betaFeedback

WorksPhysics and Astronomy

Discrete One-dimensional Models for the Electromomentum Coupling

Abstract

Willis dynamic homogenization theory revealed that the effective linear momentum of elastic composites is coupled to their effective strain. % This result, which is partially due to asymmetry at the subwavelength scale, implies that spatially-uniform but time-varying stress can induce temporal variations in the macroscopically observable momentum. Recent generalization of Willis' dynamic homogenization theory to the case of piezoelectric composites further revealed that their effective linear momentum is also coupled to the effective electric field. Here, we introduce the simplest possible model – a one-dimensional discrete model – that exhibits this so-called electromomentum coupling in subwavelength composites. We utilize our model to elucidate the physical origins of this phenomenon, illustrate its mechanism, and identify local resonances which lead to elevated Willis- and electromomentum coupling in narrow frequency bands. The results provide intuitive guidelines for the design of this coupling in piezoelectric metamaterials.

Introduction

Asymmetry has long been considered central in the emergence of unique physical behavior of multiscale systems [1]. In elastodynamics, this general notion is reinforced by the homogenization theory of Willis [2, 3, 4, 5, 6, 7, 8, 9]. One central discovery of his theory is that, in general, the effective constitutive relations for the linear momentum and stress fields are functions of both the strain and velocity fields. The Willis constitutive relations that arise from this dynamic homogenization procedure are nonlocal in space and time, namely, the response of a material point depends not only on the local fields at any instant in time, but also on neighboring points and their time history. The spatially local limit of the Willis equations, referred to as the Milton-Briane-Willis equations [10, 11], is applicable when the wavelength is much larger than the microstructure (see., e.g., the model that was first developed by Milton [12]), referred to as the metamaterial regime [13, 14], or when a single subwavelength element is analyzed1 [15, 16, 17, 18, 19].

The Willis couplings do not appear in the constitutive relations of the constituents, hence the resultant Willis materials are a type of metamaterial, whose behavior is fundamentally different from the behavior of their building blocks [20, 21, 22, 23, 24, 25, 26, 27, 28]. The surge of interest in metamaterials has resulted in a renewed interest in eponymous Willis materials, with numerous theoretical studies and experimental realizations [29, 30, 31, 32, 33, 34, 35, 36], including their application to elastic and acoustic wave control for, e.g., cloaking and sound manipulation [37, 38, 39, 40, 41, 42, 43, 44, 45]. One of the fundamental theoretical works on this topic by Sieck et al. [32] addressed the physical origins of the Willis coupling, and showed that the homogenized description must include this coupling in order to be physically meaningful. Muhlestein et al. [35] provided an experimental demonstration of this requirement, using the local response of a one-dimensional acoustic element. Milton et al. [10] observed that the Willis equations are analogous to the bianisotropic equations of electrodynamics, see also Refs. [32, 46].

Recently, Pernas-Salomón and Shmuel [47] generalized the theory of Willis to account for constituents that mechanically respond to non-mechanical stimuli, focusing on piezoelectric materials that respond to electric fields. The work demonstrated that the macroscopic linear momentum of piezoelectric composite is coupled to the electric field, and that electric displacement field is coupled to the velocity, a direct analogue to the Willis couplings. From a practical viewpoint, not only does the emergent electromomentum coupling constitute an additional degree of freedom to sense and generate elastic waves, it also opens up unique possibilities for the creation of tunable metamaterials using external electric fields. On a basic level, this occurs since the electromomentum effect, like the Willis effect, creates a direction-dependent phase angle [48], which can be used for wavefront shaping. Importantly, the phase angle that the electromomentum effect generates is tunable and can be turned on and off, by changing the electric circuit conditions. These advantages have motivated further studies of the electromomentum coupling, including alternative formulations [49, 50]; derivation of bounds [51, 52]; optimization [53, 54, 55]; and application to scattering control and cloaking [51, 56].

The mechanism behind the local component of Willis-, piezoelectric-, and electromomentum effects is similar: it is the breaking some spatial symmetry in the material properties [51, 52, 57, 58]. While continuum models provide some insight to the origins of the electromomentum effect, they yield complicated expressions for the effective coupling coefficients which do not lend themselves to an intuitive understanding of this new material response. It is therefore advantageous to develop simpler, more intuitive models to provide a better understanding of the origins of this coupling and, in turn, enable more efficient design and fabrication of these materials. Similar models have previously been introduced to illustrate, understand, and design Willis coupling in elastic and acoustic metamaterials [19, 59, 60], but have yet to be developed to analyze the electromomentum coupling. Accordingly, the objective of this work is to provide the simplest model that illustrates the mechanism of the electromomentum coupling and elucidates its physical origins.

To this end, we first introduce one-dimensional models for the Willis- and piezoelectric effects using systems of discrete masses, springs, and bound point charges. We provide in § 2 our model for the Willis effect, a model which is a generalization of the model that was introduced by Muhlestein et al. [19]. The system we consider consists of three point masses that are connected by two linear springs. We analyze cases where the masses and springs differ and provide expressions for the effective stiffness, mass density per unit length, and Willis coefficients. This lumped parameter model is then extended in § 3 to consider bound charge in order to capture the piezoelectric effect. Piezoelectricity emerges from a simple system consisting of two different masses of opposite charge that are connected by a linear spring, reminiscent of the model introduced in the classic monograph of Auld [61]. In contrast with the static analysis of Auld, we consider the inertia of the masses and hence also observe local resonances and Willis effects, which may be achieved using rationally designed metamaterials that display strong Willis- and piezoelectric couplings. The piezoelectric model serves as the building block in different assemblies with which we tailor effective material properties. Specifically, we show in § 4 that by combining two building blocks that differ by their piezoelectric coefficient, we obtain an assembly whose effective response exhibits both the Willis- and electromomentum effects. § 5 summarizes the observations from each section of the paper and discusses implications for the design of metamaterials displaying electromomentum coupling.

Discrete models for the Willis coupling

This section presents three models for Willis coupling that emerges from different types of element asymmetry. Inspired by elegantly simple model of Muhlestein et al. [19], who considered a linear spring linking two different point masses, our first model uses two identical linear springs that connect three different point masses. While the model presented here exhibits the same Willis coefficient as was first shown in Ref. [19], thanks to the middle mass it also exhibits a local resonance. Our second model consists of three identical masses that are connected by two different linear springs, a difference that also generates Willis coupling that differs from one emerging from mass asymmetry. The last model presented in this section considers the general case where both stiffness and mass are distributed asymmetrically. The results from our models are consistent with the insights from the continuum framework on Willis couplings, which show that it is a function of asymmetry in the mechanical impedance, i.e., of the mass density and elasticity. (Explicit relations between the lumped-parameters and the continuum parameters depend on proper evaluation of continuous field relationships in the long-wavelength limit2; while there are formal means to obtain such relationships, it is outside the scope of this work.)

Willis coupling by mass asymmetry

Consider three different masses, namely, m1m_1, m2m_2 and m3m_3, that are connected by linear springs of a stiffness kk, as illustrated in Figure 1a. The system is subjected to an axial force FF on both sides of the system; we denote by FLF_\mathrm{L} and FRF_\mathrm{R} the forces applied to the left and right mass, respectively, where FL≠FRF_\mathrm{L} \ne F_\mathrm{R} in general. We assume that the time dependence of the force is harmonic of the form e−iωte^{-\mathrm{i}\omega t}. We define the mass, stiffness, and force to be normalized per unit area, thus this system represents a low-order lumped parameter model of a layered system, such as those studied by Sieck et al. [32] and Pernas-Salomón et al. [48]. The resultant equations of motion of masses m1m_1, m2m_2 and m3m_3, respectively, are

Figure 1. Spring-mass models exhibiting Willis coupling owing to (a) mass asymmetry, (b) stiffness asymmetry, and (c) combined mass-stiffness asymmetry.

k(ul−um)−m1ω2ul=FL,−k(ul−2um+ur)−m2ω2um=0,k(ur−um)−m3ω2ur=−FR.(1)\begin{aligned} k(u_\mathrm{l}-u_\mathrm{m})-m_1\omega^2u_\mathrm{l} &= F_\mathrm{L},\\ -k(u_\mathrm{l}-2u_\mathrm{m}+u_\mathrm{r})-m_2\omega^2u_\mathrm{m} &= 0,\\ k(u_\mathrm{r}-u_\mathrm{m})-m_3\omega^2u_\mathrm{r} &= -F_\mathrm{R}. \tag*{(1)} \end{aligned}

here, ulu_\mathrm{l}, umu_\mathrm{m} and uru_\mathrm{r} are the displacements of the masses whose equilibrium position is at x0−Δxx_0-\Delta x, x0x_0, and x0+Δxx_0+\Delta x, respectively, where Δx\Delta x is the distance between two masses at equilibrium. We express the displacement of the middle mass as a function of the displacement of the exterior masses, such that

um=k2k−m2ω2(ul+ur)=uave[1−(ωωΣkm2)2]−1,(2)u_\mathrm{m}=\frac{k}{2k-m_2\omega^2}(u_\mathrm{l}+u_\mathrm{r})=u_\mathrm{ave}\left[1-\left(\frac{\omega}{\omega_{\Sigma km_2}}\right)^2\right]^{-1}, \tag*{(2)}

where uave:=(ul+ur)/2u_\mathrm{ave}:=(u_\mathrm{l}+u_\mathrm{r})/2 and ωΣkm22:=2Σk/m2\omega_{\Sigma km_2}^2:=2\Sigma k/m_2, i.e., the sum of the stiffness divided by the mass of the middle mass. The representation on the far right-hand side of Eq. (2) shows that um→uaveu_\mathrm{m}\to u_\mathrm{ave} for ω≪ωΣkm2\omega\ll\omega_{\Sigma km_2}, where ωΣkm2\omega_{\Sigma km_2} represents a localized resonance frequency for the motion of the central mass. Substituting this expression for umu_\mathrm{m} back into Eqs. (1) and (1) yields the following coupled equations of motion for m1m_1 and m3m_3

−m1ω2ul+kul−k2k−m2ω2(ul+ur)=FL,(3)-m_1\omega^2u_l+ku_l-\frac{k}{2k-m_2\omega^2}(u_l+u_r)=F_L, \tag*{(3)}
m3ω2ur−kur+k2k−m2ω2(ul+ur)=FR.(4)m_3\omega^2u_r-ku_r+\frac{k}{2k-m_2\omega^2}(u_l+u_r)=F_R. \tag*{(4)}

The expressions above can be rewritten in matrix form as

T[ulur]=[FLFR],where T=[−m1ω2+k−k22k−m2ω2−k22k−m2ω2k22k−m2ω2m3ω2−k+k22k−m2ω2].(5)\mathsf{T} \begin{bmatrix} u_l\\ u_r \end{bmatrix} = \begin{bmatrix} F_L\\ F_R \end{bmatrix}, \qquad\text{where }\mathsf{T}= \begin{bmatrix} -m_1\omega^2+k-\frac{k^2}{2k-m_2\omega^2} & -\frac{k^2}{2k-m_2\omega^2}\\ \frac{k^2}{2k-m_2\omega^2} & m_3\omega^2-k+\frac{k^2}{2k-m_2\omega^2} \end{bmatrix}. \tag*{(5)}

