Author's personal copy

Mechanics of Materials 69 (2014) 251–261

Contents lists available at ScienceDirect

Mechanics of Materials
journal homepage: www.elsevier.com/locate/mechmat

Composites with fractal microstructure: The effect of long range

correlations on elastic–plastic and damping behavior
R.C. Picu a,⇑, Z. Li a, M.A. Soare b, S. Sorohan c, D.M. Constantinescu c, E. Nutu c
a
Department of Mechanical, Aerospace and Nuclear Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180, United States
b
General Electric Global Research, Niskayuna, NY 12309, United States
c
Department of Strength of Materials, University POLITEHNICA of Bucharest, Bucharest, Romania

a r t i c l e i n f o a b s t r a c t

Article history: The effect of correlations of the spatial distribution of inclusions in a two-phase composite
Received 14 June 2013 is studied numerically in this work. Microstructures with fractal distribution of inclusions,
Received in revised form 22 October 2013 characterized by long-range power law correlations, are compared with random inclusion
Available online 22 November 2013
distributions of same volume fraction. The elastic–plastic response of composites with stiff
elastic inclusions and elastic–plastic matrix is studied, and it is concluded that fractal
Keywords: microstructures always lead to stiffer composites, with higher strain hardening rates, com-
Fractals
pared with the equivalent composites with randomly distributed inclusions. Composites
Damping
Plastic deformation
with filler distributions characterized by shorter range, exponential correlations exhibit
Toughening behavior intermediate between that of random and power law-correlated microstructures.
Larger variability from replica to replica is observed in the fractal case. The pressure in
inclusions is larger in the case of fractal microstructures, indicating that these are expected
to be advantageous in applications such as toughening of thermoset polymers which takes
place via the cavitation mechanism. The effect of the spatial distribution of inclusions on
the effective damping of the composite is also investigated. The matrix is considered elastic
and non-dissipative, while inclusions dissipate energy. The composite with fractal micro-
structure provides more damping than the random microstructure of same filler volume
fraction, and the effect increases with increasing fractal dimension. When damping is
introduced only in the interfaces between matrix and inclusions, the spatial distribution
of fillers becomes inconsequential for the overall composite behavior. These results are rel-
evant for the design of composites with hierarchical multiscale structure.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction in woven fiber composites. Therefore, the distribution of

inclusions in the matrix is dictated primarily by technolog-
Composite materials are broadly used in engineering for ical reasons and not by considerations related to the
their properties emerging from the interaction of the optimization of system level properties of the material.
constituent phases. In most man-made composites the Exceptions to this rule are structures designed by an
structure is either random or periodic. Particulate compos- optimization scheme aimed at achieving an optimum of
ites, made by mixing inclusions of nominally monodis- an objective function representing one or multiple macro-
perse dimensions in a matrix, have a random scopic properties. Such structures can be produced only by
microstructure. In many other situations, manufacturing specialized techniques lacking high throughput, such as
processes lead to periodic microstructures, as for example additive manufacturing (e.g. Vaezi et al., 2012).
Biological materials, on the other hand, have complex
microstructures which are optimized to perform a certain
⇑ Corresponding author. Tel.: +1 518 276 2195. function while using the minimum volume of material.
E-mail address: [email protected] (R.C. Picu).

0167-6636/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.mechmat.2013.11.002
 Author's personal copy

252 R.C. Picu et al. / Mechanics of Materials 69 (2014) 251–261

Examples are the trabecular bone (Parkinson and Fazzalari, separation between the upper and lower bounds (e.g.
2000) and various types of marine shells (Meyers et al., Beran and Molyeux, 1966; Silnutzer, 1972; Milton, 1981;
2006). Mass is distributed in the trabecular bone only in re- Milton, 1982; Phan-Tien and Milton, 1982; Quinatanilla
gions carrying large stresses, while in shells structural ele- and Torquato, 1995). The n-point bounds are written in
ments are distributed in layers such to maximize strength terms of n-point microstructural correlation functions
and toughness. Most of these materials are hierarchical which define the probability that n points with specified
and some are self-similar across a range of scales. These relative positions are all located in a certain phase of the
structures have some degree of stochasticity and are either composite. Any statistical correlation of the microstructure
only approximately periodic (e.g. the structure of abalone can be accounted for by using these methods. A review of
shells) or lack translational symmetry all together (e.g. the higher order bounds and the geometric parameters re-
the trabecular bone). quired for their evaluation is provided in Torquato (2002).
It is useful to ask why nature designs structures with Fractal structures have been studied in connection with
self-similar multiscale structure. In the context of engi- various physical processes such as transport (Dzhaparidze
neered materials, one may alternatively ask how inclusions and van Zanten, 2003), diffusion limited aggregation
should be distributed in a composite to maximize macro- (Witten and Sander, 1983), and dislocation patterning dur-
scopic properties, while preserving some level of stochas- ing deformation of metals (Zaiser and Hahner, 1999; Bako
ticity, which is mandated by the technological need to and Hoffelner, 2007), microscale plasticity (Chen et al.,
produce such materials at reasonable cost and in large vol- 2010; Ostoja-Starzewski, 2012). Fractal concepts were also
umes. This is the objective of the current work. used in percolation theory (Bergman and Kantor, 1984),
Specifically, while acknowledging that the volume quantum mechanics (Argyris et al., 2000), fracture
fraction of inclusions is the major factor controlling the mechanics (Bazant, 1997), etc. However, despite its practi-
properties of the composite, we inquire what benefits may cal importance, very few attempts have been made to
be obtained if the reinforcing phase is distributed in a spa- study the deformation of such structures or that of com-
tially correlated way. To address this question, we compare posites containing fractal inclusions.
random and spatially correlated distributions of inclusions. The elastic moduli of deterministic fractal structures
The limit case of a spatially correlated microstructure is a have been predicted using standard finite element models
fractal, in which correlations are described by a power law and renormalization group concepts to extrapolate to the
and the exponent of the correlation function depends on range of scales not accessible by direct simulation (Oshm-
the fractal dimension (Falconer, 2003). Fractal microstruc- yan et al., 2001). Dyskin applied the differential self-con-
tures lack translational symmetry, but have scaling symme- sistent method (initiated in Salganik (1973)) for media
try, i.e. remain self-similar upon a scaling operation containing self-similar distributions of spherical/ellipsoi-
(Mandelbrot, 1983; Falconer, 2003). The properties of inter- dal pores or cracks (Dyskin, 2005). The author proposes
est are the elastic–plastic response and the damping behav- to model such materials by a sequence of continua with
ior of the composite. Damping is of interest in applications effective elastic properties. This does not take into account
in which the material is subjected to intense vibrations the interaction of inclusions. Other approaches consider
(e.g. composites used for helicopter blades) and inclusions the reformulation of the governing equations to account
are added to enhance energy dissipation, while providing for the fractal nature of the inclusion domains. Tarasov
stiffness and strength. Damping may take place in the vol- studied porous materials having pores with a broad range
ume of inclusions or/and at the interface with the matrix. of sizes and in which the mass of the material within a vol-
Another problem of interest is toughening of brittle poly- ume of dimension R scales as mðRÞ  RQ , with Q non-inte-
mers (thermosets) by the addition of rubbery inclusions. ger (Tarasov, 2005a,b). The author replaces the fractal body
Toughening is triggered in such situations by cavitation with an equivalent continuum governed by a ‘‘fractal met-
within inclusions. This process is driven by the hydrostatic ric’’. The balance equations for mass, linear and angular
stress component. Inclusions are usually distributed ran- momentum for the equivalent continuum are reformulated
domly in the polymer volume and it is interesting to inquire in terms of this metric. This method was further developed
if a different stochastic distribution could lead to larger recently in Ostoja-Starzewski (2007, 2009) to represent the
toughening at prescribed filler volume fraction. mechanics of heterogeneous bodies with fractal
The effect of the distribution of inclusions on the elastic microstructure.
moduli of composites has been studied for a long time. Re- Carpinteri et al. studied the deformation of a bar in
views on the homogenization of random composites are which the strain is localized in a subset of cross-sections
presented in Nemat-Nasser and Hori (1999), Torquato forming a Cantor set (Carpinteri et al., 2004). These authors
(2002) and Dvorak (2013). Remarkable results have been use fractional operators to rewrite the balance equations,
obtained regarding the bounds on the elastic moduli of although in one dimension this is not immediately neces-
such composites. These expressions are generally given in sary. The deformation of a two-dimensional composite
terms of the volume fraction of the constituents. The clos- with Cantor-like inclusion distribution was studied in
est bounds for the bulk modulus which take into account Soare and Picu (2007). A numerical method using enriched
only the volume fraction have been derived by Hashin shape functions that account for the finer scale geometry
and Shtrikman (HS) (Hashin and Shtrikman, 1962). A fam- was developed in this work and was applied to structures
ily of higher order bounds, which take into account statis- with an arbitrary number of scales.
tical measures of the microstructure geometry, have been In Picu and Soare (2009) fractional calculus based on
proposed more recently with the purpose of reducing the local fractional operators introduced by Kolwankar and
 Author's personal copy

