# Springer

2006

Exact Results for the Homogenization of Elastic Fiber-Reinforced Solids at Finite Strainj Q.-C. HE1,jj, H. LE QUANG1 and Z.-Q. FENG2 1

Laboratoire de Me´canique, Universite´ de Marne-la-Valle´e, 19 rue A. Nobel, 77420 Champs sur Marne, France. E-mail: [email protected] 2 Laboratoire de Me´canique d’Evry, Universite´ d’Evry – Val d’Essonne, 40 rue du Pelvoux, 91020 Evry, France. Received 8 June 2005; in revised form 28 December 2005 Abstract. This work is concerned with the homogenization of solids reinforced by aligned parallel continuous fibers or weakened by aligned parallel cylindrical pores and undergoing large deformations. By alternatively exploiting the nominal and material formulations of the corresponding homogenization problem and by applying the implicit function theorem, it is shown that locally homogeneous deformations can be produced in such inhomogeneous materials and form a differentiable manifold. For every macroscopic strain associated to a locally homogeneous deformation field, the effective nominal or material stress–strain relation is exactly determined and connections are also exactly established between the effective nominal and material elastic tangent moduli. These results are microstructure-independent in the sense that they hold irrespectively of the transverse geometry and distribution of the fibers or pores. A porous medium consisting of a compressible Mooney–Rivlin material with cylindrical pores is studied in detail to illustrate the general results. Mathematics Subject Classifications (2000): 74B20, 74Q15. Key words: homogenization, fiber-reinforced composites, porous materials, homogeneous strain fields, microstructure, large deformations.

1. Introduction For the prediction of the effective mechanical properties of linear inhomogeneous or composite materials, micromechanical theories have been substantially developed and numerous general and specific results are now available (see [22, 26, 32]). A considerable effort has also been devoted in recent years to estimate or bound the effective behavior of nonlinear composite materials subjected to infinitesimal strains (see, e.g., [23, 30, 31]). By contrast, the homogenization of composite materials undergoing large deformations has been little investigated, j This work was the first time presented at the Euromech Colloqium 464 on BFiber-reinforced Solids: Constitutive Laws and Instabilities,^ September 28–October 1, 2004, Cantabria, Spain. jj Corresponding author.

Q.-C. HE ET AL.

although a framework was provided by Hill [15] more than 30 years ago and although interesting results were obtained in some general or particular situations (see, e.g., [10, 18, 19, 25, 27, 28]). The present work is concerned with the homogenization of elastic fiber-reinforced composite materials at finite strain, aiming to furnish a few exact results for this problem. In particular, by considering aligned parallel cylindrical pores in place of aligned parallel fibers, the present work is also relative to the homogenization of a class of porous materials. Within the framework of linear elasticity, Hill [14] discovered that the effective elastic moduli of a transversely isotropic fiber-reinforced composite with two transversely isotropic phases are connected by two exact relations independent of the transverse geometry and arrangement of the fibers at given volume fraction, so that the usual number of five independent elastic moduli for a transversely isotropic material reduces to three. This important result was extended by Dvorak [7] to totally anisotropic linear fiber-reinforced composites through systematically studying and exploiting the concept of homogeneous deformations in inhomogeneous materials. Inspired by the works of Hill [14] and Dvorak [7], He [12] generalized their conclusions to nonlinear elastic fiberreinforced composites subjected to infinitesimal strains by using the implicit function theorem of mathematical analysis. The present work poses and answers the following question: Is it possible to further extend the results of Hill [14] and Dvorak [7] to elastic fiber-reinforced composites undergoing large deformations? As will be seen, the key to giving a definite answer to the previous question lies in applying the implicit function theorem. However, a dilemma is encountered in the context of finite elasticity. On one side, as pointed by Hill [15] (see also [25]), for the transition from microvariables to macrovariables, the deformation gradient tensor and the nominal (or first Piola–Kirchhoff) stress tensor are the physically meaningful couple of kinematical and static variables. On the other hand, for the principle of objectivity [34] to be satisfied, the material (or Green–Lagrange) strain tensor and the material (or second Piola– Kirchhoff) stress tensor constitute a much more convenient couple of kinematical and static variables. This conflict is circumvented by alternate use of the nominal and material descriptions in solving the problem addressed in the present work. In the context of finite elasticity, the treatment of homogenization problems must account for the instability phenomena due to the fact that the strain-energy function cannot be convex with respect to the deformation gradient tensor (see, e.g., [2, 5]). Within the framework of periodic hyperelastic composites, a few authors, including Abeyaratne and Triantafyllidis [1], Triantafyllidis and Maker [33], Mu¨ller [24], Geymonat et al. [8] and Miehe et al. [21], have at different levels of generality studied the microscopic and macroscopic instabilities as well as their relations. In particular, considering a Neo-Hookean matrix with periodic cylindrical pores, Abeyaratne and Triantafyllidis [1] found that the homogenized

HOMOGENIZATION OF FIBER-REINFORCED ELASTIC SOLIDS

tangential tensor of the porous material loses strong ellipticity while the matrix material remains strongly elliptic. Examining a specific class of fiber-reinforced layered periodic composites, Triantafyllidis and Maker [33] shown that the bifurcation of such a composite at a wavelength much larger than the unit cell size is associated with the loss of strong ellipticity of the homogenized tangential tensor. The relations between microscopic and macroscopic instabilities were further investigated by Geymonat et al. [8] in the context of non-convex functional analysis, who, in a rather general way, demonstrated that long wavelength instabilities for the linearized problem lead to the loss of rank-one convexity of the homogenized strain-energy function. As pointed out by Mu¨ller [24], a major difficulty due to the occurrence of microscopic instabilities is that the representative volume element (RVE) used to formulate the homogenization problem does not a priori correspond to the unit cell of a periodic composite. The RVE must contain a sufficient number of unit cells so as to catch the energyminimizing modes (see [20, 21] for more details). In the present wok, we do not deal with any instability issue relative to the class of fiber-reinforced composites under consideration, although instability phenomena may occur at the microscopic and macroscopic levels. In fact, our work does not require that fibers be arranged periodically in the transverse plane, so that a random transverse distribution of fibers is not excluded. Except the recent one of Lopez-Pamies and Ponte Castan˜eda [19], the works reported in the literature on microscopic and macroscopic phenomena related to the homogenization of composites at finite strain are concerned almost exclusively with periodic microstructures. The relevant instability issue in the case of random microstructures remains largely open and is far beyond the scope of the present work. The results presented in this paper on the existence of homogeneous deformation fields and the exact relations should be taken to be valid only until microscopic instabilities appear. The paper is organized as follows. Section 2 is dedicated to the specification and formulation of the problem to be dealt with. In Section 3, while requiring the principle of objectivity to be checked, the implicit function theorem is applied to show that material and nominal homogeneous deformation fields can be generated in fiber-reinforced composites and form two differential manifolds which are related by the rotation group. In Section 4, the effective elastic nominal and material stress–strain relations are exactly determined for all macroscopic deformations associated to a locally homogeneous deformation field. General exact connections are also established between the effective elastic nominal and material tangent moduli evaluated at each macroscopic strain inducing a locally homogeneous deformation field. The results of Section 4 are illustrated in Section 5 through an example where a compressible Mooney– Rivlin material weakened by aligned parallel cylindrical pores is considered. The results obtained in the present paper are independent of the transverse morphology and distribution of the fibers at given volume fraction, as in the