We now limit the analysis to subwavelength microstructures which restricts the results to the metamaterial limit [32] by expanding each function about the center of the model, i.e., at x=x0x=x_0, and neglecting terms in the expansion of higher order than O(Δx)\mathcal{O}(\Delta x). Using this long-wavelength approximation together with the definition of stress and the balance of linear momentum

−F(x0)=σ,−∂F∂x∣x0:=−F′(x0)=σ,x=p˙,(6)-F(x_0)=\sigma,\qquad-\left.\frac{\partial F}{\partial x}\right|_{x_0}:=-F'(x_0)=\sigma_{,x}=\dot{p}, \tag*{(6)}

yields the following coupled relationships for the stress and momentum given external forces and the strain and velocity as a function of the displacement of the left and right masses

\begin{bmatrix} \sigma\\ p \end{bmatrix} =\mathsf{M}^{(1)} \begin{bmatrix} F_L\\ F_R \end{bmatrix}, \qquad \mathsf{M}^{(1)}= \begin{bmatrix} -\frac{1}{2} & -\frac{1}{2}\\ -\frac{1}{2\mathrm{i}\omega\Delta x} & \frac{1}{2\mathrm{i}\omega\Delta x} \end{bmatrix}, \tag*{(7)}\end{bmatrix}
\begin{bmatrix} \varepsilon\\ v \end{bmatrix} =\mathsf{M}^{(2)} \begin{bmatrix} u_l\\ u_r \end{bmatrix}, \qquad \mathsf{M}^{(2)}= \begin{bmatrix} -\frac{1}{2\Delta x} & \frac{1}{2\Delta x}\\ -\frac{\mathrm{i}\omega}{2} & -\frac{\mathrm{i}\omega}{2} \end{bmatrix}. \tag*{(8)}\end{bmatrix}

where ε:=∂u/∂x=u′\varepsilon:= \partial u/\partial x = u' and v:=∂u/∂t=u˙v := \partial u/\partial t = \dot{u} are the strain and velocity at the center of the element, respectively. Substituting these relations into Eq. (5) yields

\begin{bmatrix} \sigma\\ p \end{bmatrix} = \mathrm{M}^{(1)\mathrm{T}}\mathrm{M}^{(2)-1} \begin{bmatrix} \varepsilon\\ v \end{bmatrix} := \begin{bmatrix} \tilde{C} & \tilde{S} \\ \tilde{S}^{\dagger} & \tilde{\rho} \end{bmatrix} \begin{bmatrix} \varepsilon\\ v \end{bmatrix}, \tag*{(9)}\end{bmatrix}

where M(1)TM(2)−1\mathrm{M}^{(1)\mathrm{T}}\mathrm{M}^{(2)-1} can be cast as a matrix whose entries represent the effective material properties of the subwavelength element, which can be written as

ρ~=2k(m1+m2+m3)−m2ω2(m1+m3)2Δx(2k−m2ω2),C~=kΔx−m1+m32ω2Δx,S~=S~†=−iωm3−m12.(10)\begin{aligned} \tilde{\rho} &= \frac{2k(m_1+m_2+m_3)-m_2\omega^2(m_1+m_3)}{2\Delta x(2k-m_2\omega^2)},\\ \tilde{C} &= k\Delta x-\frac{m_1+m_3}{2}\omega^2\Delta x,\\ \tilde{S} &= \tilde{S}^{\dagger}=-i\omega\frac{m_3-m_1}{2}. \tag*{(10)} \end{aligned}

We define the length of the element as Luc=2ΔxL_{\mathrm{uc}}=2\Delta x and the following frequencies of local resonance: ωΣm2:=2kΣm/[m2(m1+m3)]\omega_{\Sigma m}^2:=2k\Sigma m/[m_2(m_1+m_3)] and ω132:=2k/(m1+m3)\omega_{13}^2:=2k/(m_1+m_3) in addition to ωΣkm22\omega_{\Sigma km_2}^2 used in Eq. (2), where Σm:=m1+m2+m3\Sigma m:=m_1+m_2+m_3. Note that ωΣm2\omega_{\Sigma m}^2 and ωΣkm22\omega_{\Sigma km_2}^2 are related via ωΣm2=ωΣkm22Σm/(m1+m3)\omega_{\Sigma m}^2=\omega_{\Sigma km_2}^2\Sigma m/(m_1+m_3), indicating that ωΣm>ωΣkm2\omega_{\Sigma m}>\omega_{\Sigma km_2} for all values of mim_i. In terms of these resonance frequencies, the effective properties can be written as

ρ~=(ΣmLuc)1−(ω/ωΣm)21−(ω/ωΣkm2)2,C~=k2Luc[1−(ω/ω13)2],S~=S~†=−iωm3−m12=−iωΔm2,(11)\begin{aligned} \tilde{\rho} &= \left(\frac{\Sigma m}{L_{\mathrm{uc}}}\right)\frac{1-(\omega/\omega_{\Sigma m})^2}{1-(\omega/\omega_{\Sigma km_2})^2},\\ \tilde{C} &= \frac{k}{2}L_{\mathrm{uc}}\left[1-(\omega/\omega_{13})^2\right],\\ \tilde{S} &= \tilde{S}^{\dagger}=-i\omega\frac{m_3-m_1}{2}=-i\omega\frac{\Delta m}{2}, \tag*{(11)} \end{aligned}

where Δm:=m3−m1\Delta m:=m_3-m_1 represents the mass asymmetry of the element.

A few important points are observed regarding this simple model. First, we note that the frequency dependence of the dynamic effective density includes a local resonance leading to negative density for frequencies near ω=ωΣkm2\omega=\omega_{\Sigma km_2}, which is when the internal mass m2m_2 oscillates out-of-phase with m1m_1 and m3m_3. Similarly, a negative effective stiffness is observed for ω>ω13\omega>\omega_{13}, where ω13\omega_{13} represents a “breathing” mode where m1m_1 and m3m_3 move out-of-phase with each other. Further, the Willis coefficients, S~\tilde{S} and S~†\tilde{S}^{\dagger} are

  • (i) equal one to another, as they should in order to satisfy reciprocity [19, 33, 57];

  • (ii) linear in ω\omega in the low-frequency limit assumed here;

  • (iii) of the same form as that of the mechanical impedance of a lumped mass that is equal to the difference between m1m_1 and m3m_3, hence identically zero when the element has a symmetric distribution of mass, i.e., when m1=m3m_1=m_3. Furthermore, if the element is inverted such that left mass equals m3m_3 and the right mass equals m1m_1, or equivalently, the coordinate system is inverted, S~\tilde{S} and S~†\tilde{S}^{\dagger} change their sign. This is in agreement with the fact that the Willis coupling models a direction-dependent

material response related to a direction-dependent characteristic impedance, a useful property for wavefront manipulation [35, 43, 48]. Finally, representation (11) clearly shows that the static limit yields the expected static benchmarks

ρ~∣ω→0=m1+m2+m32Δx=ΣmLuc,C~∣ω→0=kΔx=12kLuc,S~∣ω→0=S~†∣ω→0=0,(12)\widetilde{\rho}\big|_{\omega\to0} = \frac{m_1+m_2+m_3}{2\Delta x} = \frac{\Sigma m}{L_{\mathrm{uc}}}, \qquad\widetilde{C}\big|_{\omega\to0} = k\Delta x = \frac{1}{2}kL_{\mathrm{uc}}, \qquad\widetilde{S}\big|_{\omega\to0} = \widetilde{S}^{\dagger}\big|_{\omega\to0} = 0, \tag*{(12)}

such that the effective mass density is the arithmetic mean of the density of the constituents, and the effective stiffness is the harmonic mean of the stiffnesses of the two springs of the element.

Willis coupling by stiffness asymmetry

We now modify the element to determine effective properties that result from asymmetry in the stiffness. Accordingly, we now assume that all masses are all equal to mm, while the right and left springs are represented with stiffnesses k1k_1 and k2k_2, respectively, as shown in Figure 1b. The corresponding equations of motion of the external masses and the middle mass, respectively, are

−mω2ul+k1(ul−um)=FL,(13)-m\omega^2u_l + k_1(u_l-u_m)=F_L, \tag*{(13)}
mω2ur−k2(ur−um)=FR,(14)m\omega^2u_r-k_2(u_r-u_m)=F_R, \tag*{(14)}
−mω2um=k1(ul−um)+k2(ur−um),(15)-m\omega^2u_m=k_1(u_l-u_m)+k_2(u_r-u_m), \tag*{(15)}

from which we obtain

um=[ul(k1Σk)+ur(k2Σk)][1−(ωωΣkm2)2]−1,(16)u_m=\left[u_l\left(\frac{k_1}{\Sigma k}\right)+u_r\left(\frac{k_2}{\Sigma k}\right)\right]\left[1-\left(\frac{\omega}{\omega_{\Sigma k m_2}}\right)^2\right]^{-1}, \tag*{(16)}

where we have defined Σk:=k1+k2\Sigma k := k_1+k_2 and the local resonance ωΣkm2\omega_{\Sigma k m_2} that was defined in the previous section, now for m2=mm_2=m. This expression shows that the displacement of the central mass is a weighted average of the displacements of ulu_l and uru_r based on stiffness ratios. Substituting this expression for ωΣkm2\omega_{\Sigma k m_2} into Eqs. (13a) and (14) yields

T[ulur]=[FLFR],T=[−mω2+k1−k12k1+k2−mω2−[1−(ωωΣkm2)2]−1[1−(ωωΣkm2)2]−1−mω2(mω2−2k2)+(k2−mω2)k1k1+k2−mω2].(17)\mathrm{T}\begin{bmatrix}u_l\\u_r\end{bmatrix}=\begin{bmatrix}F_L\\F_R\end{bmatrix}, \qquad\mathrm{T}=\begin{bmatrix} -m\omega^2+k_1-\dfrac{k_1^2}{k_1+k_2-m\omega^2} & -\left[1-\left(\dfrac{\omega}{\omega_{\Sigma k m_2}}\right)^2\right]^{-1}\\ \left[1-\left(\dfrac{\omega}{\omega_{\Sigma k m_2}}\right)^2\right]^{-1} & -\dfrac{m\omega^2(m\omega^2-2k_2)+(k_2-m\omega^2)k_1}{k_1+k_2-m\omega^2} \end{bmatrix}. \tag*{(17)}

Following the same procedure as in the previous section and keeping frequency terms up to order O(ω2)\mathcal{O}\left(\omega^2\right), we obtain constitutive equations in the form of Eq. (9), with

