson, G. C. (1989). The effective elasticities of short-fiber composites with arbitrary
orientation distribution. Mech. Muter. 8, 67-73.
Ferrari, M. and Marzari, N. (1992). A Mori-Tanaka theory for short-fiber composites: application. J. Energy
Res. Tech. 144, 101-104.
Hashin, Z. and Shtrikman, S. (1963). A variational approach to the elastic behavior of multiphase materials. J.
Mech. Phys. Solids 11, 127-141.
Hasselman, D. P. H. and Fulrath, R. M. (1%5a). Effect of alumina dispersions on Young’s modulus of a glass.
J. Am. Ceram. Sot. 48(4), 218-219.
Hasselman, D. P. H. and Fulrath, R. M. (1965b). Effect of spherical tungsten dispersion on Young’s modulus
of a glass. J. Am. Ceram. Sot. 18(10), 548-549.
Mori, T. and Tanaka, K. (1973). Average stress in matrix and average elastic energy of materials with misfitting
inclusions. Actu Metall. 21, 571-574.
Nadeau, J. and Ferrari, M. (1993). (In preparation.)
Nishimatsu, C. and Gurland, J. (1960). Experimental survey of the deformation of the hard-ductile two-phase
alloy system WC-Co. Trans. ASM 52, 469-484.
Voigt, W. (1928). Lehrbuch der Krystaliphysik. Teubner, Leipzig.
Walpole, L. J. (1981). Elastic properties of composite materials: theoretical foundations. Adv. Appl. Mech. 21,
169-186.