Q.-C. HE ET AL.

case of linear elasticity. These results serve directly for partial determination of the effective elastic properties of fiber-reinforced composites undergoing large deformations and can be also used as benchmarks for analytical and numerical approximate methods elaborated to estimate them. Coordinate-free notation is adopted in the present paper. As a general rule, light-face (Greek or Latin) letters designate scalars, sets, spaces, domains or groups. Bold-face lowercase and uppercase Latin letters represent vectors and second-order tensors, respectively. Outline Latin letters are reserved for fourthorder tensors. The components of a vector, second- or fourth-order tensor are represented by the corresponding light-face letter with a suitable number of subscripts. Further, we denote by R3 the usual three-dimensional (3D) Euclidian space, by Lin the space of all second-order tensors on R3, by Sym the subspace of all second-order symmetric tensors on R3, and by Lin+ the cone of all second-order positive-determinant tensors belonging to Lin. Given any two elements A and B of Lin, we define, apart from the usual tensor product A B, two tensor products A B and A B of Kronecker type (see, e.g., [13]) by ðA BÞðu vÞ ¼ ðAuÞ ðBvÞ;

ðA BÞðu vÞ ¼ ðAvÞ ðBuÞ

for any two elements u and v of R3. It is also convenient to introduce another tensor product A B by A B ¼ 12 ðA B þ A BÞ: Next, the fourth-order identity tensor I on Lin, the transposition mapping T on Lin, and the fourth-order identity tensor 1 on Sym can be expressed in terms of the second-order identity tensor I as follows: I ¼ I I;

T ¼ I I;

1 ¼ I I:

For later use, remark that ðA BÞX ¼ AXBT ;

ðA BÞX ¼ AXT BT ;

ðA BÞX ¼ 12 A X þ XT BT

for any three elements A, B and X of Lin. In fact, the matrix components forms of A B, A B and A B are given by ðA BÞijkl ¼ Aik Bjl , ðA BÞijkl ¼ Ail Bjk and ðABÞijkl ¼ 12 Aik Bjl þ Ail Bjk . 2. Setting of the Problem In this section, we describe and formulate the problem with which the present work is concerned.

HOMOGENIZATION OF FIBER-REINFORCED ELASTIC SOLIDS

2.1. LOCAL STRESS–STRAIN RELATIONS AND MACROSCOPIC VARIABLES The inhomogeneous material investigated in this work consists of a solid matrix reinforced by aligned parallel continuous fibers (see Figure 1). The matrix and fiber phases are assumed to be individually homogeneous and perfectly bonded together across interfaces but no restrictions are imposed on the transverse geometry and distribution of the fibers, which may be random. The two-phase composite under consideration is thus homogeneous along the fiber direction while being inhomogeneous in the transverse plane. Let be the domain occupied by a specimen of the composite in the reference configuration and let (r) correspond to the sub-domain of phase r(= 1, 2). In what follows, we designate the boundary of by @, the phase interfaces by , the volume average over by < I >, the volume average over (r) by < I >r, and the initial volume fraction of phase r by c(r). For definiteness, the matrix is referred to as phase 1, the fibers are called phase 2, and the direction of the fibers is defined by the unit vector n (see Figure 1). The local (or microscopic) behavior of each phase of the composite is assumed to be nonlinearly elastic. Denoting by F the gradient of a deformation y : x 2 7! yðXÞ 2 R3 , and by P the resulting nominal stress, the local elastic stress–strain relation of the composite is described by Xr¼2 ðrÞ ðxÞPðrÞ ðFÞ: ð2:1Þ P ¼ Pðx; FÞ ¼ r¼1 Here, P(r)(F) is the nonlinear stress–strain relation of phase r and (r) is the char= acteristic function of (r) such that (r)(x) = 1 for x 2 (r) and (r)(x) = 0 for x 2 (r). The characteristic functions (r) entirely describe the microstructure of the composite. The elastic stress–strain relation (2.1) is required to satisfy the prin-

Figure 1. A fiber-reinforced composite or unidirectional pore-weakened material.

Q.-C. HE ET AL.

ciple of objectivity and the relevant material symmetry conditions [34]. Designating the 3D rotation group by SO(3), the principle of objectivity reads PðrÞ ðQFÞ ¼ QPðrÞ ðFÞ;

8F 2 Linþ :

8Q 2 SOð3Þ;

ð2:2Þ

If the material symmetry of phase r is specified by G(r) SO(3), then P(r)(F) must satisfy the condition PðrÞ ðFQÞ ¼ PðrÞ ðFÞQ;

8Q 2 GðrÞ ;

8F 2 Linþ :

ð2:3Þ

In this paper, the particularization of G(r) is not necessary. However, use will be explicitly or tacitly made of the following necessary and sufficient condition for (2.2) to be verified: PðrÞ ðFÞ ¼ RPðrÞ ðUÞ;

8F 2 Linþ ;

ð2:4Þ

where R and U are the respective rotation and right stretch tensors in the polar decomposition F = RU. In what follows, we make the additional assumption that the stress–strain relation P(r)(F) of every phase is continuously differentiable and verifies the initial condition PðrÞ ðIÞ ¼ 0:

ð2:5Þ

As will be seen, the continuous differentiability of P(r)(F) (r = 1, 2) is essential to the present work while the condition (2.5), which means that phase r is free of residual stress in the reference configuration, can be relaxed. Alternatively, the local behavior of the composite can be described by Xr¼2 ðrÞ ðxÞSðrÞ ðEÞ: ð2:6Þ S ¼ Sðx; EÞ ¼ r¼1 Here, E is the Green–Lagrange strain tensor defined by E ¼ 12 ðC IÞ

with C ¼ FT F ¼ U2 ;

ð2:7Þ

and S is the material stress tensor. The material stress–strain relation S(r)(E) of phase r is related to its nominal one P(r)(F) by PðrÞ ðFÞ ¼ FSðrÞ ðEÞ:

ð2:8Þ

With (2.1), the material symmetry requirement (2.3) and the residual-stress free condition (2.5) become SðrÞ QEQT ¼ QSðrÞ ðEÞQT ; 8Q 2 GðrÞ ; 8E 2 Sym; ð2:9Þ SðrÞ ð0Þ ¼ 0:

ð2:10Þ

HOMOGENIZATION OF FIBER-REINFORCED ELASTIC SOLIDS

It is known from continuum mechanics (see, e.g., [9]) that the main advantage of using the material stress–strain relation (2.6) in place of the nominal one (2.1) is that the principle of objectivity is automatically satisfied by the former. However, from the standpoint of micromechanics, the conjugated pair (P, F) is more appropriate. Indeed, the macroscopic nominal stress P and deformation gradient F are simply the volume averages of their microscopic counterparts over , which are dependent uniquely on the surface tractions and displacements on @ [15]: Z Z 1 1 F ¼ < F >¼ Fdv ¼ y xdv; ð2:11Þ jj jj @ 1 P ¼ < P >¼ jj