R.C. Picu et al. / Mechanics of Materials 69 (2014) 251–261 253

Gangal (1996) and Kolwankar (1998) were used to formu- of the domain A at a certain scale n, An, is seen as a reunion
late the balance equations on the fractal support. The for- of Mn cells of characteristic dimension en , of which Pn are
mulation was applied to modeling the deformation of occupied by the inclusion (fractal) material. The remaining
two-dimensional composites containing a fractal distribu- Mn  Pn cells are occupied by the matrix. The number of
tion of inclusions in a matrix. The method is readily possible configurations at iteration, n, is
n1
extendable to three-dimensional composites. ½M!=ðP!ðcM  PÞ!Þp þþpþ1 . The volume fraction of inclu-
In the stochastic case, in addition to Monte Carlo (MC) sions is given by f ¼ ðP=MÞn , while the fractal dimension
methods (Papadrakakis and Papadopoulos, 1996), various of the set is D ¼ 2logðPÞ=logðMÞ. The natural numbers M
systematic ways of approaching numerically partial differ- (M P 2) and P (1 6 P < M) are kept as parameters in this
ential equations defined on single-scale stochastic do- analysis. Fig. 1(a) shows a realization of a composite with
mains were proposed in the literature. Methods based on M = 9, P = 6 and n = 5, which has volume fraction f = 13.1%
probabilistic finite elements (second order perturbation and fractal dimension D = 1.63. Note that the fractal
PFEM) (Liu et al., 1986, 1987), or the spectral approach dimension is smaller than 2, the dimension of the embed-
for stochastic finite elements (SSFEM) (Ghanem and Spa- ding space.
nos, 1991) are relevant examples. These methods were ap- It is important to observe that the set of inclusions rep-
plied to various problems in solid and fluid mechanics such resents a fractal object only in the range of scales defined
as for example to study transport through porous media by the smallest representative length, en , and the largest
(Ghanem and Dham, 1998) and elastic deformation length scale of the problem. If the upper bound (the dimen-
(Matthies et al., 1997). The elastic deformation of compos- sion of the structure in Fig. 1(a)) is taken to be L, one com-
ites with fractal microstructure was represented using the putes en ¼ L=M n=2 , or e5 ¼ L=243 for Fig. 1(a). For any set
stochastic finite element method in Soare and Picu (2008a) with given en , the object is non-fractal and with dimension
based on an approximation of the spectral decomposition equal to that of the embedding space at all length scales
of the representation of the fractal microstructure pre- smaller than en . The composite is not defined on scales lar-
sented in Soare and Picu (2008b). ger than L in the present case, since the boundary condi-
In most problems involving fractal microstructures tions are defined at this scale. The fractal object does not
studied to date, the mechanical behavior of interest was have translation symmetry, rather it has scaling symmetry.
either the dynamic or the quasistatic elastic response. In For example, one may fill the embedding space with repli-
the present work we use ensemble averaging of realiza- cas of the fractal set bounded by L and en , but the resulting
tions modeled using finite elements and investigate a structure will have translation symmetry on scales larger
broader range of mechanical properties, as mentioned than L, and scaling symmetry between L and en .
above. This allows investigating not only the global com- An important property related to the present discussion
posite behavior, but also the local stress distribution, is that the fractal structure has long range, power law cor-
which is relevant in damage nucleation and evolution. relations. Consider the characteristic function defined on
The article begins with an overview of the models and the fractal support, taking values of 1 in inclusions and 0
methods used (Section 2), and continues with results per- elsewhere: hðx1 ; x2 Þ ¼ 1 if ðx1 ; x2 Þ 2 An , and hðx1 ; x2 Þ ¼ 0
taining to the elastic–plastic deformation (Section 3.1), if ðx1 ; x2 Þ 2 A  An . This function is probed with resolution
internal stresses in structures subjected to quasistatic en. On scales larger than en one has ACFðrÞ ¼ hðx1 þ r; x2 Þ
deformation (Section 3.2), and the investigation of the ef- hðx1 ; x2 Þðx1 ;x2 Þ  r4þ2D , as r ! 1 (Falconer, 2003; Gneiting
fect of the distribution of inclusions on the damping and Schlather, 2004), where hxiðx1 ;x2 Þ indicates ensemble
behavior of the composite (Section 3.3). Conclusions are averaging over all points defined by the coordinate pair
presented in closure. (x1, x2). Due to its stochastic isotropy, the fractal object
has ACFðrÞ ¼ hðx1 þ a1 ; x2 þ a2 Þhðx1 ; x2 Þðx1 ;x2 Þ  r4þ2D , with
r2 ¼ a21 þ a22 .
2. Models and methods The behavior of these structures is compared with that
of composites with random distribution of inclusions of
Two phase two-dimensional composites are considered. same volume fraction and having the same characteristic
The matrix fills the Euclidean space of the problem do- length en , which, in this case, represents the size of the ran-
main. In each realization, inclusions form a fractal which domly distributed inclusions. Fig. 1(b) shows a realization
is a generalization of the classical Cantor set to probabilis- of a random structure having the same f and en as the
tic structures embedded in 2D (Soare and Picu, 2008a). The structure in Fig. 1(a). The correlation of the characteristic
microstructure is constructed hierarchically by iteratively function h of the random microstructure is a Delta function
applying a set of transformation rules. The first generation of variable r when probed with resolution en . Comparing
(approximation of the fractal set with an infinite number of the mechanical behavior of composites similar to those
scales) is obtained by starting with an Euclidean domain, in Fig. 1(a) and (b) provides insight into the role of spatial
dividing it in M equal cells and selecting randomly P of correlations of the position of inclusions in defining the
them which are to be filled by inclusions. The characteris- composite mechanics.
tic length of these inclusions is e1 . The number of possible In order to put this discussion in perspective, we have
configurations at the first generation is M!=ðP!ðM  PÞ!Þ. also considered composites in which the position of inclu-
The next generations are obtained by dividing again each sions is exponentially correlated. An example is shown in
of the fractal cells in M equal parts from which M  P are Fig. 1(c). This structure has the same parameters f and en
transformed into ‘‘matrix’’ cells. Thus, the approximation with those in Fig. 1(a) and (b), but has
 Author's personal copy

