This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Generalized effective dynamic constitutive relation for heterogeneous media:
Beyond the quasi-infinite and periodic limits

Jeong-Ho Lee    Zhizhou Zhang    Grace X. Gu ggu@berkeley.edu Department of Mechanical Engineering, University of California, Berkeley, CA 94720
Abstract

Dynamic homogenization theories are powerful tools for describing and understanding the behavior of heterogeneous media such as composites and metamaterials. However, a major challenge in the dynamic homogenization theory is determining Green’s function of these media, which makes it difficult to predict their effective constitutive relations, particularly for the finite-size and/or non-periodic media in real-world applications. In this paper, we present a formulation for finding the elastodynamic effective constitutive relations for general heterogeneous media, including finite-size and non-periodic ones, by taking into account boundary terms in the Hashin-Shtrikman principle. Our proposed formulation relies on the infinite-body Green’s function of a reference homogeneous medium, making it free from the difficulty of determining Green’s function even for the homogenization of finite-size media. Additionally, we demonstrate the universal applicability of this formulation for both random and deterministic heterogeneous media. This work contributes to a better understanding of the homogenization theory and the design of next-generation metamaterials that require the accurate prediction of effective material characteristics for dynamic wave manipulation under desired operating environments.

I introduction

Heterogeneous media are considered a significant class of engineering materials, where their microstructures can be controlled to adapt to new operating environments and exhibit superior properties compared to raw materials. Metamaterials, a type of heterogeneous media consisting of building blocks large enough to be fabricated yet small enough to affect macroscopic material properties, have opened up new opportunities in the field of advanced materials. They exhibit unprecedented characteristics, particularly in manipulating dynamic material responses such as negative-index, subwavelength focusing, and invisibility cloaking for electromagnetic waves [1, 2, 3, 4, 5], collimators, superlenses, and inaudible cloaks for elastic (acoustic) waves [6, 7, 8, 9, 10, 11, 12, 13]. Extensive experimental and theoretical work has been performed to explore the rational designs of tailored metamaterials with desired macroscopic properties [14, 15, 16, 17, 18, 19].

Dynamic homogenization theories are powerful tools to describe and understand the behavior of heterogeneous media, by seeking to define effective constitutive relations and material characteristics that govern macroscopic wave propagation of a hypothetical homogenized medium. Willis first presented a linear elastodynamic homogenization theory based on the ensemble average technique and Green’s function of heterogeneous media [20, 21, 22, 23, 24, 25, 26, 27, 28, 29], from which it is discovered that the homogenized medium can exhibit non-classical features: 1) non-local material properties both in space and time; 2) tensorial density, i.e., not scalar; and 3) non-classical constitutive relation. In this non-classical constitutive relation, a coupling effect exists between linear momentum 𝝁\bm{\mu} and stress 𝝈\bm{\sigma} with velocity 𝒗\bm{v} and strain 𝜺\bm{\varepsilon}, called Willis coupling, and it has been extended in the case of piezoelectric heterogeneous media to a coupling effect between 𝝁\bm{\mu}, 𝝈\bm{\sigma}, and electric displacement 𝑫\bm{D} with 𝒗\bm{v}, 𝜺\bm{\varepsilon}, and electric field 𝑬\bm{E}, called electro-momentum coupling [30, 31, 32, 33]. Remarkably, these coupling effects occur even in the homogenization of deterministic media [34, 35]. In addition, this idea of Willis was particularized for periodic heterogeneous media based on Bloch-Floquet waves [36, 37, 38, 39, 40].

According to Willis’ book chapter [26], determining Green’s function of heterogeneous media is the biggest challenge when it comes to the dynamic homogenization theory. As Willis noted, “If the Green’s function could be determined, then all problems would be solved! Of course, this cannot be done.” To address this challenge, researchers have used the Hashin-Shtrikman principle [22, 24, 26, 28, 29, 41], which involves using Green’s function of a reference comparison medium that is homogeneous. However, previous works in the literature have constraints when determining Green’s function of the reference medium. For instance, Green’s function must satisfy certain boundary conditions, such as being zero along the boundaries (i.e., homogeneous boundary condition) or subject to the given types of boundary conditions of the target heterogeneous medium to be homogenized. Therefore, these approaches can only be suitable for infinite or quasi-infinite heterogeneous media large enough for boundaries to be not important. This limitation along with the Bloch-Floquet wave approach that only works for periodic cases poses challenges when it comes to predicting the effective constitutive relations of finite-size non-periodic heterogeneous media for real-world applications.

This paper presents two novel contributions to the linear elastodynamic homogenization of general heterogeneous media, including finite-size and non-periodic ones, for their effective constitutive relations. First, we introduce a new homogenization formulation based on the Hashin-Shtrikman principle, which takes into account boundary terms that have not been adequately considered previously. This formulation allows the use of the infinite-body Green’s function of the reference homogeneous medium. The effective constitutive relations obtained from this formulation display the non-classical features of the homogenized medium. Second, we demonstrate the universal applicability of the proposed formulation for both random and deterministic heterogeneous media. With these two contributions, this work provides a better understanding of the homogenization theory and enables the design of next-generation composites and metamaterials in which securing their effective characteristics is critical for dynamic wave manipulation.

II theory

Consider a boundary-value problem of a heterogeneous medium Ω\Omega which has random or deterministic microstructures but its boundary conditions and body force are taken to be sure, in arbitrary space dimension. Here, we proceed with piezoelectric media, which possess both Willis and electro-momentum coupling, as the homogenization formulation of standard (non-piezoelectric) media only with Willis coupling can be easily reduced by taking zero piezoelectricity as shown later in this paper. Thus, our goal is the development of a general formulation to construct effective dynamic constitutive relations of the heterogeneous medium, i.e.,

(𝝈𝑫𝝁)=[𝑪~𝑩~T𝑺~1𝑩~𝑨~𝑾~1𝑺~2𝑾~2𝝆~](:𝜺ϕ𝒗)\left(\begin{matrix}\langle\bm{\sigma}\rangle\\ \langle\bm{D}\rangle\\ \langle\bm{\mu}\rangle\end{matrix}\right)=\left[\begin{matrix}\widetilde{\bm{C}}&\widetilde{\bm{B}}^{T}&\widetilde{\bm{S}}_{1}\\ \widetilde{\bm{B}}&-\widetilde{\bm{A}}&\widetilde{\bm{W}}_{1}\\ \widetilde{\bm{S}}_{2}&\widetilde{\bm{W}}_{2}&\widetilde{\bm{\rho}}\end{matrix}\right]\left(\begin{matrix}:\langle\bm{\varepsilon}\rangle\\ \cdot\langle\bm{\nabla}\phi\rangle\\ \cdot\langle\bm{v}\rangle\end{matrix}\right) (1)