ρ~=3Σk−2mω2Luc(Σk−mω2)m,C~=Luc(4k1k2−3mω2Σk)4(Σk−mω2),S~=S~†=iωmΣk−mω2k1−k22,(18)\begin{aligned} \tilde{\rho} &= \frac{3\Sigma k-2m\omega^2}{L_{\mathrm{uc}}\left(\Sigma k-m\omega^2\right)}m,\\ \tilde{C} &= \frac{L_{\mathrm{uc}}\left(4k_1k_2-3m\omega^2\Sigma k\right)}{4\left(\Sigma k-m\omega^2\right)},\\ \tilde{S} &= \tilde{S}^{\dagger}=\frac{i\omega m}{\Sigma k-m\omega^2}\frac{k_1-k_2}{2}, \tag*{(18)} \end{aligned}

where we again note that the Willis coefficients are identically equal to another and are only nonzero when k1≠k2k_1\ne k_2. As such, Willis coupling vanishes when the element is symmetric, which is analogous to the fact that Willis coupling vanishes in the absence of mass asymmetry in the previous case. In terms of the local resonance frequencies ω3Σk2:=3Σk/(2m)\omega_{3\Sigma k}^{2}:=3\Sigma k/(2m), ωα2:=4k1/(3m)\omega_{\alpha}^{2}:=4k_1/(3m) and ωβ2:=4k2/(3m)\omega_{\beta}^{2}:=4k_2/(3m), we may write the frequency-dependent effective properties as

ρ~=(3mLuc)1−(ω/ω3Σk)21−(ω/ωΣkm2)2,C~=Luc(k1k2Σk)1−(ω/ωα)2−(ω/ωβ)21−(ω/ωΣkm2)2,S~=S~†=12(ΔkΣk)iωm1−(ω/ωΣkm2)2,(19)\begin{aligned} \tilde{\rho} &= \left(\frac{3m}{L_{\mathrm{uc}}}\right)\frac{1-\left(\omega/\omega_{3\Sigma k}\right)^2}{1-\left(\omega/\omega_{\Sigma k m2}\right)^2},\\ \tilde{C} &= L_{\mathrm{uc}}\left(\frac{k_1k_2}{\Sigma k}\right)\frac{1-\left(\omega/\omega_{\alpha}\right)^2-\left(\omega/\omega_{\beta}\right)^2}{1-\left(\omega/\omega_{\Sigma k m2}\right)^2},\\ \tilde{S} &= \tilde{S}^{\dagger}=\frac{1}{2}\left(\frac{\Delta k}{\Sigma k}\right)\frac{i\omega m}{1-\left(\omega/\omega_{\Sigma k m2}\right)^2}, \tag*{(19)} \end{aligned}

where Δk:=k1−k2\Delta k:=k_1-k_2. Note that all effective properties are frequency dependent in the metamaterial limit, if the localized resonances occur where the wavelength is much longer than the element. When 0<ω≪ωΣkm20<\omega\ll\omega_{\Sigma k m2}, the Willis coupling is linear in ω\omega, and weighted by the ratio of the difference in stiffness and the sum of the stiffnesses. We also observe that an inverted element for which k2k_2 is the left spring and k1k_1 is the right spring, or equivalently, if the coordinate system is inverted, then S~\tilde{S} and S~†\tilde{S}^{\dagger} change their sign, as one would anticipate from previous research [32]. Notably, the Willis coefficient exhibits a localized resonance, ωΣkm2\omega_{\Sigma k m2}, unlike the Willis coefficient that emerges from mass asymmetry. We may therefore expect that Willis coupling can be very large in a narrow band of frequencies around ωΣkm2\omega_{\Sigma k m2} when elements have an asymmetric stiffness distribution. This representation also clearly shows that the static limit yields the following expressions properties for density, modulus, and Willis coefficients

ρ~∣ω→0=3mLuc,C~∣ω→0=k1k2ΣkLuc,S~∣ω→0=S~†∣ω→0=0.(20)\tilde{\rho}\big|_{\omega\to0}=\frac{3m}{L_{\mathrm{uc}}},\qquad \tilde{C}\big|_{\omega\to0}=\frac{k_1k_2}{\Sigma k}L_{\mathrm{uc}},\qquad \tilde{S}\big|_{\omega\to0}=\tilde{S}^{\dagger}\big|_{\omega\to0}=0. \tag*{(20)}

Thus, the effective mass density is the arithmetic mean of the density of the constituents, the effective stiffness is the harmonic mean of the stiffnesses of the two springs of the element, and the Willis coefficient is null.

Willis coupling by combined stiffness and mass asymmetry

Finally, we consider a model that exhibits both mass and stiffness asymmetry as illustrated in Figure 1c, and repeat the same procedure as before to extract the effective properties. We find Willis couplings that are the sum of the Willis couplings found when only one property is asymmetric, namely,

S~=S~†=−iω2[Δm+(ΔkΣk)m21−(ω/ωΣkm2)2]=S~k+S~m,(21)\widetilde{S}=\widetilde{S}^{\dagger}=-\frac{\mathrm{i}\omega}{2}\left[\Delta m+\left(\frac{\Delta k}{\Sigma k}\right)\frac{m_{2}}{1-\left(\omega/\omega_{\Sigma km2}\right)^{2}}\right]=\widetilde{S}_{k}+\widetilde{S}_{m}, \tag*{(21)}

where S~k\widetilde{S}_{k} and S~m\widetilde{S}_{m} are the Willis coefficients in the asymmetric stiffness- and asymmetric mass models, respectively. The remaining effective properties are

ρ~=(ΣmLuc)1−(ω/ωk13)21−(ω/ωΣkm2)2,\widetilde{\rho}=\left(\frac{\Sigma m}{L_{\mathrm{uc}}}\right)\frac{1-\left(\omega/\omega_{k13}\right)^{2}}{1-\left(\omega/\omega_{\Sigma km2}\right)^{2}},
C~=Luc(k1k2Σk)1−(ω/ωxΣm)2−(ω/ωyΣm)21−(ω/ωΣkm2)2,(22)\widetilde{C}=L_{\mathrm{uc}}\left(\frac{k_{1}k_{2}}{\Sigma k}\right)\frac{1-\left(\omega/\omega_{x\Sigma m}\right)^{2}-\left(\omega/\omega_{y\Sigma m}\right)^{2}}{1-\left(\omega/\omega_{\Sigma km2}\right)^{2}}, \tag*{(22)}

where we have defined the following localized resonance frequencies: ωk132:=Σk/[m2(m1+m3)]\omega_{k13}^{2}:=\Sigma k/[m_{2}(m_{1}+m_{3})], ωxΣm2:=4k1/Σm\omega_{x\Sigma m}^{2}:=4k_{1}/\Sigma m and ωyΣm2:=4k2/Σm\omega_{y\Sigma m}^{2}:=4k_{2}/\Sigma m. These effective properties recover the static benchmarks for ω→0\omega\to0, as in the cases of only asymmetric mass or stiffness. These results are consistent with previous work in the literature [32, 35, 48]. However, due to the simplicity of the lumped parameter model considered here, these results provide insights into how local asymmetry leads to local Willis coupling. Furthermore, the relations they provide, which include localized resonances, may be useful in designing acoustic metamaterials that have yet to be investigated.

Comparison with periodic lumped-parameter Willis model

As discussed by Simovski [62], while effective properties that are retrieved from analysis of finite media differ from the effective properties of bulk media, there is a qualitative relation between them, and often the former provide useful approximations and insights. It is thus advantageous to compare our finite lumped Willis model with a periodic one. Such model was analyzed in the excellent paper of Nassar et al. [30], there the authors revisited Willis dynamic homogenization method and applied it to a periodic repetition of two different masses and two springs of different stiffnesses. Their model yields spatio-temporal nonlocal effective properties, being functions of both the frequency and the Bloch wavenumber.

We list the following similarities between the long-wavelength, low frequency limit of the Willis coupling in Nassar et al. [30], and our model:

  1. It is a linear function of frequency,

  2. It is purely imaginary,

(iii) It is a function of the mass asymmetry and stiffness-asymmetry.

Notably, the above features are also consistent with the local limit of continuum models for the Willis coupling [32]. We illustrate these similarities in Figure 2, where we numerically evaluate the normalized Willis coupling S^(ω)=iS~/ω\hat{S}(\omega)=i\tilde{S}/\omega, of the periodic- and finite models, for a representative example. Specifically, local Willis coupling of the periodic model by Nassar et al. [30] (dashed red), is evaluated with k1=1,k2=4,m1=1,m2=3k_1=1,k_2=4,m_1=1,m_2=3, and our finite 3-mass model (solid blue) is evaluated by setting m1=m3=1/2m_1=m_3=1/2, in order to be equivalent to a unit-cell of Nasser et al. model. In addition to the shared features of the two models that are listed above, we also see that both models predict local resonance, and follow a very similar trend below that resonance frequency.

Figure 2. Normalized Willis coupling S^=iS~/ω\hat{S}=i\tilde{S}/\omega as function of frequency, extracted from the spatially local limit of the periodic model by Nassar et al. [30] (dashed red), for the representative set k1=1,k2=4,m1=1,m2=3k_1=1,k_2=4,m_1=1,m_2=3, and our finite 3-mass model (solid blue), whose periodic repetition reproduces the model of Nassar et al. [30] by setting m1=m3=1/2m_1=m_3=1/2.

As mentioned, while finite models provide useful insights, they nevertheless differ from periodic models. A prominent example here is when only one of the mechanical parameters in the periodic model of Nassar et al. [30] are nonuniform, in which case the model exhibits mirror symmetry and the Willis coupling vanishes. By contrast, it is clear that the mirror symmetry of the finite model is broken when either one of the two parameters is nonuniform, and indeed its Willis coupling is linear in each one of the mechanical asymmetries separately.

Before we proceed to develop finite lumped-models for the piezoelectric- and electromomentum effects, we note that there are additional similarities between the effective properties of the two Willis models, e.g., both models recover the static limit of the effective mass (arithmetic mean) and effective stiffness (harmonic mean) as one would anticipate for a physically meaningful model.

Collectively, the above features reinforce our approach for extracting effective properties using finite lumped-parameter models. Finally, we note the finite models that are to be derived for the piezoelectric- (§ 3) and electromomentum effects (§ 4), can also be extended to periodic models. In short, such extension be derived by periodically connecting the elements in § 3-§ 4, and analyzing one element as unit cell of a periodic medium. As such, there are additional constraints on the connection between the forces and displacements of the ends of the unit cell. The approach to extract the effective properties in the long-wavelength limit remains similar to the approach described in this work, i.e., obtain the equations of motion, express them in terms of the physical fields at the ends of the cell, assume that the physical fields can be represented as smooth functions, expand these functions about the center of the element in order to relate the kinematic and kinetic fields, and so on. Bearing this in mind, we recall that the purpose of this work is to provide the simplest model that captures and elucidates the electromomentum effect and its origins at the sub-wavelength scale, a purpose that is achieved without analyzing the more complicated periodic model.