254 R.C. Picu et al. / Mechanics of Materials 69 (2014) 251–261

Fig. 1. The three types of composite microstructures studied in this work. The matrix is shown in white and inclusions in black. (a) Fractal distribution of
inclusions with M = 9, P = 6 and n = 5. The smallest length scale, or the dimension of an isolated inclusion is en ¼ e5 ¼ L=243. The volume fraction of
inclusions is f = 0.131 and the fractal dimension is D = 1.63. (b) Random distribution of inclusions of same f and en . (c) Microstructure characterized by an
exponential correlation function of inclusion positions and having the same f and en as the fractal structure in (a).

ACFðrÞ ¼ hðx1 þ a1 ; x2 þ a2 Þhðx1 ; x2 Þðx1 ;x2 Þ  expðr=r0 Þ for some extent, by controlling the constant J in the energy
all pairs a1 and a2 having the property r 2 ¼ a21 þ a22 , and function and its dependence on the distance between the
with r 0 a constant. The specific situations in which such interacting spins, si and sj. In this work, interactions are
structures are used and the parameters involved (e.g. r0), considered up to the second nearest neighbors of the
are discussed in Section 3.1. square lattice. The interaction strength is J = 2.5 for both
It is of interest to outline the method used to generate first and second nearest neighbors. The temperature
the exponentially correlated samples. To this end, domain parameter b was kept in the close vicinity of the critical
A is partitioned on scale en . On the resulting square lattice, point provided by the mean field solution, bc ¼ 2=ðzJÞ,
the characteristic function takes random binary values, 0 where z is the number of interacting neighbors of each site.
and 1. An Ising model is then used to evolve the system. Fig. 2 shows the correlation function obtained using this
The Ising model has been extensively studied in statistical procedure and corresponding to Fig. 1(c), along with the
physics and was used to model phase transitions in many power correlation function corresponding to the fractal
systems, e.g. see Baxter (1982). In the present context, each microstructure of Fig. 1(a). The random distribution of
cell is assigned a spin (either s = +1 or s = 1) depending on Fig. 1(b) leads to a Delta function centered at zero and is
the local value of the characteristic function. The spins not shown in Fig. 2. The corresponding best fits to the
interact such that the total energy of the system in absence two curves in Fig. 2, exponential and power law, are also
P
of an external field is E ¼  fi;jg Jsi sj , where the summation shown. Note that the characteristic length of the exponen-
is performed over all interacting cells/spins {i, j}. Parameter tial function, r0, is selected such that it provides a
J can be selected to depend on the relative position of the reasonable approximation for the power law correlation
two cells i and j. The system is evolved using a Monte Carlo
procedure which is controlled by a temperature-like
parameter, b ¼ 1=kB T. When the dimensionality is larger
than one, the mean field solution of this model predicts a
phase transition once the temperature decreases below a
critical value, or b > bc . For b < bc , spins are random and
the total magnetization, hsi, vanishes. Below the critical va-
lue, the system acquires a net magnetization. In the lan-
guage of the present application, when b < bc , function h
is either 0 or 1 with equal probability, and f = 0.5. For
b > bc , one of the two phases dominates and h can be ad-
justed such to obtain the volume fraction, f, in the desired
range of values. To this end, b is selected in the close vicin-
ity of the critical point. Numerical Ising models have pro-
vided a richer physics which could not be captured by
the mean field approach. The system exhibits residual
magnetization even above the critical temperature due to Fig. 2. Normalized ACF(r) functions for a fractal microstructure with
spin clustering and the phase transition takes place M = 9, P = 5 and n = 4 (blue circles) having D = 1.46 and f = 0.131, and for a
gradually. clustered microstructure similar to that in Fig. 1(c) and with f = 0.131
The interesting property of the Ising model is that it (filled red squares). The functions are normalized by the variance ACF(0).
The dotted red line represents an exponential fit to the ACF of the
provides exponential spatial correlations, or clustering, of clustered microstructure. The slope of the blue dashed line is 0.95. (For
the spins and the correlation of the characteristic function interpretation of the references to color in this figure legend, the reader is
h is exponentially decaying. The range can be adjusted, to referred to the web version of this article.)
 Author's personal copy