Z

1 Pdv ¼ jj

Z

ðPmÞ xdv;

ð2:12Þ

@

where m is the outward unit vector normal to @. Then, the macroscopic Green– Lagrange strain tensor E and material stress tensor S are obtained through F: T E ¼ 12 F F I ¼ 12 < F >T < F > I ; ð2:13Þ 1

S ¼ F P ¼ < F >1 < P > :

ð2:14Þ

It is important to remark that, in general, E 6¼ < E > and S 6¼ < S > (see Hill [15]). 2.2. FORMULATION OF THE HOMOGENIZATION PROBLEM Even though no limitations are imposed on the precise geometrical forms of the transverse cross-sections of the fibers and on the precise way in which the fibers are arranged in the transverse plane, we require that their transverse distribution be statistically uniform and that the largest dimension of the transverse crosssection of each fiber be much smaller than the dimensions of the specimen . These conditions make it meaningful to consider as an RVE and investigate the homogenization problem related to the composite. Since the phases are nonlinearly elastic and perfectly bonded together, the effective (or macroscopic) behavior of the composite is also nonlinearly elastic and can be described by the effective nominal stress–strain relation ^ F : P¼P

ð2:15Þ

The homogenization problem under consideration is to determine the effective ^ F from the knowledge of the microstructure characterstress–strain function P

Q.-C. HE ET AL.

ized by (r) and the phase stress–strain relations specified by P(r)(F). Together with the definition (2.12) of P, this is a boundary value problem of nonlinear inhomogeneous elasticity, which can be formulated as follows. Without loss of generality, consider the case where a linear transformation is prescribed on the boundary @ of : ð2:16Þ

yðxÞ ¼ Fx on @;

where F is a given second-order tensor belonging to Linþ . Due to the surface loading (2.16), a deformation field y, a deformation gradient field F and a nominal stress field P are produced over . From now on, it is convenient to denote the restriction of y, F and P to phase r as y(r), F(r) and P(r). In addition to the boundary condition (2.16) and to the local stress–strain law (2.1), y, F and P must verify the following field equations and perfect interface conditions: – the equation relating F(r) to y(r), FðrÞ ¼ ryðrÞ

ðrÞ ;

ð2:17Þ

over ðrÞ ;

ð2:18Þ

over

– the equilibrium of forces, DivPðrÞ ¼ 0

– the equilibrium of moments, PðrÞ FðrÞT ¼ FðrÞ PðrÞT

over ðrÞ ;

ð2:19Þ

– the deformation continuity across the interface, yð2Þ yð1Þ ¼ 0

on ;

– the stress continuity conditions across the interface, Pð2Þ Pð1Þ N? ¼ 0 on :

ð2:20Þ

ð2:21Þ

In (2.18), body forces have been neglected. In (2.21), n? is any unit vector normal to the interface between the matrix and the fibers in the reference configuration of the composite. Since the direction of the fibers is defined by the unit vector n, it follows that n? is characterized by n? n ¼ 0. The complete determination of the effective stress–strain relation (2.15) requires solving the nonlinear boundary value problem formulated by (2.16)– (2.21) together with (2.1). This problem is a very difficult one, since it is governed by a system of highly nonlinear partial differential equations with random coefficients. Consequently, it can be generally treated only numerically. However, as it will be shown below, using the fact that the composite is homogeneous along the fiber direction n, there are some special forms for the

HOMOGENIZATION OF FIBER-REINFORCED ELASTIC SOLIDS

deformation gradient tensor F in (2.16), such that the aforementioned nonlinear boundary value problem admits some particular homogeneous strain fields as exact solutions. To close this section, we remark that the homogenization of materials with unidirectional cylindrical pores can be considered as a special case of the problem formulated above. Indeed, it suffices to set Pð2Þ ¼ 0 or Sð2Þ ¼ 0.

3. Homogeneous Elastic Deformation Fields Consider a homogeneous deformation field yðxÞ ¼ F*x

ð3:1Þ

over ;

where F* 2 Lin+. Then, (2.18), (2.20) and (2.16) with F ¼ F* are immediately verified. Accounting for (2.1) and (2.17), satisfaction of the remaining equations, i.e., (2.19) and (2.21), requires that h i Pð2Þ ðF*Þ Pð1Þ ðF*Þ n? ¼ 0 PðrÞ ðF*ÞF*T ¼ F*PðrÞT ðF*Þ

on ;

ð3:2Þ

over ðrÞ :

ð3:3Þ

These two equations constitute the nominal necessary and sufficient conditions for the existence of homogeneous elastic deformations in the composite. Note that (3.2) must hold for any unit vector n? perpendicular to the fiber direction n. Thus, (3.2) can be conveniently written in the following equivalent form: h i ½IðI NÞ Pð2Þ ðF*Þ Pð1Þ ðF*Þ ¼ 0;

ð3:4Þ

where N = n n. Introducing the jump operator ½½ across the interface such that ½½ ¼ ð2Þ ð1Þ , we can further write (3.4) as ½IðI NÞ½½PðF*Þ ¼ 0. Using the objectivity principle (2.4) and the residual stress free condition (2.5), we immediately obtain the expected result that every rotation F* = R* is a trivial solution for (3.3) and (3.4). Of course, this is not a specific property of the composite under consideration but only a direct consequence of the objectivity principle together with (2.5). To find non-trivial solutions for (3.3) and (3.4), we first establish their material counterparts. Under the homogeneous deformation field condition (3.1), the material stress tensor S(r) of phase r is related to its nominal one P(r) by PðrÞ ¼ F*SðrÞ :

ð3:5Þ

Q.-C. HE ET AL.

By using (3.5) and invoking the objectivity principle condition (2.4), (3.2) and (3.3) can be equivalently rewritten as h i ð3:6Þ Sð2Þ ðE*Þ Sð1Þ ðE*Þ n? ¼ 0 on ; SðrÞT ðE*Þ ¼ SðrÞ ðE*Þ over

ðrÞ ;

ð3:7Þ

where E* is the Green–Lagrange strain tensor associated to F* by E* ¼ 12 ðC* 1Þ with

C* ¼ F*T F*:

ð3:8Þ

Equation (3.6) must hold for any unit vector n? normal to the fiber direction n. As in the work of He [12], 1defining the fourth-order orthogonal projection tensors N and N? by N? ¼ 1 N ¼ 1 N N;

ð3:9Þ

(3.6) can be written in a more convenient equivalent form: N? ½½SðE*Þ ¼ 0:

ð3:10Þ

While writing (3.10), the symmetric condition (3.7) has been tacitly employed. Bearing in mind the definition (3.9) of N? , we see that (3.10) is a system of five nonlinear equations with the six components of E* as unknowns. Accounting for the residual-stress free hypothesis (2.10), it is immediate that E* = 0 is a trivial solution of (3.10). Next, following He [12], let us show that (3.10) admits other solutions than E* = 0, i.e., non-zero homogeneous strain fields can be produced in the composite. First, the assumption that the nominal stress–strain relation P(r)(E) of the matrix or fibers is continuously differentiable has the consequence that the material tangent tensors LðrÞ ðEÞ ¼ rSðrÞ ðEÞ

ðr ¼ 1; 2Þ

ð3:11Þ

are continuous. Next, we introduce the jump of the material tangent tensor across as ½½L ¼ Lð2Þ Lð1Þ

ð3:12Þ

and define the kernel space of N ½½Lð0Þ by Ker N? ½½Lð0Þ ¼ E 2 Sym : N? ½½Lð0ÞE ¼ 0g: ?