Discrete models of piezoelectric coupling

Our next step is to incorporate the piezoelectric effect into the model by extending an approach provided by Auld [61]. We first consider two masses m1m_1 and m2m_2, connected by a linear spring of stiffness kk, carrying bound electric charges of magnitude RqRq and −Rq-Rq, respectively, and separated by Δx\Delta x, as illustrated in Figure 3a. We assume that the distance between the masses is large enough to neglect the electrical force created by the interaction between the masses, i.e., Coulomb’s force, in comparison to the mechanical force, or Fc=ke(Rq)2(Δx)2≪∣k(ur−ul)∣F_c = k_e \frac{(Rq)^2}{(\Delta x)^2} \ll\lvert k(u_r-u_l)\rvert, where kek_e is the Coulomb constant. For the case under consideration, we must take the interaction of bound charge with an externally applied electric field, denoted here by EE, in addition to the application of an axial pressure at the outer faces of the model as described in the previous section. The resultant equations of motion are

Figure 3. Schematics of mass, spring, and point charges that demonstrate piezoelectric and Willis coupling. (a) Fundamental building block that exhibits the piezoelectric coupling. (b) A pair of building blocks from (a), attached in mirror symmetry such that there is no effective piezoelectric-or Willis coupling. (c) Assembly of these two building blocks from (a), attached with the same orientation. Since the masses and charges are distributed asymmetrically, there are effective piezoelectric and Willis couplings, however the electromomentum coupling is null since both building blocks exhibit the same piezoelectric coefficient.

k(ul−ur)−m1ω2ul−ERq=FL,(23)k(u_l-u_r)-m_1\omega^2u_l-ERq=F_L, \tag*{(23)}
−k(ur−ul)+m2ω2ur−ERq=FR,(24)-k(u_r-u_l)+m_2\omega^2u_r-ERq=F_R, \tag*{(24)}

which can be written in matrix form as

T[ulur]+h=[FLFR],T=[−m1ω2+k−kkm2ω2−k],h=[−Rq−Rq]E.(25)\mathbf{T} \begin{bmatrix} u_l\\ u_r \end{bmatrix} +\mathbf{h} = \begin{bmatrix} F_L\\ F_R \end{bmatrix}, \qquad \mathbf{T} = \begin{bmatrix} -m_1\omega^2+k & -k\\ k & m_2\omega^2-k \end{bmatrix}, \qquad \mathbf{h} = \begin{bmatrix} -Rq\\ -Rq \end{bmatrix}E. \tag*{(25)}

We again expand each function about the center of the model at x=x0x=x_0, and neglect terms of higher order than O(Δx)\mathcal{O}(\Delta x). Together with Eq. (6), this yields the following expressions for the local stress, momentum, strain, and velocity

[σp]=M(1)[FLFR],M(1)=[−12−12−1iωΔx1iωΔx](26)\begin{bmatrix} \sigma\\ p \end{bmatrix} = \mathrm{M}^{(1)} \begin{bmatrix} F_L \\ F_R \end{bmatrix}, \qquad \mathrm{M}^{(1)} = \begin{bmatrix} -\frac{1}{2} & -\frac{1}{2} \\ -\frac{1}{\mathrm{i}\omega\Delta x} & \frac{1}{\mathrm{i}\omega\Delta x} \tag*{(26)} \end{bmatrix}
[εv]=M(2)[ulur],M(2)=[−1Δx1Δx−iω2−iω2](27)\begin{bmatrix} \varepsilon\\ v \end{bmatrix} = \mathrm{M}^{(2)} \begin{bmatrix} u_l \\ u_r \end{bmatrix}, \qquad \mathrm{M}^{(2)} = \begin{bmatrix} -\frac{1}{\Delta x} & \frac{1}{\Delta x} \\ -\frac{\mathrm{i}\omega}{2} & -\frac{\mathrm{i}\omega}{2} \tag*{(27)} \end{bmatrix}

Substituting these relations into Eq. (25) yields

\begin{bmatrix} \sigma\\ p \end{bmatrix} = \mathrm{M}^{(1)}\mathrm{T}\mathrm{M}^{(2)-1} \begin{bmatrix} \varepsilon\\ v \end{bmatrix} + \mathrm{M}^{(1)}\mathbf{h} = \begin{bmatrix} \widetilde{C} & \widetilde{S} \\ \widetilde{S}^{\dagger} & \widetilde{\rho} \end{bmatrix} \begin{bmatrix} \varepsilon\\ v \end{bmatrix} - \begin{bmatrix} \widetilde{B}^{\dagger} \\ 0 \end{bmatrix}E \tag*{(28)}\end{bmatrix}

where the effective properties are given by

ρ~=ΣmLuc,C~=kLuc[1−(ωω12)2],S~=S~†=−iωΔm2,B~†=−Rq;(29)\widetilde{\rho} = \frac{\Sigma m}{L_{\mathrm{uc}}}, \qquad \widetilde{C} = kL_{\mathrm{uc}}\left[1-\left(\frac{\omega}{\omega_{12}}\right)^2\right], \qquad \widetilde{S}=\widetilde{S}^{\dagger}=-\mathrm{i}\omega\frac{\Delta m}{2}, \qquad \widetilde{B}^{\dagger}=-Rq; \tag*{(29)}

here ω122:=4k/(Σm)\omega_{12}^{2}:=4k/(\Sigma m), and for this element Luc=ΔxL_{\mathrm{uc}}=\Delta x. The observations made in § 2 regarding the static limits of the effective density and stiffness also apply to this model, and we also observe that S~\widetilde{S}, which emerges from the asymmetric mass distribution and has the same linear dependence on frequency. In addition, we observe that the presence of asymmetric bound charge subjected to an external electric field produces an additional stress. We quantify this electric field-induced stress as σE=B~†E\sigma_E = \tilde{B}^{\dagger}E, where B~†\tilde{B}^{\dagger} is the piezoelectric coefficient when the relations are cast in the stress-charge form3. Since B~†\tilde{B}^{\dagger} is a property that depends on the orientation of the element, its value is flipped when the coordinate system is inverted, or when the element is flipped, such that the negative charge resides in the left mass and the positive charge resides in the right mass. We note that this directionality of the electric polarization is analogous to the directionality of the Willis coefficient in this element and in § 2. This analogy is the reason that mass and stiffness asymmetry of purely elastic or acoustic metamaterials can be referred to as having a Willis polarization.

To fully characterize the electromechanical medium in the electrostatic limit, we should also consider the dielectric response of the system. To this end, we first calculate the change in the electric dipole moment due to an externally applied strain. The electric dipole moment pDp_D is defined as pD=∑qixip_D = \sum q_i x_i, where qiq_i and xix_i represent the charge and position of the ithi^{\mathrm{th}} element and ii is summed over all charges. When the system shown in Figure 3a is subjected to displacements on the left- and right-hand sides, the resultant change in electric dipole moment, ΔpD\Delta p_D, is given by

ΔpD=Rqul−Rqur.(30)\Delta p_D = Rq u_l - Rq u_r. \tag*{(30)}

This simple expression can be rewritten as

ΔpD=[Rq−Rq][ulur]=[Rq−Rq]M(2)−1[εν]=−(Rq)Lucε.(31)\Delta p_D = \begin{bmatrix} Rq & -Rq \end{bmatrix} \begin{bmatrix} u_l \\ u_r \end{bmatrix} = \begin{bmatrix} Rq & -Rq \end{bmatrix} \mathrm{M}^{(2)-1} \begin{bmatrix} \varepsilon\\ \nu\end{bmatrix} = -(Rq)L_{\mathrm{uc}}\varepsilon. \tag*{(31)}

The polarization density induced by the mechanical deformation, PP, defined as the electric dipole moment per unit volume, thus equals

P=ΔpDLuc=−(Rq)ε.(32)P = \frac{\Delta p_D}{L_{\mathrm{uc}}} = -(Rq)\varepsilon. \tag*{(32)}

Finally, the electric displacement field in the element, DD, is the sum of this strain-induced change in electric polarization and the electric displacement field in vacuum, i.e.,

D=ϵ0E+P=ϵ0E−(Rq)ε,(33)D = \epsilon_0 E + P = \epsilon_0 E - (Rq)\varepsilon, \tag*{(33)}

where ϵ0\epsilon_0 is the permittivity of free space. Eq. (33) allows us to identify the dielectric constant A~\tilde{A} with ϵ0\epsilon_0 and the piezoelectric coefficient B~\tilde{B} with −Rq-Rq, which is equal to B~†\tilde{B}^{\dagger}, as it should to satisfy reciprocity [57]. Note that there is no relative permittivity in this case, i.e., A~=ϵ0\tilde{A} = \epsilon_0, since our model only considers two discrete masses in free space. In the more general, the presence of a third charged mass in-between the two point masses would potentially change the electric displacement-electric field relationship, as we show in the sequel.

In summary, Eqs. (28) and (33) together yield the following general constitutive relationships

[σDp]=[kLuc[1−(ωω12)2]−Rq−iωΔm2−Rq−ϵ00−iωΔm20ΣmLuc][ε−Ev]=[C~B~†S~B~−A~0S~†0ρ~][ε−Ev].(34)\begin{aligned} \begin{bmatrix} \sigma\\ D \\ p \end{bmatrix} = \begin{bmatrix} kL_{\mathrm{uc}}\left[1-\left(\frac{\omega}{\omega_{12}}\right)^2\right] & -Rq & -i\omega\frac{\Delta m}{2} \\ -Rq & -\epsilon_0 & 0 \\ -i\omega\frac{\Delta m}{2} & 0 & \frac{\Sigma_m}{L_{\mathrm{uc}}} \end{bmatrix} \begin{bmatrix} \varepsilon\\ -E \\ v \end{bmatrix} \\ &= \begin{bmatrix} \widetilde{C} & \widetilde{B}^{\dagger} & \widetilde{S} \\ \widetilde{B} & -\widetilde{A} & 0 \\ \widetilde{S}^{\dagger} & 0 & \widetilde{\rho} \end{bmatrix} \begin{bmatrix} \varepsilon\\ -E \\ v \end{bmatrix}. \tag*{(34)} \end{aligned}

Notably, the coupling between electrical and mechanical physics and the mechanical potential and kinetic energies is evident by the emergent non-zero off-diagonal parameters representing piezoelectricity, B~†=B~\widetilde{B}^{\dagger}=\widetilde{B}, and Willis coupling, S~†=S~\widetilde{S}^{\dagger}=\widetilde{S}, respectively. Further note that the momentum (electric displacement) and electric field and the electric displacement (velocity) are not coupled, since electromomentum coupling require asymmetry in the piezoelectric profile itself, as we discuss in more detail in the sequel. Finally, we observe that the effective properties converge to the static benchmarks mentioned in previous sections.

We can now use this model, which demonstrates both Willis and piezoelectric effects, as the building block from which we can tailor different emergent metamaterial properties. For example, by calculating the effective properties of an assembly made of a building block and its mirror inversion, we find that the effective piezoelectric- and Willis coefficients of the system are null, as one would anticipate, since the assembly has mirror symmetry and the net polarization is null. This composition is illustrated in Figure 3b, where such a symmetric charge distribution (without masses) was considered by Auld [61] to model a non-piezoelectric response (see figure 8.1 therein). We also note that when setting R=1R=1, the model of Auld [61] for a piezoelectric solid (figure 8.2 therein) is an assembly of two of our building blocks with the same orientation, as illustrated in Figure 3c. Note that Auld analyzed only the static response of the model, hence the effective properties presented there [61] do not include the dynamic phenomena reported here.

In view of Refs. [47, 48, 49, 54], which conclude that the electromomentum effect appears in composites with asymmetric piezoelectric profile4, we expect to observe the electromomentum effect when assembling two building blocks with different piezoelectric coefficients. The analysis of such assembly is the subject of the following section.