with the transpose operator ()T(\cdot)^{T}, in which the quantities with angle brackets mean the homogenized macroscopic ones where 𝜺=12[𝒖+(𝒖)T]\bm{\varepsilon}=\frac{1}{2}\left[\bm{\nabla}\bm{u}+(\bm{\nabla}\bm{u})^{T}\right] with spatial gradient operator \bm{\nabla} and displacement 𝒖\bm{u}, and 𝒗=𝒖˙\bm{v}=\dot{\bm{u}} with over-bar being time derivative, and ϕ=𝑬\bm{\nabla}\phi=-\bm{E} for electric potential ϕ\phi. The over-tilde is used to denote the effective material properties for stiffness 𝑪\bm{C}, dielectricity 𝑨\bm{A}, and piezoelectricity 𝑩\bm{B}, and mass density ρ\rho. We note here the effective mass density has a tensorial form. 𝑺~\widetilde{\bm{S}} and 𝑾~\widetilde{\bm{W}} are the Willis and electro-momentum coupling terms, respectively. As shown later in this paper, the effective material properties are non-local operators in general such that, for instance, (𝑪~:𝜺)(𝒙)=𝑪~(𝒙,𝒙ˇ):𝜺(𝒙ˇ)dΩ(𝒙ˇ)\left(\widetilde{\bm{C}}:\langle\bm{\varepsilon}\rangle\right)(\bm{x})=\int{\widetilde{\bm{C}}(\bm{x},\check{\bm{x}}):\langle\bm{\varepsilon}(\check{\bm{x}})\rangle}~{}d\Omega(\check{\bm{x}}) with respect to the spatial coordinates 𝒙\bm{x} and 𝒙ˇ\check{\bm{x}}.

The local mechanical and electric fields of the heterogeneous medium must satisfy the governing equations on Ω\Omega

{𝝈+𝒇=𝝁˙𝑫=q\begin{cases}\bm{\nabla}\cdot\bm{\sigma}+\bm{f}=\dot{\bm{\mu}}\\ \bm{\nabla}\cdot\bm{D}=q\end{cases} (2)

where 𝒇\bm{f} is body force, and qq is free charge density. The constitutive relations between the local kinetic and kinematic fields are written as

(𝝈𝑫𝝁)=[𝑪𝑩T0𝑩𝑨000ρ](:𝜺ϕ𝒗)\left(\begin{matrix}\bm{\sigma}\\ \bm{D}\\ \bm{\mu}\end{matrix}\right)=\left[\begin{matrix}\bm{C}&\bm{B}^{T}&0\\ \bm{B}&-\bm{A}&0\\ 0&0&\rho\end{matrix}\right]\left(\begin{matrix}:\bm{\varepsilon}\\ \cdot\bm{\nabla}\phi\\ \cdot\bm{v}\end{matrix}\right) (3)

in which the material properties depend on the microstructure materials of the heterogeneous medium at the local position. Here, we assume the standard tensor symmetries on the properties 𝑪\bm{C}, 𝑨\bm{A}, and 𝑩\bm{B}. Boundary conditions along the boundary Ω\partial\Omega are given as

{𝝈𝒏=𝒕 on Ωt;𝒖=𝒖bc on Ωu𝑫𝒏=c on Ωc;ϕ=ϕbc on Ωϕ\begin{cases}\bm{\sigma}\cdot\bm{n}=\bm{t}\text{~{}on~{}}\partial\Omega_{t};~{}~{}~{}~{}~{}~{}~{}\bm{u}=\bm{u}_{bc}\text{~{}on~{}}\partial\Omega_{u}\\ \bm{D}\cdot\bm{n}=-c\text{~{}on~{}}\partial\Omega_{c};~{}~{}~{}~{}\phi=\phi_{bc}\text{~{}on~{}}\partial\Omega_{\phi}\end{cases} (4)

where 𝒕\bm{t} and 𝒖bc\bm{u}_{bc} are prescribed traction and displacement, respectively, with the outward unit normal vector 𝒏\bm{n}, and cc and ϕbc\phi_{bc} are prescribed surface charge density and electric potential, respectively. Note that ΩtΩu=ΩcΩϕ=\partial\Omega_{t}\cap\partial\Omega_{u}=\partial\Omega_{c}\cap\partial\Omega_{\phi}=\varnothing.

Now following the Hashin-Shtrikman principle [42] schematically shown in Fig. 1, let’s introduce a homogeneous reference medium with constant properties 𝑪0\bm{C}_{0}, 𝑨0\bm{A}_{0}, 𝑩0\bm{B}_{0}, and ρ0\rho_{0}, which may or may not be one of the microstructure materials of the heterogeneous medium. Then, field polarizations are defined as

(𝝉𝝌𝝅)=([𝑪𝑩T0𝑩𝑨000ρ][𝑪0𝑩0T0𝑩0𝑨0000ρ0])(:𝜺ϕ𝒗)\left(\begin{matrix}\bm{\tau}\\ \bm{\chi}\\ \bm{\pi}\end{matrix}\right)=\left(\left[\begin{matrix}\bm{C}&\bm{B}^{T}&0\\ \bm{B}&-\bm{A}&0\\ 0&0&\rho\end{matrix}\right]-\left[\begin{matrix}\bm{C}_{0}&\bm{B}_{0}^{T}&0\\ \bm{B}_{0}&-\bm{A}_{0}&0\\ 0&0&\rho_{0}\end{matrix}\right]\right)\left(\begin{matrix}:\bm{\varepsilon}\\ \cdot\bm{\nabla}\phi\\ \cdot\bm{v}\end{matrix}\right) (5)

where 𝝉\bm{\tau}, 𝝌\bm{\chi}, 𝝅\bm{\pi} are the polarization of stress, electric displacement, and linear momentum, respectively. Hereafter, we will write matrix equations for the shake of simplicity via symbolic expression based on the notations in Fig. 1 — for instance of (3) and (5), h=Lbh=Lb and ζ=ΔLb\zeta=\Delta Lb with ΔL=LL0\Delta L=L-L_{0}, respectively. Then, the boundary-value problem of the heterogeneous medium can be equivalently represented, as shown in Fig. 1(b), via a boundary-value problem of the homogeneous reference medium with the constant properties L0L_{0} as

{𝝈r+(𝒇+𝝉𝝅˙)=𝝁˙r𝑫r=(q𝝌)\begin{cases}\bm{\nabla}\cdot\bm{\sigma}_{r}+(\bm{f}+\bm{\nabla}\cdot\bm{\tau}-\dot{\bm{\pi}})=\dot{\bm{\mu}}_{r}\\ \bm{\nabla}\cdot\bm{D}_{r}=(q-\bm{\nabla}\cdot\bm{\chi})\end{cases} (6)

where 𝝈r=𝝈𝝉\bm{\sigma}_{r}=\bm{\sigma}-\bm{\tau}, 𝝁r=𝝁𝝅\bm{\mu}_{r}=\bm{\mu}-\bm{\pi}, and 𝑫r=𝑫𝝌\bm{D}_{r}=\bm{D}-\bm{\chi} such that hr=L0bh_{r}=L_{0}b via the symbolic expression, with boundary conditions