ð3:13Þ

The dimension of this subspace of Sym is equal or greater than 1, i.e., q ¼ dim Ker N? ½½Lð0Þ 1;

ð3:14Þ

HOMOGENIZATION OF FIBER-REINFORCED ELASTIC SOLIDS

because dim Ker N? ¼ 1 and dimðKerðABÞÞ ¼ maxfdimðKerðAÞÞ; dimðKer ðBÞÞg for any two fourth-order tensors A and B. In view of the initial condition (2.10) and the fact (3.14), we can now apply the implicit function theorem of classical analysis (see, e.g., [3]) to infer the existence of a neighborhood D of E* = 0, such that the solutions of Equation (3.10) within D form a differentiable manifold M of dimension q: dimð M Þ ¼ q 1

ð3:15Þ

M ¼ E* 2 D Sym : N? ½½SðE*Þ ¼ 0 :

ð3:16Þ

with

Thus, it follows from (3.15) to (3.16) that homogeneous elastic strains can be generated in the fiber-reinforced solid under consideration. The corresponding homogeneous elastic deformation fields are obtained by (3.1) with F* ¼ R*U* ¼ R*ð2E* þ IÞ1=2 ;

R* 2 SOð3Þ;

E* 2 M:

For our purpose, it is useful to define another differentiable manifold: n o 1=2 0 M ¼ F* : F* ¼ R*ð2E* þ IÞ ; R* 2 SOð3Þ; E* 2 M :

ð3:17Þ

ð3:18Þ

We infer from (3.15) and (3.18) that dimðM 0 Þ ¼ q 4:

ð3:19Þ

In fact, the manifold M0 provides non-trivial solutions for (3.3) and (3.4). For later use, let us introduce the plane TM(E*) tangent to M at E* 2 M and the plane TM 0 ðF*Þ tangent to M0 at F* 2 M0 : TM ðE*Þ ¼ Ker N? ½½LðE*Þ ;

E* 2 M;

TM 0 ðF*Þ ¼ KerðIðI NÞÞ½½KðF*ÞÞ;

F* 2 M 0 :

ð3:20Þ ð3:21Þ

In (3.21), ½½KðF*Þ is the jump of the phase nominal tangent tensors across the interface , i.e., ½½K ¼ Kð2Þ Kð1Þ

ð3:22Þ

KðrÞ ¼ rPðrÞ

ð3:23Þ

with ðr ¼ 1; 2Þ;

Q.-C. HE ET AL.

and the kernel space of ðIðI NÞÞ½½KðF*Þ is defined as KerððIðI NÞÞ½½KðF*ÞÞ ¼ F 2 Lin :ðIðI NÞÞ½½KðF*ÞF ¼ 0 : ð3:24Þ In passing and for later use, observe that the basic formula (3.5) gives rise o the following relation between the nominal and material tangent tensors (see, e.g., [4, 6]): KðrÞ ðF*Þ ¼ ISðrÞ ðE*Þ þ ðF*IÞLðrÞ ðE*ÞðF*T IÞ:

ð3:25Þ

Lastly, it is directly deduced from (3.15) and (3.19) that dimðTM ðE*ÞÞ ¼ q 1;

dimðTM 0 ðF*ÞÞ ¼ q 4

ð3:26Þ

provided E* 2 M and F* 2 M0 .

4. Microstructure-Independent Exact Results The possibility of generating non-trivial homogeneous elastic deformations in the fiber-reinforced solid has at least two important consequences: (i) The effective nominal or material stress-strain relation can be explicitly and exactly determined for every F* 2 M0 ; (ii) exact connections exist between the effective elastic nominal (or material) tangent moduli evaluated at any F* 2 M0 (or E* 2 M). This section has the purpose of establishing these results which turn out to be independent of the transverse geometry and distribution of the fibers. 4.1. EXACT RESULTS FOR THE EFFECTIVE NOMINAL AND MATERIAL STRESS–STRAIN RELATIONS

For any surface loading specified by (2.16) with F ¼ F* 2 M 0 , the resulting deformation field is homogeneous over . Correspondingly, the nominal stress field P(x, F*) calculated by (2.1) is homogenous in each phase. Further, owing to the satisfaction of (3.4), the part ½IðI NÞPðx; F*Þ of P(x, F*) is homogeneous over . So, using (2.12), we can exactly determine the effective nominal stress– strain relation (2.15) for any F ¼ F* 2 M 0 as follows: ^ ðF*Þ ¼ < Pðx; F*Þ >¼ cð1Þ Pð1Þ ðF*Þ þ cð2Þ Pð2Þ ðF*Þ: P¼P

ð4:1Þ

Further, recalling that c(r) denotes the volume fraction of phase r, (4.1) can be split into h i ð1Þ ð1Þ ð2Þ ð2Þ ^ ðINÞPðF*Þ ¼ ðINÞ c P ðF*Þ þ c P ðF*Þ ; ð4:2Þ

HOMOGENIZATION OF FIBER-REINFORCED ELASTIC SOLIDS

^ ðF*Þ ¼ ½IðI NÞPð1Þ ðF*Þ ¼ ½IðI NÞPð2Þ ðF*Þ: ½IðI NÞP

ð4:3Þ

Using (2.14) and (4.1), we obtain the associated effective material stress: ^ ðF*Þ ¼ cð1Þ F*1 Pð1Þ ðF*Þ þ cð2Þ F*1 Pð2Þ ðF*Þ: S ¼ F*1 P

ð4:4Þ

Since P(r)(F*) verifies the principle of objectivity as specified by (2.4), we can invoke (2.8) and (2.14) to put (4.4) in a simpler equivalent form: ^ðE*Þ ¼ cð1Þ Sð1Þ ðE*Þ þ cð2Þ Sð2Þ ðE*Þ; S¼S

ð4:5Þ

^ðE*Þ ¼ F*1 P ^ðEÞ ^ ðF*Þ is the effective material stress–strain relation S where S evaluated at E ¼ E* 2 M. Interestingly, we observe that (4.5) amounts to S ¼ < Sðx; E*Þ >. This seems to be in contradiction with the general fact that E 6¼ < E > and S 6¼ < S > recalled just after equation (2.14). In reality, in the special case where a homogeneous deformation field y(x) = F*x over represents a solution for the homogenization problem, it results from (2.13) to (2.14) that E ¼ < E >¼ E* and S ¼ < S >¼ < Sðx; E*Þ >. When the phases of the composite are hyperelastic, i.e. the nominal stress– strain relation P(r)(F) of phase r in (2.1) is given by PðrÞ ðFÞ ¼

@wðrÞ ðFÞ @F

ð4:6Þ

where w(r)(F) is the strain-energy function of phase r, then the existence of homogeneous deformation fields implies that the effective strain-energy function ^ F can be exactly evaluated for all F* 2 M0 as follows: w ^ðF*Þ ¼ cð1Þ wð1Þ ðF*Þ þ cð2Þ wð2Þ ðF*Þ: w

ð4:7Þ

Thus, the Voigt bound established by Ogden [28] in the case of finite elasticity is achieved by fiber-reinforced composites whenever F* 2 M0 . 4.2. EXACT CONNECTIONS BETWEEN EFFECTIVE NOMINAL TANGENT MODULI By definition, the effective nominal tangent tensor A F associated to (2.15) is given by ^ F : A F ¼ rP

ð4:8Þ

Now, let us show that exact connections exist between the components of AðF*Þ with F* 2 M0 .