R.C. Picu et al. / Mechanics of Materials 69 (2014) 251–261 255

in the vicinity of the origin. The exponential correlation has 3.1. Elastic–plastic behavior
an early cut-off at r > r 0  4en , while the decay of the cor-
relation function of the fractal is much slower and the cor- Structures similar to those shown in Fig. 1 are consid-
relation function is much longer ranged. ered for this study. The matrix fills the square problem do-
These microstructures are discretized and their main and embeds inclusions of dimension en. Stress–strain
mechanical response is determined with a finite element curves are computed for 100 realizations, each being
model. We have used the commercial finite element pack- loaded uniaxially in displacement control up to a global
age Ansys for all simulations. The two-dimensional simula- strain of 2%. The boundaries in the direction perpendicular
tion domain for each representation is a square of length L, to the loading direction are traction free. The stress–strain
which is meshed uniformly with 4-node plane stress ele- curves and all subsequent plots represent the ensemble
ments. At a given M value, the same level of refinement average stress. The standard deviation is below 1% for the
is maintained for all scales, n. For example, all structures random case and approximately 5% for the fractal case
with M = 9 and n = 2–5 are meshed with 9722 elements, (which is about the size of the symbols used in Fig. 3).
i.e. with elements of size en =4, where en corresponds to Fig. 3 shows results for two fractal microstructures with
n = 5. We have tested in representative cases that further M = 9, P = 5 and n = 2 and 3, respectively (filled symbols),
mesh refinement does not lead to significantly different and random microstructures of the same volume fraction
system-scale results. Furthermore, we checked that at this (open symbols). The volume fractions are f = 0.308 and
level of mesh refinement the use of 8-node elements does 0.171 for n = 2 and 3, respectively. It is seen that all curves
not lead to different results. are bilinear, and that at the same volume fraction the curve
Model size effects are usually a concern when analyzing corresponding to the fractal case is above that for the random
structures with spatial correlations. In the case of ran- microstructures. However, the probability that two specific
domly distributed inclusions, the model size should be at realizations with fractal and random microstructures lead
least an order of magnitude larger than the characteristic to stress–strain curves which are in the reverse order is not
size of inclusions in order to insure that the results are zero. As n increases, f decreases and hence the curves asymp-
model size-effect free (Dvorak, 2013). This insures scale tote to the stress–strain curve of the matrix (shown by the
decoupling between the characteristic length scale of the continuous line in Fig. 3). Furthermore, as n increases, the
microstructure and the scale of observation/homogeniza- distinction between the curves for fractal and random cases
tion. A similar rule applies in presence of correlations. decreases since the volume fraction f decreases.
For example, in the case of the exponentially correlated The effective elastic modulus of the composite, Ee, and
microstructures of Fig. 1(c), L should be an order of magni- the strain hardening rate defined by the slope Ep, can be
tude larger than r0 (here r0  4en and L ¼ 243en ). The case evaluated from the stress-strain curves as suggested in
of the fractal microstructures is different as the geometry Fig. 3. These two parameters fully define the uniaxial re-
is scale-free and hence correlation decay is power law. sponse of the composite, therefore we focus attention on
The behavior of these structures is intrinsically dependent their dependence on the fractal dimension, D, and the vol-
on the two scale (upper, L, and lower, en) at which the hier- ume fraction, f. Specifically, the fractal dimension controls
archy is truncated. This ‘‘size effect’’ is an intrinsic property the exponent of the autocorrelation power function and
of the fractal microstructures which, we suggest, can be hence comparing fractal structures with same f and various
used to advantage in material design. D, one may infer the effect of the spatial correlation of the
In models used to determine the elastic–plastic re- distribution of heterogeneity on the overall composite
sponse, inclusions are linear elastic with Young’s modulus response.
and Poisson ratio E2 and m2, while the matrix is represented
with a bi-linear model with slopes E1 and 0.1E1 in the elas-
tic and plastic regimes, respectively. In all simulations
E2 = 6E1, m1 = m2 = 0.3 and kinematic hardening is used for
the plastic range. The yield strain of the matrix material
is 0.01. In models used to study the damping behavior of
the composite, the two materials are linear elastic and iso-
tropic with E2 = 6E1 and m1 = m2 = 0.3. Parameters character-
izing damping are discussed in Section 3.4. In all cases
discussed below, stress is normalized by E1 and displace-
ments by L.

3. Results and discussion

Structures with fractal and random microstructures are

studied with respect to their elastic–plastic and damping Fig. 3. Stress–strain curves for composites with fractal microstructure
(filled symbols) and random microstructures of same volume fraction
behavior. The central question posed refers to the role of (open symbols), with M = 9, P = 5, n = 2 (circles) and M = 9, P = 5, n = 3
the distribution of heterogeneity in defining the overall re- (squares). The continuous line represents the mechanical behavior of the
sponse of the composite. matrix material. The dashed lines define slopes Ee and Ep.
 Author's personal copy

256 R.C. Picu et al. / Mechanics of Materials 69 (2014) 251–261

Fig. 4(a) shows the variation of the elastic modulus, Ee/ f ð1  f ÞðK 2  K 1 Þ2
E1, with the volume fraction, f, for random structures (open K u ¼ hKi  ;
~ þ hGi
hKi n
symbols), and various fractal structures. Data are shown
for M = 9, P = 5, and n = 2–5 (blue circles), which all have f ð1  f ÞðK 2  K 1 Þ2
K l ¼ hKi  ð3aÞ
D = 1.46, for M = 4, P = 3, n = 7 and 8 (green squares), which ~ þ ðhG1 i Þ1
hKi n
have D = 1.58, and for M = 81, P = 42, n = 2 and 3 (red trian-
gles), which have D = 1.7. These structures have different
f ð1  f ÞðG2  G1 Þ2
filler volume fractions. Obtaining fractal geometries with Gu ¼ hGi  ;
~ þH
hGi
the same f and various D values is not possible. The thick
continuous orange lines represent the Hashin–Shtrikman f ð1  f ÞðG2  G1 Þ2
Gl ¼ hGi  ð3bÞ
bounds for the two-dimensional case (Hashin and Shtrik- ~ þW
hGi
man, 1962; Hashin, 1965).
The data points for the random microstructures align on where hzi ¼ ð1  f Þz1 þ fz2 , h~zi ¼ ð1  f Þz2 þ fz1 , and
a curve described by Eq. (1) and shown by the continuous hzin ¼ ð1  n2 Þz1 þ n2 z2 for any quantity, z, and defining
thin line in Fig. 4(a). phase 2 as the stiffer inclusions and phase 1 being the ma-
trix. Parameters H and W in Eq. (3b) are given by:
Ee =E1 ¼ 1:73f 2 þ 1:27f þ 1 ð1Þ
2hKin hGi2 þ hKi2 hGig
The four data points corresponding to fractal structures H¼ ð4aÞ
hK þ 2Gi2
with M = 9, P = 5, D = 1.46, are well represented by a similar
curve, which is shown by the dashed line in Fig. 4(a). It is
possible to approximate all data with the expression: W1 ¼ 2hK 1 in þ hG1 ig ð4bÞ