{𝝈r𝒏=𝒕𝝉𝒏 on Ωt;𝒖=𝒖bc on Ωu𝑫r𝒏=c𝝌𝒏 on Ωc;ϕ=ϕbc on Ωϕ.\begin{cases}\bm{\sigma}_{r}\cdot\bm{n}=\bm{t}-\bm{\tau}\cdot\bm{n}\text{~{}on~{}}\partial\Omega_{t};~{}~{}~{}~{}~{}~{}~{}\bm{u}=\bm{u}_{bc}\text{~{}on~{}}\partial\Omega_{u}\\ \bm{D}_{r}\cdot\bm{n}=-c-\bm{\chi}\cdot\bm{n}\text{~{}on~{}}\partial\Omega_{c};~{}~{}~{}~{}\phi=\phi_{bc}\text{~{}on~{}}\partial\Omega_{\phi}.\end{cases} (7)

As we consider a linear elastodynamic system, this equivalent boundary-value problem can be decomposed via the superposition principle as shown in Fig. 1(c) and (d), in which Fig. 1(c) is the reference homogeneous medium subjected to the given Neumann boundary conditions of the heterogeneous medium to be homogenized. This leads to a resultant boundary-value problem building on the system of Fig. 1(d) that is independent of both the source terms 𝒇\bm{f} and qq and the Neumann boundary conditions 𝒕\bm{t} and cc as

{𝝈+(𝝉𝝅˙)=𝝁˙𝑫=(𝝌)\begin{cases}\bm{\nabla}\cdot\bm{\sigma}^{\prime}+(\bm{\nabla}\cdot\bm{\tau}-\dot{\bm{\pi}})=\dot{\bm{\mu}}^{\prime}\\ \bm{\nabla}\cdot\bm{D}^{\prime}=(-\bm{\nabla}\cdot\bm{\chi})\end{cases} (8)

such that h=L0bh^{\prime}=L_{0}b^{\prime} as the symbolic expression, with boundary condition

{𝝈𝒏=𝝉𝒏 on Ωt;𝒖=𝒖bc𝒖bc0 on Ωu𝑫𝒏=𝝌𝒏 on Ωc;ϕ=ϕbcϕbc0 on Ωϕ\begin{cases}\bm{\sigma}^{\prime}\cdot\bm{n}=-\bm{\tau}\cdot\bm{n}\text{~{}on~{}}\partial\Omega_{t};~{}~{}~{}~{}\bm{u}=\bm{u}_{bc}-\bm{u}_{bc_{0}}\text{~{}on~{}}\partial\Omega_{u}\\ \bm{D}^{\prime}\cdot\bm{n}=-\bm{\chi}\cdot\bm{n}\text{~{}on~{}}\partial\Omega_{c};~{}~{}~{}\phi=\phi_{bc}-\phi_{bc_{0}}\text{~{}on~{}}\partial\Omega_{\phi}\end{cases} (9)

where 𝒖bc0\bm{u}_{bc_{0}} and ϕbc0\phi_{bc_{0}} are arbitrary Dirichlet boundary conditions on the system of Fig. 1(c). Then, the problem can be solved via the boundary-element method using the infinite-body Green’s function of the reference homogeneous medium. By employing the Laplace transform to replace the time derivative by the Laplace variable s=iωs=-i\omega with angular frequency ω\omega, the solution is obtained as

(𝒖(𝒙)ϕ(𝒙))w(𝒙)=w(𝒙)+(BˇG(𝒙,𝒙ˇ))Tζ(𝒙ˇ)(BˇG(𝒙,𝒙ˇ))Tζ(𝒙ˇ)dΩ(𝒙ˇ)\left(\begin{matrix}\bm{u}(\bm{x})\\ \phi(\bm{x})\end{matrix}\right)\equiv w(\bm{x})=\langle w(\bm{x})\rangle+\int{\Big{(}\check{B}G(\bm{x},\check{\bm{x}})\Big{)}^{T}\langle\zeta(\check{\bm{x}})\rangle-\Big{(}\check{B}G(\bm{x},\check{\bm{x}})\Big{)}^{T}\zeta(\check{\bm{x}})}~{}d\Omega(\check{\bm{x}}) (10)

as the symbolic expression with

Bˇ=[ˇ00ˇs0];G(𝒙,𝒙ˇ)=[𝑮11(𝒙,𝒙ˇ)𝑮12(𝒙,𝒙ˇ)𝑮21(𝒙,𝒙ˇ)G22(𝒙,𝒙ˇ)]\check{B}=\left[\begin{matrix}\check{\bm{\nabla}}&0\\ 0&\check{\bm{\nabla}}\\ s&0\end{matrix}\right];~{}~{}G(\bm{x},\check{\bm{x}})=\left[\begin{matrix}\bm{G}_{11}(\bm{x},\check{\bm{x}})&\bm{G}_{12}(\bm{x},\check{\bm{x}})\\ \bm{G}_{21}(\bm{x},\check{\bm{x}})&G_{22}(\bm{x},\check{\bm{x}})\end{matrix}\right] (11)

in which the time-reduced Green’s function matrix G(𝒙,𝒙ˇ)G(\bm{x},\check{\bm{x}}) is defined to solve [43, 44]

{(𝑮11T:𝑪0T+𝑮21𝑩0)s2ρ0𝑮11T=𝜹xxˇ(𝑮11T:𝑩0T𝑮21𝑨0T)=𝟎(𝑮12:𝑪0T+G22𝑩0)s2ρ0𝑮12=𝟎(𝑮12:𝑩0TG22𝑨0T)=δxxˇ\begin{cases}\begin{aligned} \bm{\nabla}\cdot\left(\bm{\nabla}\bm{G}_{11}^{T}:\bm{C}_{0}^{T}+\bm{\nabla}\bm{G}_{21}\cdot\bm{B}_{0}\right)-s^{2}\rho_{0}\bm{G}_{11}^{T}&=-\bm{\delta}_{x\check{x}}\\ \bm{\nabla}\cdot\left(\bm{\nabla}\bm{G}_{11}^{T}:\bm{B}_{0}^{T}-\bm{\nabla}\bm{G}_{21}\cdot\bm{A}_{0}^{T}\right)&=\bm{0}\\ \bm{\nabla}\cdot\left(\bm{\nabla}\bm{G}_{12}:\bm{C}_{0}^{T}+\bm{\nabla}G_{22}\cdot\bm{B}_{0}\right)-s^{2}\rho_{0}\bm{G}_{12}&=\bm{0}\\ \bm{\nabla}\cdot\left(\bm{\nabla}\bm{G}_{12}:\bm{B}_{0}^{T}-\bm{\nabla}G_{22}\cdot\bm{A}_{0}^{T}\right)&=-\delta_{x\check{x}}\\ \end{aligned}\end{cases} (12)

where δxxˇ\delta_{x\check{x}} is the Dirac-delta function δ(𝒙𝒙ˇ)\delta(\bm{x}-\check{\bm{x}}), and 𝜹xxˇ=δ(𝒙𝒙ˇ)𝑰\bm{\delta}_{x\check{x}}=\delta(\bm{x}-\check{\bm{x}})\bm{I} with the identity tensor 𝑰\bm{I}. We note that the over-check of Bˇ\check{B} means the gradients are associated with 𝒙ˇ\check{\bm{x}} which comes from the non-local nature of Green’s function.