Q.-C. HE ET AL.

In view of (3.21) and (3.26)2, it makes sense to introduce the orthogonal projection operator HðF*Þ from Lin to TM 0 ðF*Þ through KerððIðI NÞÞ½½KðF*ÞÞ ¼ fF* : F* ¼ ðHðF*ÞÞF; F 2 Ling

ð4:9Þ

and the complementary orthogonal projection operator H? ðF*Þ by H? ðF*Þ ¼ I HðF*Þ:

ð4:10Þ

In particular, when detð½½KðF*ÞÞ 6¼ 0, we have HðF*Þ ¼ IðI NÞ and H? ðF*Þ ¼ IN. In what follows, it should be kept in mind that HðF*Þ and H? ðF*Þ are entirely determined by ðIðI NÞÞ½½KðF*Þ. With the aid of HðF*Þ and H? ðF*Þ, every variation F of a local deformation gradient field F* 2 M0 admits the following decomposition: F ¼ F* þ F0 ;

F* ¼ HF;

F0 ¼ H? F:

ð4:11Þ

In other words, each variation F of F* 2 M 0 can be uniquely decomposed into one component F* tangent to M 0 at F* and one component F0 perpendicular to the tangent plane TM 0 ðF*Þ. Remark that F* induces a homogeneous deformation field in the fiber-reinforced composite. So, for any variation F of a macroscopic deformation gradient F ¼ F* 2 M 0 , the decomposition F ¼ F* þ F0 ;

F* ¼ HF;

F0 ¼ H? F

ð4:12Þ

holds as well. The phase nominal incremental stress–strain relations associated to (2.1) and the effective nominal incremental stress–strain relation corresponding to (2.15) take the following respective forms: Xr¼2 Xr¼2 ðrÞ ðrÞ ð x Þ P ð F; F Þ ¼ ðrÞ ðxÞKðrÞ ðFÞF; P ¼ PðX; F; FÞ ¼ r¼1 r¼1 ð4:13Þ ^ F; F ¼ rP ^ F F ¼ A F F: P ¼ P

ð4:14Þ

When F* 2 M 0 , then F ¼ F* and use of the orthogonal projection operator HðF*Þ and H? ðF*Þ allows to split (4.13) into Xr¼2 Xr¼2 ðrÞ ðrÞ ðrÞ ðrÞ HP ¼ H ð x ÞK ð x ÞK H F þ H H? F; r¼1 r¼1 ð4:15Þ H? P ¼ H?

Xr¼2

Xr¼2 ðrÞ ðrÞ ? ðrÞ ðrÞ ð x ÞK ð x ÞK H F þ H H? F; r¼1 r¼1 ð4:16Þ

HOMOGENIZATION OF FIBER-REINFORCED ELASTIC SOLIDS

and (4.14) into HP ¼ HAHF þ HAH? F;

ð4:17Þ

H? P ¼ H? AHF þ H? AH? F:

ð4:18Þ

In (4.15)–(4.18) and in what follows, for notational simplicity the dependence on F* 2 M 0 is dropped as no confusion is possible. The volume average of (4.15) and the one of (4.16) give Xr¼2 Xr¼2 ðrÞ ðr Þ ðrÞ ðrÞ ? < F0 >r ; c HK H F* þ c HK H HP ¼ r¼1 r¼1 ð4:19Þ H? P ¼

Xr¼2 ðrÞ ? ðrÞ ðrÞ ? ðrÞ ? c H K H F* þ c H K H < F0 >r : r¼1 r¼1

Xr¼2

ð4:20Þ Using F0 ¼ cð1Þ < F0 >1 þ cð2Þ < F0 >2 and comparing the above two expressions with (4.17) and (4.18), we derive HðA < K >ÞHF* þ H A Kð1Þ H? F0 ¼ cð2Þ H½½KH? < F0 >2 ; ð4:21Þ H? ðA < K >ÞHF* þ H? A Kð1Þ H? F0 ¼ cð2Þ H? ½½KH? < F0 >2 ; ð4:22Þ where < K > ¼ cð1Þ Kð1Þ þ cð2Þ Kð2Þ . To obtain a set of exact connections between the components of A, we first pose F0 ¼ 0, eliminate < F0 >2 and prescribe F* arbitrarily in (4.21) and (4.22). This gives H ½½K H? ½½KH? 1 I ðA < K >Þ H ¼ 0:

ð4:23Þ

Next, setting F* = 0, eliminating < F0 >2 and letting F0 be arbitrary in (4.21) and (4.22), yields ð4:24Þ H ½½K H? ½½KH? 1 I A Kð1Þ H? ¼ 0: 1 In (4.23) and (4.24), inverse H? ½½KH? should understood in ? the partial be 1 H? ½½KH? Þ1 ¼ H? ½½KH? ¼ H? ½½KH? the sense that H ½½KH? H? . The exact relations (4.23) and (4.24) establish connections between the effective nominal tangent moduli evaluated at each point F* 2 M 0 . These

Q.-C. HE ET AL.

connections are microstructure-independent in the sense that they hold irrespectively of the transverse geometry and distribution of the fibers. They have the direct consequence that the number of independent effective nominal tangent moduli evaluated for all point F* 2 M 0 is reduced. The degree of reduction is determined by the dimension of the kernel space of ðIðI NÞÞ½½KðF*Þ as defined by (4.9). 4.3. EXACT CONNECTIONS BETWEEN EFFECTIVE MATERIAL TANGENT MODULI By definition, the effective material tangent tensor B E associated to the ^ðEÞ is provided by effective material stress–strain relation S ^ E : B E ¼ rS

ð4:25Þ

The fact that uniform strains can be produced in the composite and constitute a differential manifold M specified by (3.16) gives rise to exact connections between the components of BðE*Þ with E* 2 M. However, the method used in the last paragraph to establish connections between the components of the effective nominal tangent tensor AðF*Þ with F* 2 M 0 cannot be applied here, because E 6¼ < E > and S 6¼ < S >. Instead, an efficient way to achieve the purpose consists in starting from the established nominal connections (4.23) and (4.24) and using the following relation ^ðE*Þ þ ðF*IÞBðE*ÞðF*T IÞ AðF*Þ ¼ IS