2
Ee =E1 ¼ 1:73gf þ 1:27gf þ 1 ð2Þ The bounds for Young’s modulus, E, are computed using Eq.
(3) and the 2D relationship between the three types of
with g ¼ 0:2ð3D þ 1Þ, which holds for D > 1.3 This indicates moduli, 4=E ¼ 1=G þ 1=K.
that the stiffness increases with D, which is a consequence n and g, are 3-point parameters characterizing the dis-
of the stronger interaction of inclusions in the fractal tribution of inclusions (phase 2) in the matrix. These are gi-
structures. ven for the two-dimensional case by Milton (1982) and
The influence of filler packing on the elastic moduli has Torquato (2002):
been discussed before. The properties of fiber composites Z 1 Z 1 Z p
4 dr ds
loaded perpendicular to the preferential direction of fibers n2 ¼ dh0 cosð2h0 Þ
have been determined for a variety of periodic arrange-
pf ð1  f Þ 0 r 0 s 0

ments (e.g. Brockenbrough et al., 1991; Nakamura and Sur-  ½S3 ðr; s; tÞ  S2 ðrÞS2 ðsÞ=f  ð5aÞ
esh, 1993). For random microstructures one can make use
Z 1 Z 1 Z p
of the n-point bounds derived in the homogenization liter- 16 dr ds
g2 ¼ dh0 cosð4h0 Þ
ature (for a review see Torquato (2002)) to investigate the pf ð1  f Þ 0 r 0 s 0
expected variation of the elastic constants in presence of  ½S3 ðr; s; tÞ  S2 ðrÞS2 ðsÞ=f  ð5bÞ
long range correlations. The three point bounds for the
bulk (K l ; K u ) and shear (Gl ; Gu moduli in two dimensions where S2(z) is the 2-point correlation function representing
are given by Milton (1982) and Torquato (2002): the probability that the ends of a segment of length z are

Fig. 4. Variation of (a) the elastic modulus (Ee/E1), and (b) strain hardening rate (Ep/E1) with the volume fraction, for microstructures with randomly
distributed inclusions (open symbols and continuous thin line), and for various fractal microstructures having M = 9, P = 5, and n = 2–5 (blue circles), M = 4,
P = 3, n = 7 and 8, M = 9, P = 6, n = 4 and 5 (green squares), M = 81, P = 42, n = 2 and 3 (red triangles). Data for microstructures with exponential correlation
function (Fig. 1(c)) are shown in (b) with crosses and dashed-dot line. The thick orange continuous lines in (a) represent the 2D Hashin–Shtrikman bounds.
The thin continuous line and the dashed line in both (a) and (b) represent the best fit to the random structures data and to the fractal structures with M = 9,
P = 5, respectively. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
 Author's personal copy

R.C. Picu et al. / Mechanics of Materials 69 (2014) 251–261 257

both in phase 2, and S3(z1, z2, z3) is the 3-point correlation from Eq. (3) without the need to perform a numerical
function representing the probability that all corners of a study.
triangle of edge lengths z1, z2, z3 are in phase 2. Length t Fig. 4(b) shows the normalized strain hardening rate
is computed in terms of r, s and h0 as versus the filler volume fraction for composites with ran-
t 2 ¼ r 2 þ s2  2rscosðh0 Þ. dom microstructures, composites with microstructures
The 2-point correlation function is evaluated numeri- having exponential correlations, and fractal microstruc-
cally from the microstructure. For all microstructures stud- tures with parameters similar to those discussed in rela-
ied S2 has the limit values S2 ð0Þ ¼ f and S2 ð1Þ ¼ f 2 . For the tion to Fig. 4(a). The strain hardening rate is independent
random microstructure S2 ðrÞ ¼ f 2 for all r > en . The fractal of strain since the composite has a bilinear stress–strain
microstructures lead to S2 functions with power law decay curve. It is seen that all microstructures with spatial corre-
on scales r > en . The 3-point correlation function S3 is esti- lations of the distribution of inclusions have larger strain
mated using the approximation in terms of 2-point corre- hardening rates than the random microstructures. The
lation functions proposed in Banissadi et al. (2012): longer ranged the correlation, the larger is the value of
Ep. The fractal dimensions of the fractal microstructures
t S2 ðrÞS2 ðsÞ s S2 ðrÞS2 ðtÞ considered is D = 1.43, 1.58, 1.63 and 1.7, for the structures
S3 ðr; s; tÞ  þ
rþsþt f rþsþt f with (M = 9, P = 5), (M = 4, P = 3), (M = 9, P = 6) and (M = 81,
r S2 ðtÞS2 ðsÞ P = 42), respectively. It is apparent that as D increases, the
þ ð6Þ data move to higher values of Ep.
rþsþt f
An interesting comparison can be made between the
which is reported to provide estimates with a maximum fractal system with M = 9, P = 5 and the system with expo-
error of 20%. The value of n2 computed using Eq. (5a) for nential correlations. The correlation functions for these
the random microstructure with f = 0.0953 is n2 ¼ 0:33. two types of structures are shown in Fig. 2. The exponen-
This is larger than the value predicted with the formula tial correlation is selected such to approximately match
n2 ¼ 0:08079ð1  f Þ þ 0:91921f (n2 = 0.1607) proposed in the power law correlation of the fractal structures at small
the literature for symmetric-cell materials of checkerboard values of r/en. The data in Fig. 4(b) indicate that the strain
type (Torquato, 2002). For the same random microstruc- hardening rate of the fractal microstructure is always lar-
ture we obtain g2 ¼ 0:126 using Eq. (5b). The correspond- ger than that of the microstructure with exponential corre-
ing parameters for the fractal microstructure with M = 9, lations. This indicates the effect of the range of spatial
P = 5, n = 4 (D = 1.465, f = 0.0953) are n2 ¼ 0:373 and correlations: as this parameter increases, the effective
g2 ¼ 0:212. modulus, Ee/E1, and the strain hardening rate, Ep/E1, in-
The HS bounds are compared with the 3-point bounds crease. This is attributed to the enhanced interaction of
evaluated using Eqs. (3) and (4) in Fig. 5. The range defined inclusions in the microstructures with non-random distri-
by the higher order bounds is much narrower than that de- butions of inclusions.
fined by the HS bounds, as expected. The observation rele- The effect of filler packing on strain hardening rates was
vant for the present discussion is that both 3-point bounds observed before for periodic structures in 2D plane strain
shift up as the range of spatial correlations increases, models (Brockenbrough et al., 1991; Nakamura and Suresh,
which is in agreement with the generic trend reported in 1993; Suresh and Brockenbrough, 1993). For example, it
this article for all effective properties discussed. In fact, was observed that the strain hardening rate was signifi-
considering that as the correlation range increases, param- cantly smaller for square diagonal-packed square fillers
eters n2 and g2 increase, this trend can be inferred directly than for square edge-packed fillers. The higher the con-
straint imposed by the filler on the deformation of the ma-
trix, the larger the strain hardening rate. Another
interesting effect was obtained when comparing circular
and square fillers at same filler volume fraction and same
periodic arrangement. It was concluded (Brockenbrough
et al., 1991) that random distribution of squares leads to
larger strain hardening rates than random distributions
of circular inclusions. This was attributed to the large con-
straining effect of fillers with stress concentrators (sharp
corners). We expect that such stress concentrations exists
in our models too; however, this has no bearing on our
conclusions since all composites compared here have the
same filler geometry.
It may be observed that the effect described here can be
interpreted to some extent as a size effect. As discussed in
Fig. 5. Hashin–Shtirkman bounds (continuous orange lines) and 3-point Section 2, no scale decoupling exists in the fractal case. For
bounds for the random (dashed black curves) and fractal (dash-dot blue given model size, L, the truncation of the hierarchy is more
curves) microstructures. The symbols represent the effective moduli of pronounced as D increases since the decay of the correla-
fractal microstructures with M = 9, P = 5, and n = 2–5 (blue circles), and of
random microstructures of same volume fraction f, from Fig. 4a. (For
tion function is slower. Hence, a stronger ‘‘size effect’’ is ex-
interpretation of the references to color in this figure legend, the reader is pected. For imposed displacement boundary conditions,
referred to the web version of this article.) one obtains an overestimate of the elastic moduli (Huet,
 Author's personal copy