The solution (10) requires only the infinite-body Green’s function of a homogeneous medium, and its integration volume extends over the region occupied by the target heterogeneous medium to be homogenized. While different homogenization theories were derived based on the reference homogeneous medium in the previous literature [22, 24, 26, 28, 29, 41], the boundary terms in the Hashin-Shtrikman principle-based approach were not considered before. This results in constraints on Green’s function of the reference medium in their homogenization. Therefore, the previous formulations only work for heterogeneous media of infinite (at least quasi-infinite) size or simple geometries to which Green’s function can be determined.

The solution (10) yields the kinematic fields as

b(𝒙)=b(𝒙)+Γ(𝒙,𝒙ˇ)ζ(𝒙ˇ)Γ(𝒙,𝒙ˇ)ζ(𝒙ˇ)dΩ(𝒙ˇ)b(\bm{x})=\langle b(\bm{x})\rangle+\int{\Gamma(\bm{x},\check{\bm{x}})\langle\zeta(\check{\bm{x}})\rangle-\Gamma(\bm{x},\check{\bm{x}})\zeta(\check{\bm{x}})}~{}d\Omega(\check{\bm{x}}) (13)

with Γ(𝒙,𝒙ˇ)=B(BˇG(𝒙,𝒙ˇ))T\Gamma(\bm{x},\check{\bm{x}})=B\Big{(}\check{B}G(\bm{x},\check{\bm{x}})\Big{)}^{T} where BB is replacing the gradients in Bˇ\check{B} to be associated with 𝒙\bm{x}. We note that Γ(𝒙,𝒙ˇ)\Gamma(\bm{x},\check{\bm{x}}) is dominated by its singularity near 𝒙=𝒙ˇ\bm{x}=\check{\bm{x}} such that it can be written as Γ(𝒙,𝒙ˇ)=𝒟δxxˇ+ΓH(𝒙,𝒙ˇ)\Gamma(\bm{x},\check{\bm{x}})=\mathcal{D}\delta_{x\check{x}}+\Gamma^{H}(\bm{x},\check{\bm{x}}). 𝒟\mathcal{D} is constant which represents the contribution of an infinitesimal region around the singularity position while ΓH(𝒙,𝒙ˇ)\Gamma^{H}(\bm{x},\check{\bm{x}}) is of the contribution outside of the singular region. This integral equation can be written via a compact integral operator form as b=b+ΓζΓζb=\langle b\rangle+\Gamma\langle\zeta\rangle-\Gamma\zeta. From this, one can find a relation between ζ\langle\zeta\rangle and b\langle b\rangle that can be expressed in a general form of

ζ=Γ+[(+ΓΔL)++]b\langle\zeta\rangle=\Gamma^{+}\left[\Big{\langle}(\mathcal{I}+\Gamma\Delta L)^{+}\Big{\rangle}^{+}-\mathcal{I}\right]\langle b\rangle (14)

in which \mathcal{I} is the identity operator, and ()+(\cdot)^{+} denotes the inverse integral operator, e.g., Γ+Γ=\Gamma^{+}\Gamma=\mathcal{I}. Therefore, by the definition of field polarizations (i.e., h=L0b+ζh=L_{0}b+\zeta), the effective material property matrix LeffL_{eff}, for the effective constitutive relations h=Leffb\langle h\rangle=L_{eff}\langle b\rangle as the compact integral operator form of (1), can be finally obtained as

Leff=L0+Γ+[(+ΓΔL)++]L_{eff}=L_{0}+\Gamma^{+}\left[\Big{\langle}(\mathcal{I}+\Gamma\Delta L)^{+}\Big{\rangle}^{+}-\mathcal{I}\right] (15)

which well exhibits the non-local feature of the effective constitutive relations of the homogenized medium. This equation can be solved by various methods such as the iteration methods, e.g., the Neumann series and the resolvent method [45], and the discretization by quadrature rules [46] that can provide a direct connection to the finite element method.

This derived formulation is general but when 𝑪T=𝑪\bm{C}^{T}=\bm{C} (i.e. Cijkl=ClkjiC_{ijkl}=C_{lkji}), LeffL_{eff} is self-adjoint based on the self-adjointness of Green’s function, which delivers 𝑺~2=𝑺~1\widetilde{\bm{S}}_{2}=\widetilde{\bm{S}}_{1}^{\dagger} and 𝑾~2=𝑾~1\widetilde{\bm{W}}_{2}=\widetilde{\bm{W}}_{1}^{\dagger} with the formal adjoint operator ()(\cdot)^{\dagger} with respect to the spatial variable, e.g., 𝑺~1(𝒙,𝒙ˇ)=𝑺~1(𝒙ˇ,𝒙)\widetilde{\bm{S}}_{1}^{\dagger}(\bm{x},\check{\bm{x}})=\widetilde{\bm{S}}_{1}(\check{\bm{x}},\bm{x}) and 𝑾~1(𝒙,𝒙ˇ)=𝑾~1(𝒙ˇ,𝒙)\widetilde{\bm{W}}_{1}^{\dagger}(\bm{x},\check{\bm{x}})=\widetilde{\bm{W}}_{1}(\check{\bm{x}},\bm{x}). As (15) is derived from (10), it is unique regardless of both source terms and boundary conditions on the heterogeneous medium to be homogenized. This uniqueness is a reminiscence of the source-driven homogenization [47], and this implies that the non-uniqueness issue in the absence of both source terms and boundary conditions can be mitigated via applying eigenstrains [27]. Furthermore, since we haven’t made any assumptions on the homogenization strategy to map microscopic quantities to homogenized macroscopic ones, the proposed homogenization formulation is universal. We note that this formulation approach can also be employed to obtain effective constitutive relations in different physics, e.g., electromagnetism.

III validation

To validate the proposed homogenization formulation and demonstrate its universal applicability, we compare the effective constitution obtained from our formulation to 1-D analytical solutions of problems for random and deterministic heterogeneous media.

First, consider a homogenization problem based on the ensemble average technique for a finite-size random heterogeneous medium with a periodically repeating unit cell, as shown in Fig. 2(a). This unit cell consists of PMMA–BaTiO3–PZT4–PVDF layers with 0.5 mm, 0.5 mm, 1 mm, and 1.5 mm of each layer thickness, a total 3.5 mm of the unit cell thickness lucl_{uc}. Relying on the periodic microstructure geometry, the position of period can be considered as a uniformly distributed random variable that has luc1l_{uc}^{-1} of uniform probability density over the unit cell domain Ωuc\Omega_{uc}. Therefore, this feature reduces the ensemble average of any lucl_{uc}-periodic function Υ\Upsilon in the homogenization to a spatial average as Υ(x)=luc1ΩucΥy0(xy)𝑑y\langle\Upsilon(x)\rangle=l_{uc}^{-1}\int_{\Omega_{uc}}{\Upsilon_{y_{0}}(x-y)}~{}dy in which the lucl_{uc}-periodic function in realization yΩucy\in\Omega_{uc} is represented using Υy0(xy)\Upsilon_{y_{0}}(x-y) with the subscript y0y_{0} denoting the realization at y=0y=0. Assuming the absence of free charge and external electromagnetic fields, the electric governing equation D=0\nabla D=0 in (2) gives zero electric displacement across the medium domain, which leads to ϕ=(BA1)ε\nabla\phi=(BA^{-1})\varepsilon. Here, the bold is omitted in a 1-D manner. In this case, the boundary-value problem of the piezoelectric medium becomes a pure mechanical problem in which the local and effective constitutive relations are modified as