ð4:26Þ

together with the corresponding phase relation (3.25). Introducing (4.26) and (3.25) into (4.23) and (4.24), we obtain 1 H½ðI½½S þ ðF*IÞ½½LðF*T IÞÞ H? ðI½½S þ ðF*IÞ½½LðF*T IÞÞH? IÞ ^ < S > þ ðF*IÞðB < L >ÞðF*T IÞÞH ¼ 0; ð4:27Þ I S 1 IÞ H½ðI½½S þ ðF*IÞ½½LðF*T IÞ H? ðI½½S þ ðF*IÞ½½LðF*T IÞÞH? ^ Sð1Þ þ ðF*IÞ B Lð1Þ ðF*T IÞ H? ¼ 0: ð4:28Þ I S These exact connections are much more complex than (4.23) and (4.24). However, in the particular case where F* = I and E* = 0, the residual-stress free ^ð0Þ ¼ 0. Correspondcondition (2.10) means that S(r)(0) = 0 and implies that S ingly, (4.27) and (4.28) can be considerably simplified into ? 1 I ðB0 < L0 >Þ H0 ¼ 0; H0 ½½L0 H? 0 ½½L0 H0

ð4:29Þ

HOMOGENIZATION OF FIBER-REINFORCED ELASTIC SOLIDS

ð1Þ ? 1 H0 ½½L0 H? I B 0 L 0 H? 0 ½½L0 H0 0 ¼ 0;

ð4:30Þ

where a quantity with the subscript 0 signifies that it is evaluated at E ¼ E* ¼ 0 or F* = I. In the case of hyperelastic infinitesimal strains and under the condition that ð2Þ ð1Þ ð2Þ ð1Þ L0 L0 is not singular, i.e., detðL0 L0 Þ 6¼ 0, the connections (4.29) and (4.30) reduce to those given by Dvorak [7] for linear elasticity. 5. Example The general results given in the previous sections are now illustrated by considering a simple but physically meaningful example where phase 1 is a compressible Mooney–Rivlin material and phase 2 consists of unidirectional cylindrical pores. The matrix phase of the porous material under consideration is characterized by the Mooney–Rivlin strain-energy function (see, e.g., [5, 16]): ð1Þ 1=2 J3 ð1Þ þ 2 ðJ1

e ð1Þ ðJi Þ ¼ 1 wð1Þ ðFÞ ¼ w

2 ð1Þ ð1Þ 1 2 þ 23 ln J3 ð1Þ

3Þ þ 3 ðJ2 3Þ;

ð5:1Þ

where i(1) Q 0 are the material parameters of the matrix phase and Ji are the principal invariants of the right Cauchy–Green strain tensor C = FTF defined by J1 ¼ trC;

J2 ¼

1 2

h i ðtrCÞ2 tr C2 ;

J3 ¼ det C:

ð5:2Þ

Clearly, for phase 2 made of cylindrical pores, we have w(2)(F) = 0. Applying the classical formula (see, e.g., [29]) Sð1Þ ¼ 2

e ð1Þ @w ¼2 @C

e ð1Þ e ð1Þ e ð1Þ e ð1Þ 1 @w @w @w @w I þ J1 C þ J3 C @J1 @J2 @J2 @J3 ð5:3Þ

to (5.1) gives the material stress of the matrix phase S

ð1Þ

¼2

ð1Þ 2

þ

ð1Þ 3 J1

h i ð1Þ ð1Þ ð1Þ ð1Þ 1=2 I 23 C þ 2 1 J3 J3 2 23 C1 : ð5:4Þ

Q.-C. HE ET AL.

Next, the expression for the corresponding nominal stress is obtained as follows: ð1Þ ð1Þ ð1Þ Pð1Þ ¼ FSð1Þ ¼ 2 2 þ 3 J1 F 23 FC h i ð1Þ 1=2 ð1Þ ð1Þ ð5:5Þ þ 2 1 J3 J3 2 23 FT : Regarding phase 2, we have S(2) = P(2) = 0. As shown in Section 3, finding homogeneous material strain fields amounts to solving the system (3.10) of nonlinear equations. To this end, let us introduce an orthonormal basis {e1, e1, e3} with the pore direction n coinciding with e1. Then, substituting (5.4) into the system of equations N? ½½SðEÞC ¼ 0 which is equivalent to (3.10), it follows that n h io 1=2 N? ð2 þ 3 J1 ÞC 3 C2 þ 1 J3 J3 2 23 I ¼ 0; ð5:6Þ where N? ¼ 1 e1 e1 e1 e1 . In (5.6) and in what follows, we set i = (1) i (i = 1, 2, 3) for notational simplicity. To find all the solutions for the system of nonlinear equation (5.6), we first observe the fact that, given any right Cauchy–Green strain tensor C, an orthonormal basis {e1, e2, e3} can always be chosen such that e1 corresponds to the direction n of cylindrical pores and the matrix of C relative to {e1, e2, e3} takes the form 2 3 C11 C12 C13 C ¼ 4 C12 C22 0 5: ð5:7Þ 0 C33 C13 In fact, it is always possible to rotate the transverse-plane orthonormal basis {e2, e3} about the axis e1 so that C23 = C32 = 0. Correspondingly, J1 ¼ C11 þ C22 þ C33 ;

ð5:8aÞ

2 2 J2 ¼ C11 C22 þ C22 C33 þ C33 C11 C12 C13 ;

ð5:8bÞ

2 2 C33 C13 C22 : J3 ¼ C11 C22 C33 C12

ð5:8cÞ

Introducing (5.7) and (5.8a, 5.8b, 5.8c) into (5.6), we obtain five nonlinear equations: 2 C12 þ 3 C33 C12 ¼ 0;

ð5:9Þ

2 C13 þ 3 C22 C13 ¼ 0;

ð5:10Þ

3 C12 C13 ¼ 0;

ð5:11Þ

HOMOGENIZATION OF FIBER-REINFORCED ELASTIC SOLIDS

qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 2 2 2 C C2 C 1 C11 C22 C33 C12 C33 C13 C22 C11 C22 C33 C12 33 13 22 2 þ 2 C22 þ 3 C11 C22 þ C33 C22 C12 2 23 ¼ 0; ð5:12Þ qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 2 2 2 C C2 C 1 C11 C22 C33 C12 C33 C13 C22 C11 C22 C33 C12 33 13 22 2 þ 2 C33 þ 3 C11 C33 þ C22 C33 C13 2 23 ¼ 0: ð5:13Þ In Appendix A, it is proved that, due to the requirement that C be positive definite, quations (5.9)–(5.13) admit only the solution expressed by C* ¼ U*2 ¼ 21 e1 e1 þ 22 ðI e1 e1 Þ

ð5:14Þ

with 1 ðtÞ ¼

1 t þ

pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 21 t2 4ð3 þ 1 t2 Þ½3 ðt4 2Þ þ 2 ðt2 1Þ ; 2ð3 þ 1 t2 Þt

2 ðt Þ ¼ t ð5:15Þ

where t 2]0, + 1 [ is a positive parameter. The tensors E* = (C* j I)/2 with C* specified by (5.14) and (5.15) are transversely isotropic with respect to e1 and form a manifold M of dimension q = 1 in the space Sym. Furthermore, using (3.18), we obtain the manifold n o M 0 ¼ F* : F* ¼ R*ð2E þ IÞ1=2 ; R* 2 SOð3Þ; E* 2 M