258 R.C. Picu et al. / Mechanics of Materials 69 (2014) 251–261

1990) and, as D increases at given filler volume fraction the release of the triaxial stress state as a consequence of
one expects to obtain systematically larger values of the filler cavitation. Recently it was suggested that the domi-
moduli. However, this is not the only reason for the larger nant effect is the dissipation in the matrix.
Ee/E1 and Ep/E1 obtained at larger D. The stronger interac- Fig. 6(b) shows the PDF of the hydrostatic stress (rh =
tion of fillers in fractals with larger fractal dimension is rii/3) in inclusions for the fractal case with M = 9, P = 6
responsible to a larger extent for the effect discussed in and n = 5, and for the random microstructure of same vol-
this section. ume fraction. The horizontal axis is normalized such to
bring the two PDFs to mean 1. A more significant difference
is observed in this case, with the fractal microstructure
3.2. Internal stresses having a broader PDF, and exhibiting a slowly decaying tail
on the tensile side of the distribution. This indicates that in
The stress distribution in the microstructure is also of fractal microstructures cavitation occurs at a lower far field
interest. The maximum principal stress controls fracture stress. Considering that fractal microstructures are in aver-
under monotonic and fatigue loading conditions. Probabil- age stiffer than their random counterparts, the observation
ity distribution functions (PDF) of the stress in the matrix suggests that cavitation occurs at smaller strains, which is
elements have been computed for all microstructures. beneficial in most applications. The origin of the different
Fig. 6(a) shows the distribution function of the maximum PDF of the fractal microstructure can be associated with
(tensile) principal stress in the fractal microstructure with the stronger interaction of inclusions in this type of geom-
M = 9, P = 6 and n = 5, along with the PDF of the same quan- etry. The inter-particle distance has a broader distribution
tity in the matrix of the material with random microstruc- function than in the random microstructure case and
ture of same volume fraction (f = 0.131). The far field hence inclusion interactions and the plastic deformation
loading is uniaxial tension and the stress is evaluated at of matrix ligaments between inclusions are more
a total strain of 2%. The stress was normalized in both cases pronounced.
by the far field mean stress: 0.012E1 for the random micro-
structure, and 0.0122E1 for the fractal microstructure.
Therefore, the means of the two normalized PDFs are at 3.3. Damping behavior
1. It is seen that the distribution corresponding to the frac-
tal microstructure is shifted to smaller values of stress. The damping behavior of these composites is discussed
However, the decay of the tails is much slower. The tails next. This is relevant for many applications in which com-
at large stresses cross over, with the fractal carrying more posites are used as energy absorbing materials. In such sit-
extreme stress values than the random microstructure. uations, it is useful to inquire how should the inclusions be
This discussion holds for other fractal cases as well. distributed in order to obtain maximum damping at given
It is also interesting to look at the PDF of the pressure volume fraction of filler material. The objective of this sec-
within inclusions. As discussed in the introduction, this is tion is to provide a quantitative assessment of this issue
relevant for situations in which fillers are used to toughen and to determine the role of the spatial correlation of the
a brittle polymer. Toughening of epoxies using rubbery distribution of inclusions in defining the damping behavior
particles (e.g. Hayes and Seferis, 2001; Kinloch, 2003) is of the composite.
currently used in commercial products. In these applica- Two situations are discussed. In the first case the matrix
tions, the hydrostatic stress in inclusions produces cavita- is considered linear elastic with no damping and dissipa-
tion and/or filler–matrix interface debonding. Energy tion being allowed within inclusions. In the second case,
dissipation, leading to macroscopic toughness, is due to both matrix and inclusions are linear elastic and non-dissi-
both the deformation of inclusions during the cavitation pative, while damping is introduced in the interfaces.
process, and the plastic deformation of the matrix between These two limit situations are representative for different
inclusions. The matrix plastic deformation is promoted by classes of composites and the results are expected to shed

Fig. 6. Probability distribution functions for (a) the maximum principal stress in the matrix, rmax, and (b) the hydrostatic stress in inclusions, rh, in a fractal
microstructure with M = 9, P = 6, n = 5 (filled symbols) and the random microstructure with the same filler volume fraction (open symbols). These quantities
are normalized with the respective far-field (or system average) values.
 Author's personal copy

R.C. Picu et al. / Mechanics of Materials 69 (2014) 251–261 259

light on the microstructural optimization of these

materials.
Structures similar with those in Fig. 1 are modeled
using finite elements. The equation of motion is of the
form:
€ g þ ½Cfug
½Mfu _ þ ½Kfug ¼ f0g; ð7Þ

where [M], [C] and [K] are the mass, viscous damping and
stiffness matrices. A solution of the form fug ¼ f/gekt is
sought for Eq. (7), which leads to an eigenvalue problem
with eigenvalues appearing in complex conjugate pairs of
the form kr ¼ rr ixr , r = 1, . . ., n, where n is the total num-
ber of degrees of freedom of the problem. For subcritical
damping, the damping ratio
rr Fig. 8. Increase of the damping ratio in fractal microstructures relative to
fr ¼  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð8Þ
r2r þ x2r the corresponding random microstructures of same volume fraction,
versus the fractal dimension, D.
defines the ratio between the damping coefficient and the
critical damping coefficient for mode r. p
Note that for pro-
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
portional damping, the relation x0r ¼ x2r þ r2r can be Fig. 4, the dashed line is fitted to the blue circles which cor-
written between the eigenfrequency of mode r with damp- respond to fractal structures with M = 9, P = 5 and various n
ing, xr , and the corresponding eigenfrequency of the un- values. The data points and error bars are evaluated from
damped system, x0r . Therefore, x0r is always larger than sets of 100 replicas for each configuration.
xr . Below, we report the damping ratio f1 corresponding As the fractal dimension increases, the departure from
to the mode with lowest eigenfrequency. the random case is more pronounced. For the fractal struc-
When energy dissipation takes place only in interfaces ture with the largest fractal dimension considered, M = 91,
between fillers and matrix, interface elements are used to P = 42, n = 3, D = 1.701, the damping ratio is 53% larger than
introduce damping. that for the random microstructure of same volume frac-
tion. This significant increase is due entirely to the interac-
tion between inclusions which is enhanced by their
3.3.1. Damping in the filler volume hierarchical distribution.
Fig. 7 shows the damping ratio for the lowest eigenfre- All fractal structures considered exhibit enhancements
quency mode, f1 , function of the volume fraction of inclu- relative to the corresponding random cases, the increase
sions for various systems. The open symbols and the thick being function of D. Fig. 8 shows the percentage increase
line correspond to the random distribution of inclusions of the damping ratio relative to the random case of same
(Fig. 1(b)), while the other symbols correspond to fractal volume fraction versus the fractal dimension, for all cases
microstructures with various fractal dimensions. As in considered. A linear relationship emerges, with gains close
to 100% for fractal dimensions close to 1.9.