(σμ)=[Ca00ρ](εv);(σμ)=[C~aS~1aS~2aρ~a](εv)\left(\begin{matrix}\sigma\\ \mu\end{matrix}\right)=\left[\begin{matrix}C_{a}&0\\ 0&\rho\end{matrix}\right]\left(\begin{matrix}\varepsilon\\ {v}\end{matrix}\right);~{}~{}~{}\left(\begin{matrix}\langle\sigma\rangle\\ \langle\mu\rangle\end{matrix}\right)=\left[\begin{matrix}\widetilde{C}_{a}&\widetilde{S}_{1_{a}}\\ \widetilde{S}_{2_{a}}&\widetilde{\rho}_{a}\end{matrix}\right]\left(\begin{matrix}\langle\varepsilon\rangle\\ \langle v\rangle\end{matrix}\right) (16)

where the subscript aa denotes the apparent quantities as Ca=C+B2A1C_{a}=C+B^{2}A^{-1} for the microscopic property, and C~a=C~+B~2A~1\widetilde{C}_{a}=\widetilde{C}+\widetilde{B}^{2}\widetilde{A}^{-1}, S~1a=S~1+B~W~1A~1\widetilde{S}_{1_{a}}=\widetilde{S}_{1}+\widetilde{B}\widetilde{W}_{1}\widetilde{A}^{-1}, S~2a=S~2+B~W~2A~1\widetilde{S}_{2_{a}}=\widetilde{S}_{2}+\widetilde{B}\widetilde{W}_{2}\widetilde{A}^{-1}, and ρ~a=ρ~+W~1W~2A~1\widetilde{\rho}_{a}=\widetilde{\rho}+\widetilde{W}_{1}\widetilde{W}_{2}\widetilde{A}^{-1} for the effective properties. By the symbolic expression, we write (16) as ha=Labah_{a}=L_{a}b_{a} and ha=Leffaba\langle h_{a}\rangle=L_{eff_{a}}\langle b_{a}\rangle respectively for the local and effective relation, with ha=(σμ)Th_{a}=(\sigma~{}\mu)^{T} and ba=(εv)Tb_{a}=(\varepsilon~{}v)^{T}. Then, (13) is reduced, via the compact integral operator form, as

ba=ba+ΓaζaΓaζab_{a}=\langle b_{a}\rangle+\Gamma_{a}\langle\zeta_{a}\rangle-\Gamma_{a}\zeta_{a} (17)

where ζa=(τπ)T\zeta_{a}=(\tau~{}\pi)^{T}, and Γa(x,xˇ)=Ba(BˇaTGa(x,xˇ))\Gamma_{a}(x,\check{x})=B_{a}\Big{(}\check{B}_{a}^{T}G_{a}(x,\check{x})\Big{)} with Bˇa=(ˇs)T\check{B}_{a}=(\check{\nabla}~{}s)^{T} and BaB_{a} replacing the gradient to be associated with xx, which can be decomposed as Γa(x,xˇ)=𝒟aδxxˇ+ΓaH(x,xˇ)\Gamma_{a}(x,\check{x})=\mathcal{D}_{a}\delta_{x\check{x}}+\Gamma_{a}^{H}(x,\check{x}), where 𝒟a\mathcal{D}_{a} is constant, based on its singularity near x=xˇx=\check{x}. The apparent infinite-body time-reduced Green’s function Ga(x,xˇ)G_{a}(x,\check{x}) is defined to solve

(GaC0a)s2ρ0Ga=δ(xxˇ)\nabla\left(\nabla G_{a}C_{0_{a}}\right)-s^{2}\rho_{0}G_{a}=-\delta(x-\check{x}) (18)

where C0aC_{0_{a}} is the apparent stiffness of the reference homogeneous medium, i.e., [48]

Ga(x,xˇ)=(2ik0aC0a)1eik0a|xxˇ|G_{a}(x,\check{x})=(2ik_{0_{a}}C_{0_{a}})^{-1}e^{-ik_{0_{a}}|{x-\check{x}}|} (19)

with k0a=ωρ0/C0ak_{0_{a}}=\omega\sqrt{\rho_{0}/C_{0_{a}}}. Then, this leads to

Leffa=L0a+Γa+[(+ΓaΔLa)++]L_{eff_{a}}=L_{0_{a}}+\Gamma_{a}^{+}\left[\Big{\langle}(\mathcal{I}+\Gamma_{a}\Delta L_{a})^{+}\Big{\rangle}^{+}-\mathcal{I}\right] (20)

where ΔLa=LaL0a\Delta L_{a}=L_{a}-L_{0_{a}} with L0aL_{0_{a}} being the apparent material property matrix of the reference homogeneous medium. We note that taking the piezoelectricity B0B\rightarrow 0 exactly recovers the effective constitutive relations of standard heterogeneous media with only Willis coupling. One can easily expand it to the multidimensional case. Fig. 2(b-e) show macroscopic stress and linear momentum fields across the medium domain, under prescribing displacements at its left and right ends respectively by zero and 0.2% of the medium size, for the different number of the unit cell at time frequency fq(=ω2π)=0.24f_{q}(=\frac{\omega}{2\pi})=0.24 MHz. The fields obtained using (20) are compared with the ensemble average of analytical solutions of each ensemble. These results show that our proposed homogenization formulation provides an excellent estimate of the effective material properties for finite-size random heterogeneous media and displays well the necessity of the non-classical coupling terms in the effective constitutive relations of the homogenized medium, i.e., Willis and electro-momentum coupling. Fig. 2(f) and (g) visualize the apparent coupling coefficients and confirm their non-local feature in space and the adjoint relation between S~1a\widetilde{S}_{1_{a}} and S~2a\widetilde{S}_{2_{a}}.

Second, we extract the non-classical coupling terms of the homogenized medium for a finite-size non-periodic deterministic heterogeneous medium and compare them to benchmark data of [35]. Consider a piezoelectric heterogeneous element made of 2.62 mm of PZT4, 0.19 mm of BaTiO3, and 0.19 mm of PVDF. We here also assume the absence of free charge and external electric fields. The microstructures of this element possess wavelength much larger than those characteristic lengths, i.e., kl1kl\ll 1 with wavenumber kk and length ll of the layer. When a heterogeneous medium is at the subwavelength scale, or when a single isolated element is used rather than a bulk of mutually interacting elements, the effective constitutive relations become local in space, termed Milton-Briane-Willis equation [49, 50]. It has been studied that the local coupling terms originate from broken inversion asymmetry of the microstructures [51, 47]. Then, the apparent kinematic field vector bab_{a} in the modified constitutive relations (16) can be related to the state vector ς=(uσ)T\varsigma=(u~{}\sigma)^{T} via the material properties as ba=Mςb_{a}=M\varsigma and ba=Meffς\langle b_{a}\rangle=M_{eff}\langle\varsigma\rangle via the symbolic expression, with