Discrete model of electromomentum coupling

We now assemble two building blocks, the first of which is defined by m1=2m2=mm_1=2m_2=m and R=R1R=R_1, and the second by m2=2m1=mm_2=2m_1=m and R=R2R=R_2, as shown in Figure 4a and Figure 4b, respectively. Using the relations provided in Eq. (34), we obtain the following effective properties for the building block shown in Figure 4a

Figure 4. Schematics of mass, spring, and point charges that produce piezoelectric, Willis, and electromomentum coupling. The building blocks in (a) and (b) exhibit the piezoelectric- and Willis effects. The building block in (b) has inverse polarity in mass with respect to block (a), and therefore has the opposite Willis coefficient of block (a). It also displays a piezoelectric coefficient with a different magnitude from that of block (a) because its point charge magnitude is different. (c) Assembly of these building blocks, which exhibits the electromomentum effect. (d) Assembly of two building blocks with two different stiffnesses and masses. Assemblies (c) and (d) exhibit both the electromomentum- and Willis couplings.

ρ~1=3m2Δx,C~1=(k+3mω28)Δx,A~1=ϵ0,S~1=−iωm4,B~1=R1q.(35)\begin{aligned} \tilde{\rho}_1 &= \frac{3m}{2\Delta x}, \qquad\tilde{C}_1 = \left(k+\frac{3m\omega^2}{8}\right)\Delta x, \qquad\tilde{A}_1 = \epsilon_0,\\ \tilde{S}_1 &= -\frac{i\omega m}{4}, \qquad\tilde{B}_1 = R_1q. \tag*{(35)} \end{aligned}

The properties of the building block shown Figure 4b are the same, except for (i) the Willis coefficient, whose sign opposite to the sign provided in Eq. (35), since the masses in the second building block are flipped with respect to those in the first building block; (ii) the piezoelectric coefficient, which is related to the coefficient of the first building block, B~1\tilde{B}_1, via the ratio of their charges, i.e., B~2=B~1R2/R1\tilde{B}_2=\tilde{B}_1R_2/R_1. The assembly of these two building blocks as indicated in Figure 4c yields a 3-mass model, since there is no spacing between the m/2m/2 masses and they can therefore be considered as a single mass mm. Accordingly, the left-, center-, and right-masses in the assembly are charged with R1qR_1q, −R1q+R2q-R_1q+R_2q and −R2q-R_2q, respectively, as shown in Figure 4c. Under the application of electric field EE and axial force FF on the outer masses, the equations of motion are

−mω2ul+k(ul−um)−ER1q=FL,(36)-m\omega^2u_l+k(u_l-u_m)-ER_1q=F_L, \tag*{(36)}
−mω2um=k(ul−um)+k(ur−um)+(R2−R1)Eq,(37)-m\omega^2u_m=k(u_l-u_m)+k(u_r-u_m)+(R_2-R_1)Eq, \tag*{(37)}
mω2ur−k(ur−um)−ER2q=FR,(38)m\omega^2u_r-k(u_r-u_m)-ER_2q=F_R, \tag*{(38)}

from which we have

um=k2k−mω2ul+k2k−mω2ur+(R2−R1)Eq2k−mω2=12[ul+ur+ΔRqkE][1−(ωωΣkm2)2]−1(39)u_m=\frac{k}{2k-m\omega^2}u_l+\frac{k}{2k-m\omega^2}u_r+\frac{(R_2-R_1)Eq}{2k-m\omega^2}=\frac{1}{2}\left[u_l+u_r+\frac{\Delta Rq}{k}E\right]\left[1-\left(\frac{\omega}{\omega_{\Sigma km^2}}\right)^2\right]^{-1} \tag*{(39)}

where ΔR:=R2−R1\Delta R:=R_2-R_1. We note that this expression is similar to Eq. (2) but with an addition contribution to the displacement, due to the interaction of the bound charge with the electric field. Substituting Eq. (39) into Eqs. (34a) and (38) and rearranging terms yield

T[ulur]+h=[FLFR],(40)\mathrm{T}\begin{bmatrix}u_l\\u_r\end{bmatrix}+\mathbf{h}=\begin{bmatrix}F_L\\F_R\end{bmatrix}, \tag*{(40)}

where

T=[−mω2+k−k22k−mω2−k22k−mω2k22k−mω2mω2−k+k22k−mω2],h=−[q[(k−mω2)R1+kR2]2k−mω2q[kR1+(k−mω2)R2]2k−mω2]E.\mathrm{T}=\begin{bmatrix} -m\omega^2+k-\dfrac{k^2}{2k-m\omega^2} & -\dfrac{k^2}{2k-m\omega^2}\\ \dfrac{k^2}{2k-m\omega^2} & m\omega^2-k+\dfrac{k^2}{2k-m\omega^2} \end{bmatrix}, \qquad \mathbf{h}=-\begin{bmatrix} \dfrac{q[(k-m\omega^2)R_1+kR_2]}{2k-m\omega^2}\\ \dfrac{q[kR_1+(k-m\omega^2)R_2]}{2k-m\omega^2} \end{bmatrix}E.

By repeating the procedure introduced in § 3 to relate σ\sigma and pp (ε\varepsilon and vv) to FLF_L and FRF_R (ulu_l and uru_r), we obtain

\begin{bmatrix}\sigma\\p\end{bmatrix}=\mathrm{M}^{(1)}\begin{bmatrix}F_L\\F_R\end{bmatrix}, \qquad \mathrm{M}^{(1)}=\begin{bmatrix} -\frac{1}{2} & -\frac{1}{2}\\ -\frac{1}{2i\omega\Delta x} & \frac{1}{2i\omega\Delta x} \end{bmatrix}, \tag*{(41)}\end{bmatrix}
\begin{bmatrix}\varepsilon\\v\end{bmatrix}=\mathrm{M}^{(2)}\begin{bmatrix}u_l\\u_r\end{bmatrix}, \qquad \mathrm{M}^{(2)}=\begin{bmatrix} -\frac{1}{2\Delta x} & \frac{1}{2\Delta x}\\ -\frac{i\omega}{2} & -\frac{i\omega}{2} \end{bmatrix}. \tag*{(42)}\end{bmatrix}

We then combine Eqs. (40)–(42) to end up the following expressions relating the stress and momentum of the charged mass-spring system to the externally applied strain, velocity, and electric fields

\begin{bmatrix}\sigma\\p\end{bmatrix}=\mathrm{M}^{(1)}\mathrm{T}\mathrm{M}^{(2)-1}\begin{bmatrix}\varepsilon\\v\end{bmatrix}+\mathrm{M}^{(1)}\mathbf{h}=\begin{bmatrix}\widetilde{C}&0\\0&\widetilde{\rho}\end{bmatrix}\begin{bmatrix}\varepsilon\\v\end{bmatrix}-\begin{bmatrix}\widetilde{B}^{\dagger}\\\widetilde{W}^{\dagger}\end{bmatrix}E, \tag*{(43)}\end{bmatrix}

where

ρ~=3km−m2ω2(2k−mω2)Δx,C~=(k−mω2)Δx,\widetilde{\rho}=\frac{3km-m^2\omega^2}{(2k-m\omega^2)\Delta x}, \qquad \widetilde{C}=(k-m\omega^2)\Delta x,
B~†=−(R1+R2)q2=B~1†+B~2†2,(44)\widetilde{B}^{\dagger}=-\frac{(R_1+R_2)q}{2}=\frac{\widetilde{B}_1^{\dagger}+\widetilde{B}_2^{\dagger}}{2}, \tag*{(44)}
W~†=iωm(R2−R1)q2(2k−mω2)Δx=−iωmΔx(2k−mω2)B~2†−B~1†2.\widetilde{W}^{\dagger}=\frac{i\omega m(R_2-R_1)q}{2(2k-m\omega^2)\Delta x}=-\frac{i\omega m}{\Delta x(2k-m\omega^2)}\frac{\widetilde{B}_2^{\dagger}-\widetilde{B}_1^{\dagger}}{2}.

Rewriting the effective properties in terms of the characteristic localized resonance frequencies ω3km2:=3k/m\omega_{3km}^{2}:=3k/m, and ω1km2:=k/m\omega_{1km}^{2}:=k/m yields

ρ~=ΣmLuc1−(ω/ω3km)21−(ω/ωΣkm2)2,C~=k2Luc[1−(ω/ω1km)2],B~†=−(R1+R2)q2=B~1†+B~2†2,W~†=iωmΣkLuc[11−(ω/ωΣkm2)2](B~1†−B~2†).(45)\begin{aligned} \tilde{\rho} &= \frac{\Sigma m}{L_{\mathrm{uc}}}\frac{1-(\omega/\omega_{3km})^{2}}{1-(\omega/\omega_{\Sigma km2})^{2}}, \qquad\tilde{C} = \frac{k}{2}L_{\mathrm{uc}}\left[1-(\omega/\omega_{1km})^{2}\right],\\ \tilde{B}^{\dagger} &= -\frac{(R_{1}+R_{2})q}{2}=\frac{\tilde{B}_{1}^{\dagger}+\tilde{B}_{2}^{\dagger}}{2},\\ \tilde{W}^{\dagger} &= \frac{i\omega m}{\Sigma kL_{\mathrm{uc}}}\left[\frac{1}{1-(\omega/\omega_{\Sigma km2})^{2}}\right]\left(\tilde{B}_{1}^{\dagger}-\tilde{B}_{2}^{\dagger}\right). \tag*{(45)} \end{aligned}

The utility of the simple charged spring-mass system model is made clear through inspection of Eqs. (43)–(45), which illustrate three important points about the Willis- and electromomentum couplings. First, we note that by combining two elements of opposite Willis polarization/coupling, we observe zero effective Willis coupling since the masses and springs of the assembly are distributed symmetrically. Second, the piezoelectric coefficient of the assembly is the average of the piezoelectric coefficient of its elements. If we set R1=−R2R_{1}=-R_{2}, we get B~†=0\tilde{B}^{\dagger}=0, as one would anticipate, since the charge is distributed symmetrically in that case. Importantly, the linear momentum is coupled with the electric field through W~†\tilde{W}^{\dagger}, which we identify as the electromomentum coupling coefficient. Its leading term in a low frequency expansion is linear in ω\omega, and it is proportional to the asymmetry in the piezoelectric coefficient of the building blocks, i.e., W~†∝B~2†−B~1†\tilde{W}^{\dagger}\propto\tilde{B}_{2}^{\dagger}-\tilde{B}_{1}^{\dagger}. We also identify a local resonance that can amplify W~†\tilde{W}^{\dagger} when ω→ωΣkm2\omega\to\omega_{\Sigma km2}. This observation reinforces the continuum analysis of Pernas-Salomón and Shmuel [47], which also identified a resonance frequency of the electromomentum coefficient (see figure 4 therein).