Fig. 7. Variation of the damping ratio, Eq. (8), with the volume fraction
for composites with random (open circles) and fractal microstructures.
The continuous and dashed lines are fitted to the results for random and Fig. 9. Variation of the damping ratio with the total length of interfaces
fractal (M = 9, P = 5, blue circles) microstructures. The other filled symbols per unit area of the model, for the case in which damping takes place in
correspond to fractal microstructures with M = 4, P = 3, n = 7 and 8, M = 9, the interfaces between matrix and fillers only, for random microstruc-
P = 6, n = 4 and 5 (green squares), M = 81, P = 42, n = 2 (red triangles), tures (open symbols and line) and for fractal microstructures with M = 9,
M = 36, P = 10, n = 2, M = 49, P = 14, n = 2, M = 64, P = 18, n = 2 (gray P = 5 and n = 2–5 (blue filled symbols). (For interpretation of the
circles). (For interpretation of the references to color in this figure legend, references to color in this figure legend, the reader is referred to the
the reader is referred to the web version of this article.) web version of this article.)
 Author's personal copy

260 R.C. Picu et al. / Mechanics of Materials 69 (2014) 251–261

3.3.2. Damping in the filler–matrix interfaces The present results can be used to improve the design of
Let us consider now that both matrix and filler materi- composites for structural applications. At present, the only
als are linear elastic and non-dissipative, and dissipation technology able to produce microstructures with specific
takes place in the interface between matrix and inclusions. distributions of inclusions is additive manufacturing. The
The comparison is performed between composites with low throughput and rather large cost of these methods
fractal and random microstructures having the same total prohibit their use for large scale composite fabrication, ex-
interface length. cept in few cases, such as when the final product is in the
Fig. 9 shows the damping ratio corresponding to the form of thin films. Powder metallurgy and electrolytic co-
lowest eigenfrequency mode for fractal microstructures deposition have been used to fabricate composites with
with M = 9, P = 5 and n = 2–5, and for the corresponding graded compositions, however, these methods do not in-
random microstructures. The emerging physical picture is sure obtaining the desired, position independent spatial
quite different from that discussed in Section 3.3.1. The correlation of properties discussed here. Further research
damping ratio increases linearly with the total interfacial is needed in the manufacturing area to develop methods
length, and the way the inclusions are distributed has no adequate for the fabrication of such materials.
influence on the overall dissipation. This observation is
interesting and provides guidance to composite design Acknowledgments
for damping applications.
This work was supported in part by a grant from the
Romanian National Authority for Scientific Research, CNCS
– UEFISCDI, project number PN-II-ID-PCE-2011-3-0120,
4. Conclusions contract 293/2011.

The objective of this work is to establish the role of spa-

References
tial correlations of the distribution of inclusions in defining
the mechanical behavior of particulate composites. In par- Argyris, J., Ciubotariu, C.I., Weingaertner, W.E., 2000. Fractal space
ticular, fractal microstructures having power law-corre- signatures in quantum physics and cosmology – I. Space, time,
lated characteristic functions are compared with random matter, fields and gravitation. Chaos Solitons Fractals 11, 1671–1719.
Bako, B., Hoffelner, W., 2007. Cellular dislocations patterning during
microstructures which have no spatial correlation. The plastic deformation. Phys. Rev. B 76, 214108.
central question is whether fractal microstructures have Banissadi, M., Ahzi, S., Garmestani, H., Ruch, D., Remond, Y., 2012. New
enhanced properties relative to the random microstruc- approximate solution for N-point correlation functions for
heterogeneous materials. J. Mech. Phys. Solids 60, 104–119.
tures with identical filler volume fraction. Baxter, R.J., 1982. Exactly Solved Models in Statistical Mechanics.
It is observed that composites with fractal microstruc- Academic Press, London.
tures are stiffer and have larger strain hardening rates, Bazant, Z., 1997. Scaling of quasibrittle fracture: hypotheses of invasive
and lacunar fractality, their critique and Weibull connection. Int. J.
the effect increasing with the fractal dimension. A pro- Frac. 83, 41–65.
nounced effect results when both materials are elastic, Beran, M.J., Molyeux, J., 1966. Continuum Theories. Wiley, NY.
but energy dissipation is allowed within fillers. In this case, Bergman, D.J., Kantor, Y., 1984. Critical properties of an elastic fractal.
Phys. Rev. Lett. 53, 511–514.
fractal microstructures dissipate up to 100% more energy Brockenbrough, J.R., Suresh, S., Wienecke, H.A., 1991. Acta Metall. Mater.
compared with the random microstructures of same filler 39, 735.
volume fraction. Interestingly, if viscous damping is intro- Carpinteri, A., Chiaia, B., Cornetti, P., 2004. A fractal theory for the
mechanics of elastic materials. Mater. Sci. Eng. A 365, 235–240.
duced only in the interface between fillers and matrix, the
Chen, Y.S., Choi, W., Papanikolaou, S., Sethna, J.P., 2010. Bending crystals:
distribution of inclusions has no effect on the overall emergence of fractal dislocation structures. Phys. Rev. Lett. 105,
damping of the composite. 105501.
On the local scale, the pressure within inclusions has Dvorak, G.J., 2013. Mechanics of Composite Materials. Springer, New York.
Dyskin, A.V., 2005. Effective characteristics and stress concentrations in
more extreme values in the fractal composite. This is materials with self-similar microstructure. Int. J. Solids Struct. 42,
important in situations in which cavitation in inclusions 477–502.
is used as a method to enhance the toughness of the com- Dzhaparidze, K., van Zanten, H., 2003. Krein’s spectral theory and the
Paley–Wiener expansion for fractional brownian motion. Stochastic
posite. The present results indicate that cavitation takes Section Report 13. Vrije Univ., Amsterdam.
place at a lower overall stress in the composite with fractal Falconer, K., 2003. Fractal Geometry – Mathematical Foundations and
microstructure. Applications. Wiley, Sussex. England.
Ghanem, R., Dham, S., 1998. Stochastic finite element analysis for
In conclusion, microstructures with long-range corre- multiphase flow in heterogeneous porous media. Transport Porous
lated distributions of inclusions are preferable in some Med. 32, 239–262.
applications, as they exhibit significantly different behav- Ghanem, G., Spanos, P.D., 1991. Stochastic Finite Elements: A Spectral
Approach. Springer-Verlag, New York.
ior relative to that of composites with random distribution
Gneiting, T., Schlather, M., 2004. Stochastic models that separate fractal
of inclusions. However, the mechanical response of fractal dimension and Hurst effect. SIAM Rev. 46, 269–282.
microstructures depends on the range of scales of the frac- Hashin, Z., 1965. On elastic behavior of fiber reinforced materials of
arbitrary transverse phase geometry. J. Mech. Phys. Solids 13, 119–
tal hierarchy and indirectly on the model/sample size rela-
134.
tive to the size of the smallest inclusions. This size effect is Hashin, Z., Shtrikman, S., 1962. On some variational principles in
absent in the case of random microstructures or micro- anisotropic and nonhomogeneous elasticity. J. Mech. Phys. Solids
structures with short-range correlations, provided the 10, 335–342.
Hayes, B.S., Seferis, J.C., 2001. Modification of thermosetting resins and
model is at least an order of magnitude larger than the cor- composites through preformed polymer particles: a review. Polym.
relation length. Compos. 22, 451–467.
 Author's personal copy