M=[0Ca1s0];Meff=[sS~1aC~a1C~a1s0]M=\left[\begin{matrix}0&C_{a}^{-1}\\ s&0\end{matrix}\right];~{}~{}~{}M_{eff}=\left[\begin{matrix}-s\widetilde{S}_{1_{a}}\widetilde{C}_{a}^{-1}&\widetilde{C}_{a}^{-1}\\ s&0\end{matrix}\right] (21)

where the derivative of the state vector can also be calculated using the material properties and itself as ς=KMς\nabla\varsigma=KM\varsigma and ς=KeffMeffς\nabla\langle\varsigma\rangle=K_{eff}M_{eff}\langle\varsigma\rangle, with

K=[100sρ];Keff=[10sS~2asρ~a].K=\left[\begin{matrix}1&0\\ 0&s\rho\end{matrix}\right];~{}~{}~{}K_{eff}=\left[\begin{matrix}1&0\\ s\widetilde{S}_{2_{a}}&s\widetilde{\rho}_{a}\end{matrix}\right]. (22)

For the ii-th layer, the local state vectors at its two ends xL(i)x_{L}^{(i)} and xR(i)x_{R}^{(i)} (xL(i)<xR(i)x_{L}^{(i)}<x_{R}^{(i)}) are related via ς(xR(i))=Tl(i)ς(xL(i))\varsigma(x_{R}^{(i)})=T_{l}^{(i)}\varsigma(x_{L}^{(i)}) with the layer’s transfer matrix Tl(i)T_{l}^{(i)} as

Tl(i)=[cos(ka(i)l(i))(Ca(i)ka(i))1sin(ka(i)l(i))Ca(i)ka(i)sin(ka(i)l(i))cos(ka(i)l(i))]T_{l}^{(i)}=\left[\begin{matrix}\cos(k_{a}^{(i)}l^{(i)})&(C_{a}^{(i)}k_{a}^{(i)})^{-1}\sin(k_{a}^{(i)}l^{(i)})\\ -C_{a}^{(i)}k_{a}^{(i)}\sin(k_{a}^{(i)}l^{(i)})&\cos(k_{a}^{(i)}l^{(i)})\end{matrix}\right] (23)

with ka(i)=ωρ(i)/Ca(i)k_{a}^{(i)}=\omega\sqrt{\rho^{(i)}/C_{a}^{(i)}}. By the subwavelength characteristic, ΓaH(x,xˇ)\Gamma_{a}^{H}(x,\check{x}) can be expressed by expanding it around the element ends upto the linear term such that

ΓaH(x,xˇ)\displaystyle\Gamma_{a}^{H}(x,\check{x}) =ΓaH(xxˇ)\displaystyle=\Gamma_{a}^{H}(x-\check{x}) (24)
{ΓaH(lt)+ΓaH(lt)[xxˇlt](xxˇ)ΓaH(lt)+ΓaH(lt)[xxˇ+lt](x<xˇ)\displaystyle\approx\begin{cases}\begin{aligned} \Gamma_{a}^{H}(l_{t})+\nabla\Gamma_{a}^{H}(l_{t})\left[x-\check{x}-l_{t}\right]~{}~{}~{}~{}~{}~{}~{}&(x\geq\check{x})\\ \Gamma_{a}^{H}(-l_{t})+\nabla\Gamma_{a}^{H}(-l_{t})\left[x-\check{x}+l_{t}\right]~{}~{}~{}&(x<\check{x})\end{aligned}\end{cases}

with the element total length ltl_{t}. Then, based on the continuity of the state vector at the layer intersection and ζa=(LeffaL0a)ba\langle\zeta_{a}\rangle=(L_{eff_{a}}-L_{0_{a}})\langle b_{a}\rangle, integrating (17) over Ω(x)\Omega(x) gives an equation to find LeffaL_{eff_{a}} of the deterministic heterogeneous element at the subwavelength scale as

[inl(1(i)+2(i)3(i))(~1+~2~3)]ς(xL(1))=0\left[\sum_{i}^{n_{l}}\left(\mathcal{E}_{1}^{(i)}+\mathcal{E}_{2}^{(i)}-\mathcal{E}_{3}^{(i)}\right)-(\widetilde{\mathcal{E}}_{1}+\widetilde{\mathcal{E}}_{2}-\widetilde{\mathcal{E}}_{3})\right]\varsigma(x_{L}^{(1)})=0 (25)

in terms of the arbitrary ς(xL(1))\varsigma(x_{L}^{(1)}) with nln_{l} being the number of layer, in which the fact that the state vector is sure along the boundary is used, i.e., ς(xL(1))=ς(xL(1))\langle\varsigma(x_{L}^{(1)})\rangle=\varsigma(x_{L}^{(1)}), with

1(i)=(I+𝒟aΔLa(i))K(i)1(T(i)T(i1))\mathcal{E}_{1}^{(i)}=\left(I+\mathcal{D}_{a}\Delta L_{a}^{(i)}\right)K^{(i)^{-1}}\left(T^{(i)}-T^{(i-1)}\right) (26a)
2(i)=ΓaHΔLa(i){K(i)1(xR(i)T(i)xL(i)T(i1))(K(i)M(i)K(i))1(T(i)T(i1))}\mathcal{E}_{2}^{(i)}=\Gamma_{a-}^{H}\Delta L_{a}^{(i)}\left\{K^{(i)^{-1}}\left(x_{R}^{(i)}T^{(i)}-x_{L}^{(i)}T^{(i-1)}\right)-\left(K^{(i)}M^{(i)}K^{(i)}\right)^{-1}\left(T^{(i)}-T^{(i-1)}\right)\right\} (26b)
3(i)=Γa+HΔLa(i){K(i)1((xR(i)lt)T(i)(xL(i)lt)T(i1))(K(i)M(i)K(i))1(T(i)T(i1))}\mathcal{E}_{3}^{(i)}=\Gamma_{a+}^{H}\Delta L_{a}^{(i)}\left\{K^{(i)^{-1}}\left((x_{R}^{(i)}-l_{t})T^{(i)}-(x_{L}^{(i)}-l_{t})T^{(i-1)}\right)-\left(K^{(i)}M^{(i)}K^{(i)}\right)^{-1}\left(T^{(i)}-T^{(i-1)}\right)\right\} (26c)
~1=(I+𝒟a(LeffaL0a))Keff1(T(nl)I)\widetilde{\mathcal{E}}_{1}=\Big{(}I+\mathcal{D}_{a}(L_{eff_{a}}-L_{0_{a}})\Big{)}K_{eff}^{-1}\left(T^{(n_{l})}-I\right) (26d)
~2=ΓaH(LeffaL0a){Keff1(ltT(nl))(KeffMeffKeff)1(T(nl)I)}\widetilde{\mathcal{E}}_{2}=\Gamma_{a-}^{H}(L_{eff_{a}}-L_{0_{a}})\Bigg{\{}K_{eff}^{-1}\left(l_{t}T^{(n_{l})}\right)-\left(K_{eff}M_{eff}K_{eff}\right)^{-1}\left(T^{(n_{l})}-I\right)\Bigg{\}} (26e)
~3=Γa+H(LeffaL0a){Keff1(ltI)(KeffMeffKeff)1(T(nl)I)}\widetilde{\mathcal{E}}_{3}=\Gamma_{a+}^{H}(L_{eff_{a}}-L_{0_{a}})\Bigg{\{}K_{eff}^{-1}\left(l_{t}I\right)-\left(K_{eff}M_{eff}K_{eff}\right)^{-1}\left(T^{(n_{l})}-I\right)\Bigg{\}} (26f)