ð5:16Þ

in the space Lin. Then, all admissible homogeneous deformation fields for the porous material under consideration are given by (3.1) with F* 2 M 0 . It is worth noting that the necessary and sufficient condition (5.6) for the existence of homogeneous deformation fields in the porous material is nothing else than the demand that the stress field in the matrix phase be uniaxial along the direction of the cylindrical pores. This can be expected, since the strain field in the matrix is required to be uniform while no stresses are exerted on the pore surfaces. Furthermore, (5.14) and (5.15) can be viewed as characterizing the admissible strain states of the matrix phase associated to the uniaxial stress state. Next, applying the formulas (4.5), (4.1) and (4.7), we can exactly evaluate the effective material stress–strain relation at every E* 2 M and the effective

Q.-C. HE ET AL.

nominal stress–strain relation and strain-energy function at each F* 2 M for the porous material through ^ðE*Þ ¼ cð1Þ Sð1Þ ðE*Þ S¼S ¼ 2cð1Þ ð2 þ 3 J *1 ÞI 2cð1Þ 3 C* þ 2cð1Þ pﬃﬃﬃﬃﬃﬃ 1 J *3 J *3 2 23 C*1 ;

ð5:17Þ

^ ðF*Þ ¼ cð1Þ Pð1Þ ðF*Þ P¼P ¼ 2cð1Þ ð2 þ 3 J *1 ÞRU* 2cð1Þ 3 RU*3 pﬃﬃﬃﬃﬃﬃ þ 2cð1Þ 1 J *3 J *3 2 23 RU*1 ;

ð5:18Þ

^ðF*Þ ¼ cð1Þ wð1Þ ðF*Þ w 2 pﬃﬃﬃﬃﬃﬃﬃ ¼ cð1Þ 1 J3* 1 ð2 þ 23 Þ ln J *3 þ 2 ðJ *1 3Þ þ 3 ðJ *2 3Þ; ð5:19Þ where use is made of (5.4), (5.5) and (5.1). Substituting the expressions of C* and U* specified by (5.14)–(5.15) and the expressions of J*i specified by J*1 ðtÞ ¼ 21 ðtÞ þ 2t2 ;

J*2 ¼ 221 ðtÞt2 þ t4 ;

J*3 ¼ 21 ðtÞt4 ;

into (5.17)–(5.19), we obtain

1 t2 2 þ 23 ð1Þ 4 2 S ¼ 2c 1 t þ 23 t þ 2 e1 e1 ; 1 21 P ¼ 2c

ð1Þ

4 2 þ 23 1 t þ 23 t2 þ 2 1 1 t2 1

ð5:20Þ

ð5:21Þ

ðR*e1 Þ e1 ; ð5:22Þ

h 2 i ^ ðF*Þ ¼ cð1Þ 1 1 t2 1 ð2 þ 23 Þ ln 21 t4 w þ cð1Þ 2 21 þ 2t2 3 þ 3 221 t2 þ t4 3 ;

ð5:23Þ

where 1 is given by (5.15)1 and R* 2 SO(3). As expected, the effective behavior of the porous material evaluated at E* 2 M or F* 2 M0 is identical to the one of the matrix with the volume fraction c(1) as a multiplier. So, under the surface loading (2.6) with F ¼ F* 2 M 0 , the porous

HOMOGENIZATION OF FIBER-REINFORCED ELASTIC SOLIDS

* Þ along the direction of Figure 2. Effective material stress–strain relation S 11 ðE 11 cylindrical pores.

* to the uniform transverse strain E *22. Figure 3. Ratio of the uniform longitudinal strain E11

Q.-C. HE ET AL.

material under consideration behaves macroscopically as a compressible Mooney–Rivlin material. To graphically represent the relations (5.15) and (5.21), we take the numerical values cð1Þ ¼ 0:75;

1 ¼ 500 Pa; 2 ¼ 50 Pa;

3 ¼ 20 Pa

for the material parameters of the matrix. The effective material stress–strain relation S 11 ðE *11Þ along the direction of the cylindrical pores is shown in Figure 2, * /E 22 * is illustrated in Figure 3. and the ratio E 11 After determining all homogeneous deformation fields through (5.14) and (5.15), the exact connections (4.23)–(4.24) or (4.27)–(4.28) can be in principle specified for the porous material in question with no difficulties. However, the corresponding explicit expressions are algebraically cumbersome. Therefore, we do not detail them here. 6. Final Remarks In this work, the important results provided by Hill [14] and Dvorak [7] have been generalized to elastic solids reinforced by elastic fibers, or weakened by unidirectional cylindrical pores, and subjected to large deformations. The method elaborated to accomplish this generalization can be directly used to extend to the case of large deformations other microstructure-independent relations, such as the well-known Levin’s one and Rosen–Hashin’s one (see, e.g., [22]), established within the framework of linear elasticity or thermoelasticity and on the basis of the concept of uniform fields. Indeed, the main difficulty encountered in doing these extensions in the context of large deformations is due to the dilemma pointed out at the beginning of the paper. The method proposed in this work and consisting in alternatively exploiting the nominal and material description has turned out to be very efficient to solve the dilemma. All the results derived in Sections 2–5 are relative to a representative element (RVE) and hence valid for the homogenization of elastic fiber-reinforced composites under consideration. It is interesting to note that all these results remain valid even when is smaller than an RVE, under the condition that the volume fractions of the fibers and matrix in correspond to those prescribed for the composite. In this case, the overall properties of should not be considered as Feffective or homogenized_ but Fapparent_ (see, e.g., [11, 17]). As pointed out in the Introduction, an important issue which has not been addressed in this work is that of microscopic and macroscopic instabilities due to the non-convexity of the energy function of an elastic material at finite strain. All the results given in Sections 2–5 are valid only before the occurrence of microscopic instability. In other words, they are meaningful only within the microscopic stability domain. Even though the determination of the latter is an open problem for composites with random microstructures, we conjecture that,

HOMOGENIZATION OF FIBER-REINFORCED ELASTIC SOLIDS

except in some very special cases, the microscopic stability domain in the space of Green–Lagrange strain tensors E is in general a domain containing the zero strain tensor E = 0 as an interior point. This conjecture is strongly supported by the recent results of Lopez-Pamies and Ponte Castan˜eda [19]. Thus, we believe that the results obtained in this work are useful for the homogenization of fiberreinforced composites undergoing large deformations. Appendix A. Solution of the System of Equations (5.9)–(5.13) First of all, note that any physically meaningful solution for the system of nonlinear equations (5.9)–(5.13) must be such that the tensor C is positive definite, and that every material constant i involved in (5.9)–(5.13) is such that i Q 0. Then, we begin with the simplest equation (5.11). Satisfaction of the latter implies that C12 = 0 or C13 = 0. Without loss of generality, assume that C13 = 0. Then, equation (5.10) is immediately verified and equation (5.9) leads to C12 = 0 or C33 = j2/3. The last expression for C33 is not admissible, since it is in contradiction with the requirement that C be positive definite. So, at this stage, we can conclude that C12 ¼ C13 ¼ 0:

ðA:1Þ

This conclusion allows us to simplify (5.12) and (5.13) into pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 1 C11 C22 C33 C11 C22 C33 þ 2 C22 þ 3 ðC11 þ C33 ÞC22 2 23 ¼ 0; ðA:2Þ pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 1 C11 C22 C33 C11 C22 C33 þ 2 C33 þ 3 ðC11 þ C22 ÞC33 2 23 ¼ 0: ðA:3Þ Subtracting (A.3) from (A.2) gives ð2 þ 3 C11 ÞðC22 C33 Þ ¼ 0: The solution C11 = j2/3 is not admissible and, hence, the following one must hold: C22 ¼ C33 :

ðA:4Þ

Next, introducing (A.4) into (A.2) results in pﬃﬃﬃﬃﬃﬃﬃ 2 ð3 þ 1 C22 ÞC22 C11 1 C22 C11 þ 3 C22 þ 2 C22 2 23 ¼ 0: ðA:5Þ

Q.-C. HE ET AL.