R.C. Picu et al. / Mechanics of Materials 69 (2014) 251–261 261

Huet, C., 1990. Application of variational concepts to size effects in elastic Parkinson, I.H., Fazzalari, N.L., 2000. Methodological principles for fractal
heterogeneous bodies. J. Mech. Phys. Solids 38, 813–841. analysis of trabecular bone. J. Microsc. 198, 134–142.
Kinloch, A.J., 2003. Toughening epoxy adhesives to meet today’s Phan-Tien, N., Milton, G.W., 1982. New bounds on the effective thermal
challenges. MRS Bull. 28, 445–448. conductivity of n-phase materials. Proc. R. Soc. London A 380, 333–
Kolwankar, K.M., 1998. Studies of Fractal Structures and Processes Using 348.
Methods of Fractional Calculus (Ph.D. thesis). University of Pune, Picu, R.C., Soare, M.A, 2009. Mechanics of materials with self-similar
India. hierarchical microstructure. In: Galvanetto, U., Aliabadi, M.F. (Eds.),
Kolwankar, K.M., Gangal, A.D., 1996. Fractional differentiability of Multiscale Modeling in Solid Mechanics – Computational Approaches.
nowhere differentiable functions and dimensions. Chaos 6, 505–524. Imperial College Press, London, pp. 295–332.
Liu, W.K., Belytschko, T., Mani, A., 1986. Probabilistic finite elements for Quinatanilla, J., Torquato, S., 1995. New bounds on the elastic moduli of
nonlinear structural dynamics. Comput. Meth. Appl. Mech. Eng. 56, suspensions of spheres. Appl. Phys. 77, 4361–4372.
61–81. Salganik, R.L., 1973. Mechanics of bodies with many cracks. Mech. Solids
Liu, W.K., Mani, A., Belytschko, T., 1987. Finite element methods in 8, 135–143.
probabilistic mechanics. Prob. Eng. Mech. 2, 201–213. Silnutzer, N.R., 1972. Effective Constants of Statistically Homogeneous
Mandelbrot, B.B., 1983. The Fractal Geometry of Nature. Freeman, New Materials (Ph.D. thesis). University of Pennsylvania, Philadelphia.
York. Soare, M.A., Picu, R.C., 2007. An approach to solving mechanics problems
Matthies, H.G., Brenner, C.E., Bucher, C.G., Soares, C.G., 1997. Uncertainties for materials with multiscale self-similar microstructure. Int. J. Solids
in probabilistic numerical analysis of structures and solids – Struct. 44, 7877–7890.
stochastic finite elements. Struct. Saf. 19, 283–336. Soare, M.A., Picu, R.C., 2008a. Boundary value problems defined on
Meyers, M.A., Lin, A.Y.M., Seki, Y., Chen, P.Y., Kad, B.K., Bodde, S., 2006. stochastic self-similar multiscale geometries. Int. J. Numer. Methods
Structural biological composites: an overview. J. Mater. 58, 35–41. Eng. 74, 668–696.
Milton, G.W., 1981. Bounds on the electromagnetic, elastic, and other Soare, M.A., Picu, R.C., 2008b. Spectral decomposition of random fields
properties of two-component composites. Phys. Rev. Lett. 46, 542– defined over the generalized Cantor set. Chaos Soliton Fractals 37,
545. 566–573.
Milton, G.W., 1982. Bound on the elastic and transport properties of two Suresh, S., Brockenbrough, J.R., 1993. Continuum models of deformation:
component composites. J. Mech. Phys. Sol. 30 (3), 177–191. metals reinforced with continuous fibers. In: Suresh, S., Mortensen, A.,
Nakamura, T., Suresh, S., 1993. Effect of thermal residual stress and fiber Needleman, A. (Eds.), Fundamentals of Metal Matrix Composites.
packing on deformation of metal–matrix composites. Acta Metall. Butterworth-Heinemann.
Mater. 41, 1665–1681. Tarasov, V.E., 2005a. Fractional hydrodynamics equations for fractal
Nemat-Nasser, S., Hori, M., 1999. Micromechanics: Overall Properties of media. Ann. Phys. 318, 286–307.
Heterogeneous Materials. North-Holland, Amsterdam. Tarasov, V.E., 2005b. Continuous medium model for fractal media. Phys.
Oshmyan, V.G., Patlashan, S.A., Timan, S.A., 2001. Elastic properties of Lett. A 336, 167–174.
Sierpinski-like carpets: finite-element-based simulation. Phys. Rev. E Torquato, S., 2002. Random Heterogeneous Materials: Microstructure and
64, 1–10, 056108. Macroscopic Properties. Springer, New-York.
Ostoja-Starzewski, M., 2007. Towards thermoelasticity of fractal media. J. Vaezi, M., Seitz, H., Yang, S., 2012. A review on 3D micro-additive
Therm. Stresses 30, 889–896. manufacturing technologies. Int. J. Adv. Manuf. Tech. http://
Ostoja-Starzewski, M., 2009. Extremum and variational principles for dx.doi.org/10.1007/s 00170-012-4605-2.
elastic and inelastic media with fractal geometries. Acta Mech. 205, Witten, T.A., Sander, L.M., 1983. Diffusion-limited aggregation. Phys. Rev.
161–170. 27, 5686–5697.
Ostoja-Starzewski, M., 2012. Elastic-plastic transitions in 3D random Zaiser, M., Hahner, P., 1999. Fractal analysis of deformation induced
materials: massively parallel simulations, fractal morphogenesis and dislocation patterns. Acta Mater. 47, 2463–2476.
scaling functions. Philos. Mag. 92, 2733–2758.
Papadrakakis, M., Papadopoulos, V., 1996. Robust and efficient methods
for stochastic finite element analysis using Monte Carlo simulations.
Comput. Meth. Appl. Mech. Eng. 134, 325–340.