where T(i)=Tl(i)Tl(i1)Tl(1)T^{(i)}=T_{l}^{(i)}T_{l}^{(i-1)}\ldots T_{l}^{(1)} with Tl(0)=IT_{l}^{(0)}=I, Γa+H=ΓaH(lt)\Gamma_{a+}^{H}=\Gamma_{a}^{H}(l_{t}) and ΓaH=ΓaH(lt)+ΓaH(lt)lt\Gamma_{a-}^{H}=\Gamma_{a}^{H}(-l_{t})+\nabla\Gamma_{a}^{H}(-l_{t})l_{t}, and II is the identity matrix. This equation can be easily solved by using numerical methods, such as the multivariate Newton–Raphson method, with respect to the components in LeffaL_{eff_{a}}, i.e., C~a\widetilde{C}_{a}, S~1a\widetilde{S}_{1_{a}}, S~2a\widetilde{S}_{2_{a}}, and ρ~a\widetilde{\rho}_{a}. Fig. 3 shows the pure imaginary apparent coupling coefficients of the target heterogeneous element and its inversion, i.e., when the first and third layers are interchanged. The coupling coefficients obtained from our formulation have S~1a=S~2a\widetilde{S}_{1_{a}}=\widetilde{S}_{2_{a}} and are opposite in sign for the regular and inverted elements, i.e., a direction-dependent response, which is coincident with the observation in [35]. The obtained coupling coefficient values are also well agreed with the benchmark data. These results demonstrate the universal applicability of our proposed formulation not only for ensemble average-based homogenization on random heterogeneous media but also for deterministic cases.

IV conclusion

In conclusion, we have derived a formulation to find the dynamic effective constitutive relations for general heterogeneous media, including finite-size and non-periodic ones, by carefully taking into account boundary terms in the Hashin-Shtrikman principle. Our formulation relies on the infinite-body Green’s function of a reference homogeneous medium, making it free from the difficulty of determining Green’s function even for the homogenization of finite-size media. We note that as our theory generalizes the homogenization theory on two-phase disordered heterogeneous media [52, 53, 54], our formulation satisfies rapid convergence at least up to two-phase disordered heterogeneous media, with the decomposition of Γ(𝒙,𝒙ˇ)\Gamma(\bm{x},\check{\bm{x}}). It is envisioned that this study will enable more accurate predictions of the effective material properties of heterogeneous media and expand to more real-world applications.

Acknowledgements.
This research was supported by the Defense Advanced Research Projects Agency (Fund Number: W911NF2110363) and Alfred P. Sloan Foundation.