To complete the calculation of the effective relations, it is left to find expressions for the electric displacement, DD, and to verify that it is indeed coupled with the velocity through some W~\tilde{W} that is identical to W~†\tilde{W}^{\dagger}, a condition that results from reciprocity [57]. To do that, we calculate again the change in the electric dipole moment, and the corresponding polarization density P=ΔpD/LucP=\Delta p_{\mathrm{D}}/L_{\mathrm{uc}}, namely,

ΔpD=R1qul+(R2−R1)qum−R2qur=ΔR2q22k[1−(ω/ωΣkm2)2]E−iωmΔRq2k[1−(ω/ωΣkm2)2]v−qΣRΔxε,(46)\begin{aligned} \Delta p_{\mathrm{D}} &= R_{1}qu_{l}+(R_{2}-R_{1})qu_{m}-R_{2}qu_{r}\\ &= \frac{\Delta R^{2}q^{2}}{2k\left[1-(\omega/\omega_{\Sigma km2})^{2}\right]}E-\frac{i\omega m\Delta Rq}{2k\left[1-(\omega/\omega_{\Sigma km2})^{2}\right]}v-q\Sigma R\Delta x\varepsilon, \tag*{(46)} \end{aligned}

where ΣR=R1+R2\Sigma R=R_{1}+R_{2} and we have used Eqs. (39), (41), and (42) to express ΔpD\Delta p_{\mathrm{D}} in terms of ε\varepsilon, vv, and EE. The resultant electric displacement field is then written as

D=ε0E+P=[ΔR2q22k(1−(ω/ωΣkm2)2)Luc+ε0]E−iωmΔRq2k[1−(ω/ωΣkm2)2]Lucv−ΣRq2ε,(47)D=\varepsilon_{0}E+P=\left[\frac{\Delta R^{2}q^{2}}{2k\left(1-(\omega/\omega_{\Sigma km2})^{2}\right)L_{\mathrm{uc}}}+\varepsilon_{0}\right]E-\frac{i\omega m\Delta Rq}{2k\left[1-(\omega/\omega_{\Sigma km2})^{2}\right]L_{\mathrm{uc}}}v-\frac{\Sigma Rq}{2}\varepsilon, \tag*{(47)}

or D~=B~ε+W~v+A~E\widetilde{D}=\widetilde{B}\varepsilon+\widetilde{W}v+\widetilde{A}E, with

B~=B~†=−ΣRq2,W~=W~†=−iωmΔRq2k[1−(ω/ωΣkm2)2]Luc,\widetilde{B}=\widetilde{B}^{\dagger}=-\frac{\Sigma Rq}{2},\qquad\widetilde{W}=\widetilde{W}^{\dagger}=-\frac{\mathrm{i}\omega m\Delta Rq}{2k\left[1-\left(\omega/\omega_{\Sigma km2}\right)^2\right]L_{\mathrm{uc}}},
A~=ΔR2q2(2k−mω2)Luc+ε0=ΔR2q22kLuc[1−(ω/ωΣkm2)2]+ε0.(48)\widetilde{A}=\frac{\Delta R^2q^2}{(2k-m\omega^2)L_{\mathrm{uc}}}+\varepsilon_0=\frac{\Delta R^2q^2}{2kL_{\mathrm{uc}}\left[1-\left(\omega/\omega_{\Sigma km2}\right)^2\right]}+\varepsilon_0. \tag*{(48)}

Indeed, Eq. (48) shows that not only B~=B~†\widetilde{B}=\widetilde{B}^{\dagger}, but also that W~=W~†\widetilde{W}=\widetilde{W}^{\dagger}, i.e., the electric displacement field is coupled with the velocity in the same way that the linear momentum is coupled with the electric field. Note that now the relative permittivity is not equal to zero, i.e., A~≠ε0\widetilde{A}\ne\varepsilon_0, since our model now considers three discrete masses in free space. Applying electric field will move each of the external masses freely and the middle mass will affect the polarization. Furthermore, we can see that the effective dielectric constant depends on the difference of the electric charge and on the springs’ stiffness such that if R1=R2R_1=R_2 the middle mass has no electric charge and the model will be equivalent to the model introduced in § 3 (A~=ε0\widetilde{A}=\varepsilon_0). In addition, if the stiffness will be big enough, again we will receive an equivalent model to the last model because the middle mass will not move and the ability of our model to polarize will be equivalent to two charges in free space. If we would change the electric charges of the left-, middle- and right mass to be R1q,R2q,R3qR_1q,R_2q,R_3q respectively, we could see that the dielectric constant depends only on the middle mass charge, i.e., R2qR_2q. Therefore, in order to receive a non-zero relative permittivity, we must model at least three charged mass that can move in space.

Collectively, Eqs. (43) and (47) establish the effective constitutive equations of the model, which we cast in the following matrix form

[σDp]=[k[1−(ω/ω1km)2]Δx−ΣRq20−ΣRq2−ΔR2q22kLuc[1−(ω/ωΣkm2)2]−ε0−iωmΔRq2k[1−(ω/ωΣkm2)2]Luc0−iωmΔRq2k[1−(ω/ωΣkm2)2]LucΣmLuc1−(ω/ω3km)21−(ω/ωΣkm2)2][ε−Ev]==[C~B~†0B~−A~W~0W~†ρ~][ε−Ev].(49)\begin{aligned} \begin{bmatrix} \sigma\\ D\\ p \end{bmatrix} = \begin{bmatrix} k\left[1-\left(\omega/\omega_{1km}\right)^2\right]\Delta x & -\frac{\Sigma Rq}{2} & 0\\ -\frac{\Sigma Rq}{2} & -\frac{\Delta R^2q^2}{2kL_{\mathrm{uc}}\left[1-\left(\omega/\omega_{\Sigma km2}\right)^2\right]}-\varepsilon_0 & -\frac{\mathrm{i}\omega m\Delta Rq}{2k\left[1-\left(\omega/\omega_{\Sigma km2}\right)^2\right]L_{\mathrm{uc}}}\\ 0 & -\frac{\mathrm{i}\omega m\Delta Rq}{2k\left[1-\left(\omega/\omega_{\Sigma km2}\right)^2\right]L_{\mathrm{uc}}} & \frac{\Sigma m}{L_{\mathrm{uc}}}\frac{1-\left(\omega/\omega_{3km}\right)^2}{1-\left(\omega/\omega_{\Sigma km2}\right)^2} \end{bmatrix} \begin{bmatrix} \varepsilon\\ -E\\ v \end{bmatrix} = \\ &= \begin{bmatrix} \widetilde{C} & \widetilde{B}^{\dagger} & 0\\ \widetilde{B} & -\widetilde{A} & \widetilde{W}\\ 0 & \widetilde{W}^{\dagger} & \widetilde{\rho} \end{bmatrix} \begin{bmatrix} \varepsilon\\ -E\\ v \end{bmatrix}. \tag*{(49)} \end{aligned}

Note that in the absence of mechanical asymmetry, the Willis coefficients are equal to zero in Eqs. (49), in contrast with the analysis of the continuum model by Pernas-Salomón et al. [48]. There, the Willis coefficients are nonzero even when there is no mechanical asymmetry, provided that the piezoelectric profile is asymmetric. We believe that the contribution of the piezoelectric asymmetry to the Willis coefficients is absent here since we have neglected the electric force between the charges (see § 3, where we assumed that Fc≪∣k(ur−ul)∣F_c \ll\lvert k(u_r-u_l)\rvert). These internal electric forces act in effect as an additional spring distribution, whose profile is asymmetric when the piezoelectric profile is asymmetric, and hence would contribute to the Willis coefficients, as in the continuum model of Pernas-Salomón et al. [48]. Our assumption here that Fc≪∣k(ur−ul)∣F_c \ll\lvert k(u_r-u_l)\rvert and therefore we can neglect FcF_c is consistent with the fact that the contribution of the asymmetry in the piezoelectric profile to the Willis coefficient in Ref. [48] is of order of magnitude smaller than the contribution of the mechanical asymmetry.

We can generalize Eq. (49) to a form that includes the Willis coupling by breaking also the symmetry in the mechanical parameters, i.e., creating a difference in the springs and/or the masses (Figure 4d). Accordingly, we change the stiffness of the left and right springs to k1k_1 and k2k_2, respectively, and change the density of building blocks masses, such that the resultant left, middle, and right masses in the assembly are equal to m1m_1, m2m_2 and m3m_3, respectively. By repeating the same procedure as before, we obtain

[σDp]=[Luck1k2Σk1−(ω/ωk1k2)21−(ω/ωΣkm2)2−[(2k2−m2ω2)R1+(2k1−m2ω2)R2]q2Σk[1−(ω/ωΣkm2)2]−iω(ΣkΔm+Δkm2)2Σk[1−(ω/ωΣkm2)2]−[(2k2−m2ω2)R1+(2k1−m2ω2)R2]q2Σk[1−(ω/ωΣkm2)2]−(qΔR)2ΣkLuc[1−(ω/ωΣkm2)2]−ϵ0iωm2LucΣkqΔR1−(ω/ωΣkm2)2−iω(ΣkΔm+Δkm2)2Σk[1−(ω/ωΣkm2)2]iωm2LucΣkqΔR1−(ω/ωΣkm2)2ΣmLuc1−(ω/ωΣkΣm)21−(ω/ωΣkm2)2][ε−Ev](50)\begin{bmatrix} \sigma\\ D \\ p \end{bmatrix} = \begin{bmatrix} \frac{L_{\mathrm{uc}}k_1k_2}{\Sigma k}\frac{1-(\omega/\omega_{k_1k_2})^2}{1-(\omega/\omega_{\Sigma k m_2})^2} & -\frac{\left[(2k_2-m_2\omega^2)R_1+(2k_1-m_2\omega^2)R_2\right]q}{2\Sigma k\left[1-(\omega/\omega_{\Sigma k m_2})^2\right]} & -\frac{i\omega(\Sigma k\Delta m+\Delta k m_2)}{2\Sigma k\left[1-(\omega/\omega_{\Sigma k m_2})^2\right]} \\[1.2em] -\frac{\left[(2k_2-m_2\omega^2)R_1+(2k_1-m_2\omega^2)R_2\right]q}{2\Sigma k\left[1-(\omega/\omega_{\Sigma k m_2})^2\right]} & -\frac{(q\Delta R)^2}{\Sigma kL_{\mathrm{uc}}\left[1-(\omega/\omega_{\Sigma k m_2})^2\right]}-\epsilon_0 & \frac{i\omega m_2}{L_{\mathrm{uc}}\Sigma k}\frac{q\Delta R}{1-(\omega/\omega_{\Sigma k m_2})^2} \\[1.2em] -\frac{i\omega(\Sigma k\Delta m+\Delta k m_2)}{2\Sigma k\left[1-(\omega/\omega_{\Sigma k m_2})^2\right]} & \frac{i\omega m_2}{L_{\mathrm{uc}}\Sigma k}\frac{q\Delta R}{1-(\omega/\omega_{\Sigma k m_2})^2} & \frac{\Sigma m}{L_{\mathrm{uc}}}\frac{1-(\omega/\omega_{\Sigma k\Sigma m})^2}{1-(\omega/\omega_{\Sigma k m_2})^2} \end{bmatrix} \begin{bmatrix} \varepsilon\\ -E \\ v \tag*{(50)} \end{bmatrix}

where ωk1k22:=4k1k2/ΣkΣm\omega_{k_1k_2}^2:=4k_1k_2/\Sigma k\Sigma m and ωΣkΣm2:=ΣmΣk/[m2(m1+m3)]\omega_{\Sigma k\Sigma m}^2:=\Sigma m\Sigma k/[m_2(m_1+m_3)]. Once again, we observe that the effective properties, except A~\tilde{A}, exhibit a resonance frequency at ωΣkm2=Σk/m2\omega_{\Sigma k m_2}=\Sigma k/m_2. In contrast with Eq. (49), we note that B~\tilde{B} and S~\tilde{S} (which was null) also exhibit strong frequency-dependence due to localized resonance. Interestingly, we also observe that S~\tilde{S} and W~\tilde{W} are related via

W~=2qΔRm2Luc(ΣkΔm+Δkm2)S~.(51)\tilde{W}=\frac{2q\Delta Rm_2}{L_{\mathrm{uc}}(\Sigma k\Delta m+\Delta k m_2)}\tilde{S}. \tag*{(51)}

This implies that the relative magnitude and sign of these coefficients can be set through the ratio of the charge, stiffness, and mass contrasts, and their relative orientation in space.