This is a quadratic equation of C 1/2 11 and has the following admissible solution: pﬃﬃﬃﬃﬃﬃﬃ 1 C22 þ C11 ¼

qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 2 4C ð þ C Þ C 2 2 þ ðC 1Þ 21 C22 22 3 1 22 3 2 22 22 2ð3 þ 1 C22 ÞC22

:

ðA:6Þ Combining (5.7), (A.1), (A.4) and (A.6), we finally reach the conclusion that the system of equations (5.9)–(5.13) admits only the solution specified by (5.14) and 1/2 (5.15) where C 1/2 11 = 1 and C 22 = 2 = t with t Q 0. References 1. 2. 3. 4. 5. 6. 7. 8.

9. 10. 11. 12. 13. 14. 15. 16. 17. 18.

R. Abeyaratne and N. Triantafyllidis, An investigation of localization in a porous elastic material using homogenization theory. ASME J. Appl. Mech. 51 (1984) 481–486. J.M. Ball, Convexity conditions and existence theorems in nonlinear elasticity. Arch. Rational Mech. Anal. 63 (1977) 337–403. R.B. Borden, A Course in Advanced Calculus. Dover, New York (1998). P. Chadwick and R.W. Ogden, A theorem of tensor calculus and its application to isotropic elasticity. Arch. Rational Mech. Anal. 44 (1971) 54–68. P.G. Ciarlet, Mathematical Elasticity – Volume I: Three-Dimensional Elasticity. NorthHolland, Amsterdam (1988). A. Curnier, Computational Methods in Solid Mechanics. Kluwer, Dordrecht (1994). G.J. Dvorak, On uniform fields in heterogeneous media. Proc. R. Soc. Lond. A 431 (1990) 89–110. G. Geymonat, S. Mu¨ller and N. Triantafyllidis, Homogenization of nonlinear elastic materials, microscopic bifurcation and macroscopic loss of rank-one convexity. Arch. Rational Mech. Anal. 122 (1993) 231–290. M.E. Gurtin, An Introduction to Continuum Mechanics. Academic, Boston (1981). Z. Hashin, Large isotropic elastic deformations of composite and porous media. Int. J. Solids Structures 21 (1985) 711–720. Q.C. He, Effects of size and boundary conditions on the yield strength of heterogeneous materials. J. Mech. Phys. Solids 49 (2001) 2557–2575. Q.C. He, Uniform strain fields and microstructure-independent relations in nonlinear elastic fibrous composites. J. Mech. Phys. Solids 47 (1999) 1781–1793. Q.C. He and A. Curnier, A more fundamental approach to damaged elastic stress–strain relations. Int. J. Solids Structures 32 (1995) 1433–1457. R. Hill, Theory of mechanical properties of fiber-strengthened materials – I. Elastic behavior. J. Mech. Phys. Solids 12 (1964) 199–212. R. Hill, On constitutive macro-variables for heterogeneous solids at finite strain. Proc. R. Soc. London A 326 (1972) 131–147. G.A. Holzapfel, Nonlinear Solid Mechanics: A Continuum Approach for Engineering. Wiley, Chichester (2000). C. Huet, Application of variational concepts to size effects in elastic heterogeneous bodies. J. Mech. Phys. Solids 38 (1990) 813–841. O. Lopez-Pamies and P.P Castan˜eda, Second-order estimates for the large-deformation response of particle-reinforced rubbers. C. R. Mecanique 331 (2003) 1–8.

HOMOGENIZATION OF FIBER-REINFORCED ELASTIC SOLIDS

19.

20.

21.

22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33.

34.

O. Lopez-Pamies and P.P. Castan˜eda, Second-order estimates for the macroscopic response and loss of ellipticity in porous rubbers at large deformations. J. Elasticity 76 (2004) 247– 287. C. Miehe, Computational micro-to-macro transitions for discretized micro-structures of heterogeneous materials at finite strains based on the minimization of averaged incremental energy. Comput. Methods. Appl. Mech. Eng. 192 (2003) 559–591. C. Miehe, J. Schro¨der and M. Becker, Computational homogenization analysis in finite elasticity: Material and structural instabilities on the micro- and macro-scales of periodic composites and their interaction. Comput. Methods. Appl. Mech. Eng. 191 (2002) 4971–5005. G.W. Milton, The Theory of Composites. Cambridge University Press, Cambridge (2002). G.W. Milton and S.K. Serkov, Bounding the current in nonlinear conducting composites. J. Mech. Phys. Solids 48 (2000) 1295–1324. S. Mu¨ller, Homogenization of nonconvex integral functionals and cellular elastic materials. Arch. Rational Mech. Anal. 99 (1987) 189–212. S. Nemat-Nasser, Averaging theorems in finite deformation plasticity. Mech. Mater. 31 (1999) 493–523. S. Nemat-Nasser and M. Hori, Micromechanics: Overall Properties of Heterogeneous Materials. Elsevier, Amsterdam (1999). R.W. Ogden, On the overall moduli of non-linear elastic composite materials. J. Mech. Phys. Solids 22 (1974) 541–553. R.W. Ogden, Extremum principles in non-linear elasticity and their application to composites – I. Theory. Int. J. Solids Structures 14 (1978) 265–282. R.W. Ogden, Nonlinear Elastic Deformations. Ellis Horwood and Wiley, Chichester (1984). P. Ponte Castan˜eda, Second-order homogenization estimates for nonlinear composites incorporating field fluctuations. I. Theory. J. Mech. Phys. Solids 50 (2002) 737–757. P.P. Castan˜eda and P. Suquet, Nonlinear composites. In: E. van der Giessen and T.W. Wu (eds.). Advances in Applied Mechanics, vol. 34, Academic, San Diego (1998) pp. 171–302. S. Torquato, Random Heterogeneous Media: Microstructure and Macroscopic Properties. Springer, Berlin Heidelberg New York (2001). N. Triantafyllidis and B.N. Maker, On the comparison between microscopic and macroscopic instability mechanisms in a class of fiber-reinforced composites. ASME J. Appl. Mech. 52 (1985) 794–800. C. Truesdell and W. Noll, The Non-Linear Field Theories of Mechanics. Springer, Berlin Heidelberg New York (1992).