References

  • Smith and Kroll [2000] D. R. Smith and N. Kroll, Negative refractive index in left-handed materials, Physical review letters 85, 2933 (2000).
  • Pendry [2000] J. B. Pendry, Negative refraction makes a perfect lens, Physical review letters 85, 3966 (2000).
  • Schurig et al. [2006] D. Schurig, J. J. Mock, . B. Justice, S. A. Cummer, J. B. Pendry, A. F. Starr, and D. R. Smith, Metamaterial electromagnetic cloak at microwave frequencies, Science 314, 977 (2006).
  • Pendry et al. [2006] J. B. Pendry, D. Schurig, and D. R. Smith, Controlling electromagnetic fields, science 312, 1780 (2006).
  • Leonhardt and Tyc [2009] U. Leonhardt and T. Tyc, Broadband invisibility by non-euclidean cloaking, Science 323, 110 (2009).
  • Fang et al. [2006] N. Fang, D. Xi, J. Xu, M. Ambati, W. Srituravanich, C. Sun, and X. Zhang, Ultrasonic metamaterials with negative modulus, Nature materials 5, 452 (2006).
  • Ambati et al. [2007] M. Ambati, N. Fang, C. Sun, and X. Zhang, Surface resonant states and superlensing in acoustic metamaterials, Physical Review B 75, 195447 (2007).
  • Zhang et al. [2009] S. Zhang, L. Yin, and N. Fang, Focusing ultrasound with an acoustic metamaterial network, Physical review letters 102, 194301 (2009).
  • Zhang et al. [2011] S. Zhang, C. Xia, and N. Fang, Broadband acoustic cloak for ultrasound waves, Physical review letters 106, 024301 (2011).
  • Quan et al. [2015] L. Quan, F. Qian, X. Liu, X. Gong, and P. A. Johnson, Mimicking surface plasmons in acoustics at low frequency, Physical Review B 92, 104105 (2015).
  • Cummer et al. [2016] S. A. Cummer, J. Christensen, and A. Alù, Controlling sound with acoustic metamaterials, Nature Reviews Materials 1, 1 (2016).
  • Quan et al. [2018] L. Quan, Y. Ra’di, D. L. Sounas, and A. Alù, Maximum willis coupling in acoustic scatterers, Physical review letters 120, 254301 (2018).
  • Lee et al. [2022] J.-H. Lee, Z. Zhang, and G. X. Gu, Maximum electro-momentum coupling in piezoelectric metamaterial scatterers, Journal of Applied Physics 132, 125108 (2022).
  • Engheta and Ziolkowski [2006] N. Engheta and R. W. Ziolkowski, Metamaterials: physics and engineering explorations (John Wiley & Sons, 2006).
  • Liu et al. [2009] R. Liu, C. Ji, J. Mock, J. Chin, T. Cui, and D. Smith, Broadband ground-plane cloak, Science 323, 366 (2009).
  • Wilt et al. [2020] J. K. Wilt, C. Yang, and G. X. Gu, Accelerating auxetic metamaterial design with deep learning, Advanced Engineering Materials 22, 1901266 (2020).
  • Kadic et al. [2015] M. Kadic, T. Bückmann, R. Schittny, and M. Wegener, Experiments on cloaking in optics, thermodynamics and mechanics, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 373, 20140357 (2015).
  • Shaltout et al. [2019] A. M. Shaltout, V. M. Shalaev, and M. L. Brongersma, Spatiotemporal light control with active metasurfaces, Science 364, eaat3100 (2019).
  • Chen et al. [2020] C.-T. Chen, D. C. Chrzan, and G. X. Gu, Nano-topology optimization for materials design with atom-by-atom control, Nature communications 11, 3745 (2020).
  • Willis [1980a] J. Willis, Polarization approach to the scattering of elastic waves—i. scattering by a single inclusion, Journal of the Mechanics and Physics of Solids 28, 287 (1980a).
  • Willis [1980b] J. Willis, A polarization approach to the scattering of elastic waves—ii. multiple scattering from inclusions, Journal of the Mechanics and Physics of Solids 28, 307 (1980b).
  • Willis [1981a] J. R. Willis, Variational and related methods for the overall properties of composites, Advances in applied mechanics 21, 1 (1981a).
  • Willis [1981b] J. R. Willis, Variational principles for dynamic problems for inhomogeneous elastic media, Wave Motion 3, 1 (1981b).
  • Willis [1983] J. R. Willis, The overall elastic response of composite materials,   (1983).
  • Willis [1985] J. R. Willis, The nonlocal influence of density variations in a composite, International Journal of Solids and Structures 21, 805 (1985).
  • Willis [1997] J. R. Willis, Dynamics of composites, in Continuum Micromechanics, edited by P. Suquet (Springer Vienna, Vienna, 1997) pp. 265–290.
  • Willis [2011] J. R. Willis, Effective constitutive relations for waves in composites and metamaterials, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 467, 1865 (2011).
  • Willis [2012] J. R. Willis, The construction of effective relations for waves in a composite, Comptes Rendus Mécanique 340, 181 (2012).
  • Willis [2019] J. Willis, From statics of composites to acoustic metamaterials, Philosophical Transactions of the Royal Society A 377, 20190099 (2019).
  • Pernas-Salomón and Shmuel [2020] R. Pernas-Salomón and G. Shmuel, Symmetry breaking creates electro-momentum coupling in piezoelectric metamaterials, Journal of the Mechanics and Physics of Solids 134, 103770 (2020).
  • Muhafra et al. [2022] A. Muhafra, M. Kosta, D. Torrent, R. Pernas-Salomón, and G. Shmuel, Homogenization of piezoelectric planar willis materials undergoing antiplane shear, Wave Motion 108, 102833 (2022).
  • Zhang et al. [2022] Z. Zhang, J.-H. Lee, and G. X. Gu, Rational design of piezoelectric metamaterials with tailored electro-momentum coupling, Extreme Mechanics Letters 55, 101785 (2022).
  • Kosta et al. [2022] M. Kosta, A. Muhafra, R. Pernas-Salómon, G. Shmuel, and O. Amir, Maximizing the electromomentum coupling in piezoelectric laminates, International Journal of Solids and Structures 254, 111909 (2022).
  • Shuvalov et al. [2011] A. Shuvalov, A. Kutsenko, A. Norris, and O. Poncelet, Effective willis constitutive equations for periodically stratified anisotropic elastic media, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 467, 1749 (2011).
  • Pernas-Salomón et al. [2021] R. Pernas-Salomón, M. R. Haberman, A. N. Norris, and G. Shmuel, The electromomentum effect in piezoelectric willis scatterers, Wave Motion 106, 102797 (2021).
  • Nemat-Nasser and Srivastava [2011] S. Nemat-Nasser and A. Srivastava, Overall dynamic constitutive relations of layered elastic composites, Journal of the Mechanics and Physics of Solids 59, 1953 (2011).
  • Nemat-Nasser et al. [2011] S. Nemat-Nasser, J. R. Willis, A. Srivastava, and A. V. Amirkhizi, Homogenization of periodic elastic composites and locally resonant sonic materials, Physical Review B 83, 104103 (2011).
  • Srivastava [2015] A. Srivastava, Elastic metamaterials and dynamic homogenization: a review, International Journal of Smart and Nano Materials 6, 41 (2015).
  • Nassar et al. [2015] H. Nassar, Q.-C. He, and N. Auffray, Willis elastodynamic homogenization theory revisited for periodic media, Journal of the Mechanics and Physics of Solids 77, 158 (2015).
  • Nassar et al. [2016] H. Nassar, Q.-C. He, and N. Auffray, A generalized theory of elastodynamic homogenization for periodic media, International Journal of Solids and Structures 84, 139 (2016).
  • Milton and Willis [2007] G. W. Milton and J. R. Willis, On modifications of newton’s second law and linear continuum elastodynamics, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 463, 855 (2007).
  • Hashin and Shtrikman [1963] Z. Hashin and S. Shtrikman, A variational approach to the theory of the elastic behaviour of multiphase materials, Journal of the Mechanics and Physics of Solids 11, 127 (1963).
  • Norris [1994] A. N. Norris, Dynamic green’s functions in anisotropic piezoelectric, thermoelastic and poroelastic solids, Proceedings of the Royal Society of London. Series A: Mathematical and Physical Sciences 447, 175 (1994).
  • Wang and Zhang [2005] C.-Y. Wang and C. Zhang, 3-d and 2-d dynamic green’s functions and time-domain bies for piezoelectric solids, Engineering Analysis with Boundary Elements 29, 454 (2005).
  • Lewis et al. [2022] B. J. Lewis, E. N. Onder, and A. A. Prudil, Chapter 13 - integral equations, in Advanced Mathematics for Engineering Students, edited by B. J. Lewis, E. N. Onder, and A. A. Prudil (Butterworth-Heinemann, 2022) pp. 349–372.
  • Baker [1977] C. T. Baker, The numerical treatment of integral equations (Oxford University Press, 1977).
  • Sieck et al. [2017] C. F. Sieck, A. Alù, and M. R. Haberman, Origins of willis coupling and acoustic bianisotropy in acoustic metamaterials through source-driven homogenization, Physical Review B 96, 104303 (2017).
  • Duffy [2015] D. G. Duffy, Green’s functions with applications (Chapman and Hall/CRC, 2015).
  • Milton et al. [2006] G. W. Milton, M. Briane, and J. R. Willis, On cloaking for elasticity and physical equations with a transformation invariant form, New Journal of Physics 8, 248 (2006).
  • Milton [2007] G. W. Milton, New metamaterials with macroscopic behavior outside that of continuum elastodynamics, New Journal of Physics 9, 359 (2007).
  • Muhlestein et al. [2016] M. B. Muhlestein, C. F. Sieck, A. Alù, and M. R. Haberman, Reciprocity, passivity and causality in willis materials, Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 472, 20160604 (2016).
  • Torquato [1997a] S. Torquato, Effective stiffness tensor of composite media—i. exact series expansions, Journal of the Mechanics and Physics of Solids 45, 1421 (1997a).
  • Torquato [1997b] S. Torquato, Exact expression for the effective elastic tensor of disordered composites, Physical review letters 79, 681 (1997b).
  • Kim and Torquato [2020] J. Kim and S. Torquato, Effective elastic wave characteristics of composite media, New Journal of Physics 22, 123050 (2020).
Refer to caption
Figure 1: Schematics of the Hashin–Shtrikman principle-based approach to equivalently represent a heterogeneous medium (a) via a homogeneous reference medium (b), with the field polarization and its consequent source and boundary terms (e), which is decomposed into two homogeneous systems (c) and (d).
Refer to caption
Figure 2: (a) Schematics of the 1-D finite-size heterogeneous medium with the nn repeating unit cell. (b-e) Comparison of the kinetic fields. The black line is the ensemble average of analytical solutions of each ensemble. The blue line denotes macroscopic fields using our formulation whereas the red line is the ones not considering the coupling terms. The yellow line is a homogenization using the arithmetic mean of material properties. (f and g) Visualization of the apparent coupling coefficients S~1a\widetilde{S}_{1_{a}} and S~2a\widetilde{S}_{2_{a}}.
Refer to caption
Figure 3: The pure imaginary apparent coupling coefficient. The circles are coefficient values obtained from our proposed formulation while the solid lines are the benchmark data of [35]. The blue and yellow colors mean the regular and inverted elements, respectively.