Summary

The generalization by Pernas-Salomón and Shmuel [47] to the homogenization scheme of Willis [7], from elastic- to piezoelectric constitutes, revealed that the effective linear momentum (electric displacement field) of piezoelectric composites is coupled with the electric (velocity) field. In this work, we have developed the simplest model—a one-dimensional assembly of charged masses and springs—that exhibits these electromomentum couplings. To this end, we have first generalized the discrete models of Muhlestein et al. [19] and Auld [61] for the Willis and piezoelectric effects, respectively. The final model, which demonstrates the electromomentum effect, comprises two elements, each of which exhibits a different piezoelectric coupling. The resultant expressions for the effective properties satisfy reciprocity, recover quasistatic benchmarks, and exhibit localized resonances which may be exploited to elicit very strong narrow-band Willis- and electromomentum couplings. When the excitation frequency is much lower than the characteristic frequency, we show that the Willis- and electromomentum coefficients are linear in the frequency, and weighted by the mechanical- and electromechanical asymmetry, respectively. This conclusion reinforces and unifies previous studies that used other methods of analysis [19, 32, 47, 48, 49, 54]. While this conclusion, together with some of the other conclusions above, were reported before, here they emerge thanks to a simple model. This model allows for a more intuitive understanding and compact representation of the cross-couplings, thereby providing a useful basis for the design of piezoelectric metamaterials.

Acknowledgments

We thank anonymous reviewers for constructive feedback that helped improve this paper. This project was funded by the European Union (ERC, EXCEPTIONAL, Project No. 101045494); and the Israel Science Foundation, funded by the Israel Academy of Sciences and Humanities (Grant no. 2061/20). MRH acknowledges support from the Defense Advance Research Project Agency (DARPA) and the Army Research Office and was accomplished under Grant Number W911NF-20-1-0349.

References

  1. [1]Philip W Anderson. More is different: broken symmetry and the nature of the hierarchical structure of science. Science, 177(4047):393–396, 1972.
  2. [2]J R Willis. The nonlocal influence of density variations in a composite. Int. J. Solids Struct., 21(7):805–817, 1985. ISSN 0020-7683. doi: https://doi.org/10.1016/0020-7683(85)90084-8. URL http://www.sciencedirect.com/science/article/pii/0020768385900848.
  3. [3]J R Willis. Variational principles for dynamic problems for inhomogeneous elastic media. Wave Motion, 3(1):1–11, 1981. ISSN 0165-2125. doi: https://doi.org/10.1016/0165-2125(81)90008-1. URL http://www.sciencedirect.com/science/article/pii/0165212581900081.
  4. [4]John R Willis. Variational and related methods for the overall properties of composites. Advances in applied mechanics, 21:1–78, 1981.DOI
  5. [5]J R Willis. Continuum Micromechanics. chapter Dynamics o, pages 265–290. Springer-Verlag New York, Inc., New York, NY, USA, 1997. ISBN 3-211-82902-4. URL http://dl.acm.org/citation.cfm?id=265753.265760.
  6. [6]J R Willis. Exact effective relations for dynamics of a laminated body. Mech. Mater., 41(4):385–393, 2009. ISSN 0167-6636. doi: http://dx.doi.org/10.1016/j.mechmat.2009.01.010. URL http://www.sciencedirect.com/science/article/pii/S0167663609000118.
  7. [7]J R Willis. Effective constitutive relations for waves in composites and metamaterials. Proc. R. Soc. London A Math. Phys. Eng. Sci., 467(2131):1865–1879, 2011. ISSN 1364-5021. doi: 10.1098/rspa.2010.0620. URL http://rspa.royalsocietypublishing.org/content/467/2131/1865.DOI
  8. [8]J.R. Willis. A comparison of two formulations for effective relations for waves in a composite. Mechanics of Materials, 47:51 – 60, 2012. ISSN 0167-6636. doi: https://doi.org/10.1016/j.mechmat.2011.12.008. URL http://www.sciencedirect.com/science/article/pii/S0167663612000105.
  9. [9]John R. Willis. The construction of effective relations for waves in a composite. Comptes Rendus Mécanique, 340(4):181 – 192, 2012. ISSN 1631-0721. doi: https://doi.org/10.1016/j.crme.2012.02.001. URL http://www.sciencedirect.com/science/article/pii/S1631072112000381. Recent Advances in Micromechanics of Materials.
  10. [10]G W Milton, M Briane, and J R Willis. On cloaking for elasticity and physical equations with a transformation invariant form. New J. Phys., 8(10):248, 2006. URL http://stacks.iop.org/1367-2630/8/i=10/a=248.DOI
  11. [11]Graeme W. Milton. A unifying perspective on linear continuum equations prevalent in physics. part ii: Canonical forms for time-harmonic equations. arXiv: Analysis of PDEs, 2020.
  12. [12]Graeme W Milton. New metamaterials with macroscopic behavior outside that of continuum elastodynamics. New Journal of Physics, 9(10):359–359, oct 2007. doi: 10.1088/1367-2630/9/10/359. URL https://doi.org/10.1088%2F1367-2630%2F9%2F10%2F359.
  13. [13]Michael R Haberman and Matthew D Guild. Acoustic metamaterials. Phys. Today, 69(6):42–48, 2016.DOI
  14. [14]Steven A. Cummer, Johan Christensen, and Andrea Alù. Controlling sound with acoustic metamaterials. Nature Reviews Materials, 1:16001 EP –, 02 2016. URL https://doi.org/10.1038/natrevmats.2016.1.
  15. [15]Timothy Leighton. The acoustic bubble. Academic press, 2012.
  16. [16]Valentin Leroy, Martin Devaud, and Jean-Claude Bacri. The air bubble: Experiments on an unusual harmonic oscillator. American Journal of Physics, 70(10):1012–1019, 2002.DOI
  17. [17]Kyle S Spratt, Mark F Hamilton, Kevin M Lee, and Preston S Wilson. Radiation damping of, and scattering from, an arbitrarily shaped bubble. The Journal of the Acoustical Society of America, 142(1):160–166, 2017.DOI
  18. [18]Sam H Lee and Oliver B Wright. Origin of negative density and modulus in acoustic metamaterials. Physical Review B, 93(2):024302, 2016.DOI
  19. [19]Michael B Muhlestein, Caleb F Sieck, Andrea Alù, and Michael R Haberman. Reciprocity, passivity and causality in Willis materials. Proc. R. Soc. London A Math. Phys. Eng. Sci., 472(2194), 2016. ISSN 1364-5021. doi: 10.1098/rspa.2016.0604. URL http://rspa.royalsocietypublishing.org/content/472/2194/20160604.DOI
  20. [20]Muamer Kadic, Graeme W. Milton, Martin van Hecke, and Martin Wegener. 3d metamaterials. Nature Reviews Physics, 1(3):198–210, 2019. doi: 10.1038/s42254-018-0018-y. URL https://doi.org/10.1038/s42254-018-0018-y.
  21. [21]Johan Christensen, Muamer Kadic, Oliver Kraft, and Martin Wegener. Vibrant times for mechanical metamaterials. MRS Communications, 5:453–462, 2015. doi: 10.1557/mrc.2015.51.DOI
  22. [22]Richard V Craster and Sébastien Guenneau. Acoustic metamaterials: Negative refraction, imaging, lensing and cloaking, volume 166. Springer Science & Business Media, 2012.DOI
  23. [23]Ankit Srivastava. Elastic metamaterials and dynamic homogenization: a review. International Journal of Smart and Nano Materials, 6(1):41–60, 2015.arxiv.org/abs/1411.3001
  24. [24]C. R. Simovski. Material parameters of metamaterials (a review). Optics and Spectroscopy, 107(5):726, 2009. doi: 10.1134/S0030400X09110101. URL https://doi.org/10.1134/S0030400X09110101.
  25. [25]Ben Lustig, Guy Elbaz, Alan Muhafra, and Gal Shmuel. Anomalous energy transport in laminates with exceptional points. Journal of the Mechanics and Physics of Solids, page 103719, 2019. ISSN 0022-5096. doi: https://doi.org/10.1016/j.jmps.2019.103719. URL http://www.sciencedirect.com/science/article/pii/S0022509619306921.
  26. [26]A Sridhar, V G Kouznetsova, and M G D Geers. A general multiscale framework for the emergent effective elastodynamics of metamaterials. J. Mech. Phys. Solids, 111:414–433, 2018. ISSN 0022-5096. doi: https://doi.org/10.1016/j.jmps.2017.11.017. URL http://www.sciencedirect.com/science/article/pii/S0022509617306245.
  27. [27]Joo Hwan Oh, Shuibao Qi, Yoon Young Kim, and Badreddine Assouar. Elastic metamaterial insulator for broadband low-frequency flexural vibration shielding. Phys. Rev. Applied, 8:054034, Nov 2017. doi: 10.1103/PhysRevApplied.8.054034. URL https://link.aps.org/doi/10.1103/PhysRevApplied.8.054034.
  28. [28]Ankit Srivastava. Causality and passivity: From electromagnetism and network theory to metamaterials. Mechanics of Materials, 154:103710, 2021. ISSN 0167-6636. doi: https://doi.org/10.1016/j.mechmat.2020.103710. URL https://www.sciencedirect.com/science/article/pii/S0167663620307377.
  29. [29]Shixu Meng and Bojan B Guzina. On the dynamic homogenization of periodic media: Willis’ approach versus two-scale paradigm. Proc. R. Soc. London A Math. Phys. Eng. Sci., 474(2213), 2018. ISSN 1364-5021. doi: 10.1098/rspa.2017.0638. URL http://rspa.royalsocietypublishing.org/content/474/2213/20170638.
  30. [30]Hussein Nassar, Qi-Chang He, and Nicolas Auffray. Willis elastodynamic homogenization theory revisited for periodic media. Journal of the Mechanics and Physics of Solids, 2015.
  31. [31]A N Norris, A L Shuvalov, and A A Kutsenko. Analytical formulation of three-dimensional dynamic homogenization for periodic elastic systems. Proc. R. Soc. London A Math. Phys. Eng. Sci., 2012. ISSN 1364-5021. doi: 10.1098/rspa.2011.0698. URL http://rspa.royalsocietypublishing.org/content/early/2012/02/14/rspa.2011.0698.
  32. [32]Caleb F Sieck, Andrea Alù, and Michael R Haberman. Origins of Willis coupling and acoustic bianisotropy in acoustic metamaterials through source-driven homogenization. Phys. Rev. B, 96(10):104303, 2017. doi: 10.1103/PhysRevB.96.104303. URL https://link.aps.org/doi/10.1103/PhysRevB.96.104303.
  33. [33]Ankit Srivastava. Causality and passivity in elastodynamics. Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, 471 (2180), 2015. ISSN 1364-5021. doi: 10.1098/rspa.2015.0256. URL http://rspa.royalsocietypublishing.org/content/471/2180/20150256.
  34. [34]Daniel Torrent, Yan Pennec, and Bahram Djafari-Rouhani. Resonant and nonlocal properties of phononic metasolids. Phys. Rev. B, 92(17):174110, 2015. doi: 10.1103/PhysRevB.92.174110. URL https://link.aps.org/doi/10.1103/PhysRevB.92.174110.DOI
  35. [35]Michael B. Muhlestein, Caleb F. Sieck, Preston S. Wilson, and Michael R. Haberman. Experimental evidence of Willis coupling in a one-dimensional effective material element. Nature Communications, 8:15625 EP –, 06 2017. URL https://doi.org/10.1038/ncomms15625.
  36. [36]Anton Melnikov, Yan Kei Chiang, Li Quan, Sebastian Oberst, Andrea Alù, Steffen Marburg, and David Powell. Acoustic meta-atom with experimentally verified maximum Willis coupling. Nature Communications, 10(1):3148, 2019. doi: 10.1038/s41467-019-10915-5. URL https://doi.org/10.1038/s41467-019-10915-5.
  37. [37]A. J. Lawrence, Benjamin M. Goldsberry, Samuel P. Wallen, and Michael R. Haberman. Numerical study of acoustic focusing using a bianisotropic acoustic lens. The Journal of the Acoustical Society of America, 148(4):EL365–EL369, 10 2020. ISSN 0001-4966. doi: 10.1121/10.0002137.
  38. [38]Jeonghoon Park, Dongwoo Lee, and Junsuk Rho. Recent advances in non-traditional elastic wave manipulation by macroscopic artificial structures. Applied Sciences, 10(2), 2020. ISSN 2076-3417. doi: 10.3390/app10020547. URL https://www.mdpi.com/2076-3417/10/2/547.DOI
  39. [39]Yan Kei Chiang, Sebastian Oberst, Anton Melnikov, Li Quan, Steffen Marburg, Andrea Alù, and David A Powell. Reconfigurable acoustic metagrating for high-efficiency anomalous reflection. Physical Review Applied, 13(6):064067, 2020.DOI
  40. [40]Yangyang Chen, Xiaopeng Li, Gengkai Hu, Michael R. Haberman, and Guoliang Huang. An active mechanical Willis meta-layer with asymmetric polarizabilities. Nature Communications, 11(1):3681, 2020. doi: 10.1038/s41467-020-17529-2. URL https://doi.org/10.1038/s41467-020-17529-2.
  41. [41]Bogdan-Ioan Popa, Yuxin Zhai, and Hyung-Suk Kwon. Broadband sound barriers with bianisotropic metasurfaces. Nature Communications, 9(1):5299, 2018. doi: 10.1038/s41467-018-07809-3. URL https://doi.org/10.1038/s41467-018-07809-3.
  42. [42]Aurélien Merkel, Vicent Romero-García, Jean-Philippe Groby, Jensen Li, and Johan Christensen. Unidirectional zero sonic reflection in passive 𝒫𝒯-symmetric Willis media. Phys. Rev. B, 98:201102, Nov 2018. doi: 10.1103/PhysRevB.98.201102. URL https://link.aps.org/doi/10.1103/PhysRevB.98.201102.DOI
  43. [43]Junfei Li, Chen Shen, Ana D’iaz-Rubio, Sergei A Tretyakov, and Steven A Cummer. Systematic design and experimental demonstration of bianisotropic metasurfaces for scattering-free manipulation of acoustic wavefronts. Nat. Commun., 9(1):1342, 2018. doi: 10.1038/s41467-018-03778-9. URL https://doi.org/10.1038/s41467-018-03778-9.
  44. [44]Yongquan Liu, Zixian Liang, Jian Zhu, Lingbo Xia, Olivier Mondain-Monval, Thomas Brunet, Andrea Alù, and Jensen Li. Willis metamaterial on a structured beam. Phys. Rev. X, 9:011040, Feb 2019. doi: 10.1103/PhysRevX.9.011040. URL https://link.aps.org/doi/10.1103/PhysRevX.9.011040.DOI
  45. [45]Junfei Li, Ailing Song, and Steven A. Cummer. Bianisotropic acoustic metasurface for surface-wave-enhanced wavefront transformation. Phys. Rev. Applied, 14:044012, Oct 2020. doi: 10.1103/PhysRevApplied.14.044012. URL https://link.aps.org/doi/10.1103/PhysRevApplied.14.044012.DOI
  46. [46]Xiaoshi Su and Andrew N. Norris. Retrieval method for the bianisotropic polarizability tensor of willis acoustic scatterers. Phys. Rev. B, 98:174305, Nov 2018. doi: 10.1103/PhysRevB.98.174305. URL https://link.aps.org/doi/10.1103/PhysRevB.98.174305.DOI
  47. [47]René Pernas-Salomón and Gal Shmuel. Symmetry breaking creates electro-momentum coupling in piezoelectric metamaterials. Journal of the Mechanics and Physics of Solids, 134:103770, 2020. ISSN 0022-5096. doi: https://doi.org/10.1016/j.jmps.2019.103770. URL http://www.sciencedirect.com/science/article/pii/S0022509619306386.
  48. [48]René Pernas-Salomón, Michael R. Haberman, Andrew N. Norris, and Gal Shmuel. The electromomentum effect in piezoelectric willis scatterers. Wave Motion, page 102797, 2021. doi: https://doi.org/10.1016/j.wavemoti.2021.102797. URL https://www.sciencedirect.com/science/article/pii/S0165212521000950.
  49. [49]Alan Muhafra, Majd Kosta, Daniel Torrent, René Pernas-Salomón, and Gal Shmuel. Homogenization of piezoelectric planar willis materials undergoing antiplane shear. Wave Motion, 108:102833, 2022. ISSN 0165-2125. doi: https://doi.org/10.1016/j.wavemoti.2021.102833. URL https://www.sciencedirect.com/science/article/pii/S0165212521001311.
  50. [50]Jeong-Ho Lee, Zhizhou Zhang, and Grace X. Gu. Generalized effective dynamic constitutive relation for heterogeneous media: Beyond the quasi-infinite and periodic limits, 2023.arxiv.org/abs/2303.04389
  51. [51]Jeong-Ho Lee, Zhizhou Zhang, and Grace X. Gu. Maximum electro-momentum coupling in piezoelectric metamaterial scatterers. Journal of Applied Physics, 132(12):125108, 2022.DOI
  52. [52]Samuel P Wallen, Matthew A Casali, Benjamin M Goldsberry, and Michael R Haberman. Polarizability of electromomentum coupled scatterers. In Proceedings of Meetings on Acoustics 182ASA, volume 46, page 065002. Acoustical Society of America, 2022.DOI
  53. [53]Zhizhou Zhang, Jeong-Ho Lee, and Grace X. Gu. Rational design of piezoelectric metamaterials with tailored electro-momentum coupling. Extreme Mechanics Letters, 55:101785, 2022. ISSN 2352-4316. doi: https://doi.org/10.1016/j.eml.2022.101785. URL https://www.sciencedirect.com/science/article/pii/S2352431622001079.
  54. [54]Majd Kosta, Alan Muhafra, Rene Pernas-Salomón, Gal Shmuel, and Oded Amir. Maximizing the electromomentum coupling in piezoelectric laminates. International Journal of Solids and Structures, 254-255:111909, 2022. ISSN 0020-7683. doi: https://doi.org/10.1016/j.ijsolstr.2022.111909. URL https://www.sciencedirect.com/science/article/pii/S0020768322003705.
  55. [55]Hai D. Huynh, Xiaoying Zhuang, Harold S. Park, S.S. Nanthakumar, Yabin Jin, and Timon Rabczuk. Maximizing electro-momentum coupling in generalized 2d willis metamaterials. Extreme Mechanics Letters, 61:101981, 2023. ISSN 2352-4316. doi: https://doi.org/10.1016/j.eml.2023.101981. URL https://www.sciencedirect.com/science/article/pii/S2352431623000275.
  56. [56]Jeong-Ho Lee, Zhizhou Zhang, and Grace X. Gu. Reaching new levels of wave scattering via piezoelectric metamaterials and electro-momentum coupling. The Journal of the Acoustical Society of America, 153:A163–A163, 03 2023. ISSN 0001-4966. doi: 10.1121/10.0018518.DOI
  57. [57]René Pernas-Salomón and Gal Shmuel. Fundamental Principles for Generalized Willis Metamaterials. Phys. Rev. Applied, 14:064005, Dec 2020. doi: 10.1103/PhysRevApplied.14.064005. URL https://link.aps.org/doi/10.1103/PhysRevApplied.14.064005.
  58. [58]Shahrokh Sepehrirahnama, Sebastian Oberst, Yan Kei Chiang, and David A. Powell. Willis coupling-induced acoustic radiation force and torque reversal. Phys. Rev. Lett., 129:174501, Oct 2022. doi: 10.1103/PhysRevLett.129.174501. URL https://link.aps.org/doi/10.1103/PhysRevLett.129.174501.DOI
  59. [59]G W Milton and J R Willis. On modifications of Newton’s second law and linear continuum elastodynamics. Proc. R. Soc. London A Math. Phys. Eng. Sci., 463(2079):855–880, 2007. ISSN 1364-5021. doi: 10.1098/rspa.2006.1795. URL http://rspa.royalsocietypublishing.org/content/463/2079/855.DOI
  60. [60]Hongfei Qu, Xiaoning Liu, and Gengkai Hu. Mass-spring model of elastic media with customizable willis coupling. International Journal of Mechanical Sciences, 224:107325, 2022. ISSN 0020-7403. doi: https://doi.org/10.1016/j.ijmecsci.2022.107325. URL https://www.sciencedirect.com/science/article/pii/S0020740322002375.
  61. [61]B.A. Auld. Acoustic fields and waves in solids. A Wiley-Interscience publication. Wiley, 1973. ISBN 978885013437. URL https://books.google.co.il/books?id=_2MWAwAAQBAJ.
  62. [62]C R Simovski. On electromagnetic characterization and homogenization of nanostructured metamaterials. Journal of Optics, 13(1):013001, nov 2010. doi: 10.1088/2040-8978/13/1/013001. URL https://doi.org/10.1088/2040-8978/13/1/013001.
  63. [63]Linear theory of piezoelectricity. IEEE Transactions on Sonics and Ultrasonics, 31(2):10–13, March 1984. ISSN 0018-9537. doi: 10.1109/T-SU.1984.31467.DOI

Paper details

vlnc:W_r4h26emz
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents