Abstract
In a previous study (Marigo et al. in J. Mech. Phys. Solids 143:104029, 2020) we have studied the effect of a periodic array of subwavelength plates or beams over a semi-infinite elastic ground on the propagation of waves hitting the interface. The study was restricted to the low frequency regime where only flexural resonances take place. Here, we present a generalization to higher frequencies which allows us to account for both flexural and longitudinal resonances and to evaluate their interplay. An effective model is obtained using asymptotic analysis and homogenization techniques, which can be expressed in terms of the ground alone with an effective dynamic (frequency-dependent) boundary conditions of the Robin’s type. For an in-plane wave at oblique incidence, the scattered displacement fields and the reflection coefficients are obtained in closed forms and their effectiveness to reproduce the actual scattering is inspected by comparison with direct numerics in a two-dimensional setting.
Similar content being viewed by others
Notes
The dispersion relation of the flexural resonances reads \(D(\kappa )=(1+\cos \kappa h_{\text{\tiny b}}\text{ch}\kappa h_{\text{\tiny b}})=0\) resulting in \(\kappa h_{\text{\tiny b}}=1.875\), 4.694, 7.855, 10.995, ⋯; the approximation \(\kappa h_{\text{\tiny b}}\sim \frac{\pi }{2}+n\pi \) provides \(\kappa h_{\text{\tiny b}}=\) 1.571, 4.712, 7.854, 10.995 ⋯ and the agreement gets better as \(n\) increases.
The parameter \({\mathcal{K}}\) can be written \({\mathcal{K}}=\frac{K}{\kappa }= \frac{h_{\text{\tiny F}}}{h_{\text{\tiny L}}}\); at prescribed frequency, it measures the relative wavelengths associated with flexural and longitudinal motions.
At the occurrence of the first longitudinal resonance, \(f_{\text{\tiny L}}\) diverges in (44) resulting in \(u_{z}(x,0)=0\) in (43); this was not captured in the model of [1] as \(f_{\text{\tiny L}}=\varphi \frac{\rho _{\text{\tiny b}}}{\rho _{\text{\tiny s}}}k_{\text{\tiny T}}h_{\text{\tiny b}}\) was obtained as an approximation of \(f_{\text{\tiny L}}\) in (43) for \(Kh_{\text{\tiny b}}\ll 1\).
In Fig. 7, spectra are reported against \(\theta _{\text{\tiny L}}\in (0,90^{\circ })\) in their lower-parts. As \(\theta _{\text{\tiny L}}=90^{\circ }\), \(\theta _{\text{\tiny T}}=\theta _{c}\) with \(\theta _{c}=\text{asin}\sqrt{ \frac{\mu _{\text{\tiny s}}}{\lambda _{\text{\tiny s}}+2\mu _{\text{\tiny s}}}} \simeq 37.8^{\circ }\). Increasing further the incidence of the transverse wave above \(\theta _{c}\), the longitudinal waves become evanescent and no mode conversion occurs (hence \(|R_{\text{\tiny TT}}|=1\) and Fig. 7 shows the real part). Note that the expressions of the reflection coefficients in (50) remain valid, with \(\sin \theta _{\text{\tiny L}}=\frac{\beta }{k_{\text{\tiny L}}}>1\), hence \(\alpha _{\text{\tiny L}}\) becomes purely imaginary.
It is worth noting that the discrepancy between our model (dotted black lines) and that neglecting flexural motions (dashed-dotted green lines) is more prominent for the absolute values reported in Fig. 9 than for the real parts in Figs. 7 and 8. This is due to the fact that imaginary parts of the reflexion coefficients have more pronounced variations around resonances (they cancel for the substrate on its own, (50)).
Expressions of the reflection coefficients (53) apply at longitudinal resonances, \(|f_{\text{\tiny L}}|\to \infty \). Symmetrically, at flexural resonances, \(|f_{\text{\tiny F}}|\to \infty \), (50) simplify to
$$ \displaystyle R_{\text{\tiny TT}}= \frac{ \cos \theta _{\text{\tiny T}}-i\xi ^{-1}f_{\text{\tiny L}}\cos (\theta _{\text{\tiny L}}+\theta _{\text{\tiny T}})}{\cos \theta _{\text{\tiny T}}-i\xi ^{-1}f_{\text{\tiny L}}\cos (\theta _{\text{\tiny L}}-\theta _{\text{\tiny T}})}, \quad \displaystyle R_{\text{\tiny LT}}= \frac{-i\sin 2\theta _{\text{\tiny T}}f_{\text{\tiny L}}}{\cos \theta _{\text{\tiny T}}-i\xi ^{-1}f_{\text{\tiny L}}\cos (\theta _{\text{\tiny L}}-\theta _{\text{\tiny T}})}, $$hence strictly, we do not have \(|R_{\text{\tiny TT}}|=1\) and \(|R_{\text{\tiny LT}}=0\) at flexural resonances. This is less visible in Fig. 9 as most of these resonances take place as \(|f_{\text{\tiny F}}|\ll 1\) (far from a longitudinal resonance).
References
Marigo, J.-J., Pham, K., Maurel, A., Guenneau, S.: Effective model for elastic waves propagating in a substrate supporting a dense array of plates/beams with flexural resonances. J. Mech. Phys. Solids 143, 104029 (2020)
Xiao, Y., Wen, J., Wen, X.: Flexural wave band gaps in locally resonant thin plates with periodically attached spring–mass resonators. J. Phys. D, Appl. Phys. 45(19), 195401 (2012)
Torrent, D., Mayou, D., Sánchez-Dehesa, J.: Elastic analog of graphene: Dirac cones and edge states for flexural waves in thin plates. Phys. Rev. B 87(11), 115143 (2013)
Maznev, A., Gusev, V.: Waveguiding by a locally resonant metasurface. Phys. Rev. B 92(11), 115422 (2015)
Williams, E.G., Roux, P., Rupin, M., Kuperman, W.: Theory of multiresonant metamaterials for a 0 lamb waves. Phys. Rev. B 91(10), 104307 (2015)
Yoritomo, J.Y., Weaver, R.L., Roux, P., Rupin, M., Williams, E.G.: On band gap predictions for multiresonant metamaterials on plates. J. Acoust. Soc. Am. 139(3), 1282–1284 (2016)
Colquitt, D., Colombi, A., Craster, R., Roux, P., Guenneau, S.: Seismic metasurfaces: sub-wavelength resonators and Rayleigh wave interaction. J. Mech. Phys. Solids 99, 379–393 (2017)
Wootton, N., Erbas, B., Kaplunov, J., Wootton, P.: Approximate analysis of surface wave-structure interaction. J. Mech. Mater. Struct. 13(2), 297–309 (2018)
Wootton, P., Kaplunov, J., Colquitt, D.: An asymptotic hyperbolic–elliptic model for flexural-seismic metasurfaces. Proc. R. Soc. A 475(2227), 20190079 (2019)
Lott, M., Roux, P.: Effective impedance of a locally resonant metasurface. Phys. Rev. Mater. 3(6), 065202 (2019)
Lott, M., Roux, P.: Locally resonant metamaterials for plate waves: the respective role of compressional versus flexural resonances of a dense forest of vertical rods. In: Fundamentals and Applications of Acoustic Metamaterials: From Seismic to Radio Frequency, vol. 1, pp. 25–45 (2019)
Wootton, P.T., Kaplunov, J., Prikazchikov, D.: A second-order asymptotic model for Rayleigh waves on a linearly elastic half plane. IMA J. Appl. Math. 85(1), 113–131 (2020)
Achaoui, Y., Khelif, A., Benchabane, S., Robert, L., Laude, V.: Experimental observation of locally-resonant and Bragg band gaps for surface guided waves in a phononic crystal of pillars. Phys. Rev. B 83(10), 104201 (2011)
Colombi, A., Roux, P., Guenneau, S., Gueguen, P., Craster, R.V.: Forests as a natural seismic metamaterial: Rayleigh wave bandgaps induced by local resonances. Sci. Rep. 6, 19238 (2016)
Lott, M., Roux, P., Garambois, S., Guéguen, P., Colombi, A.: Evidence of metamaterial physics at the geophysics scale: the metaforet experiment. Geophys. J. Int. 220(2), 1330–1339 (2020)
Achenbach, J.: Wave Propagation in Elastic Solids, vol. 16. Elsevier, Amsterdam (2012)
Nieves, M., Carta, G., Jones, I., Movchan, A., Movchan, N.: Vibrations and elastic waves in chiral multi-structures. J. Mech. Phys. Solids 121, 387–408 (2018)
Maurel, A., Pham, K.: Multimodal method for the scattering by an array of plates connected to an elastic half-space. J. Acoust. Soc. Am. 146, 4402 (2019)
Pham, K., Maurel, A., Félix, S., Guenneau, S.: Hybridized love waves in a guiding layer supporting an array of plates with decorative endings. Materials 13(7), 1632 (2020)
Looker, J.R., Sader, J.E.: Flexural resonant frequencies of thin rectangular cantilever plates. J. Appl. Mech. 75(1), 011007 (2008)
Acknowledgements
K.P. acknowledges support of the Agence de l’Innovation de Défense (AID) from the Direction Générale de l’Armement (DGA) under grant 2019 65 0070 and the Agence Nationale de la Recherche under grant ANR-19-CE08-0006.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendices
Appendix A: Proofs of Some Properties
Throughout the appendix, we shall use that for a symmetric tensor \(\tau _{ij}\) defined for \(\hat{\mathbf{x}}' \in \hat{\mathsf{s}}_{\text{\tiny b}}\), and for a vector \(a_{i}\) with antisymmetric tensor \(\partial _{j} a_{i}\), we have
since \(\partial _{j}a_{i}\tau _{ij}=0\).
1.1 A.1 Determination of \(({\mathbf{V}}^{0},{\mathbf{V}}^{1})\) and of \(({\boldsymbol{\varSigma }}^{0},{\mathbf{V}}^{2})\)
As previously said in §3.1.3, the dependence of \({\mathbf{V}}^{0}\) and \({\mathbf{V}}^{1}\) on \(\hat{\mathbf{x}}'=(\hat{x},\hat{y})\) is obtained starting from \((\text{C'})^{-2}\) and \((\text{C'})^{-1}\) in (13) which tell us that there exists \(({\mathbf{U}}^{0}({\mathbf{x}},\tilde{z}),\varOmega ^{0}({\mathbf{x}},\tilde{z}))\) and \(({\mathbf{U}}^{1}({\mathbf{x}},\tilde{z}),\varOmega ^{1}({\mathbf{x}},\tilde{z}))\) such that
Next, as \(\frac{\partial }{\partial \hat{x}}V_{z}^{1}=- \frac{\partial }{\partial \tilde{z}}V_{x}^{0}\), \(\frac{\partial }{\partial \hat{y}}V_{z}^{1}=- \frac{\partial }{\partial \tilde{z}}V_{y}^{0}\) from \((\text{C'})^{-1}\), we have \(\frac{\partial ^{2} }{\partial \hat{x}\partial \hat{y}}V_{z}^{1}=\pm \frac{\partial }{\partial \tilde{z}}\varOmega ^{0}=0\), hence
The dependence of \({\mathbf{V}}^{2}\) with respect to \(\hat{\mathbf{x}}'\) is more intricate as it requires to solve the following problem set in the section \(\hat{\mathbf{x}}'\in \hat{\mathsf{s}}_{\text{\tiny b}}\) with \(({\boldsymbol{\varSigma }}^{0},{\mathbf{V}}^{2})\) as unknowns:
with the input \({\mathbf{V}}^{0}\) and \({\mathbf{V}}^{1}\) given by (56). For a circular section, this problem can be solved explicitly. Specifically there exists \(({\mathbf{U}}^{2}({\mathbf{x}},\tilde{z}),\varOmega ^{2}({\mathbf{x}},\tilde{z}))\) such that the displacement \({\mathbf{V}}^{2}\) is given by
and the corresponding stress tensor \({\boldsymbol{\varSigma }}^{0}\) by
1.2 A.2 Proof of \(P_{1} :\frac{\partial U_{z}^{1}}{\partial \tilde{z}}=0\) in (15)
We use \((\text{E}_{z})^{-1}\) in (12), specifically
that we integrate over \(\hat{\mathbf{x}}'\in \hat{\mathsf{s}}_{\text{\tiny b}}\). Accounting for the boundary condition \(\varSigma _{xz}^{1} n_{x}+\varSigma _{yz}^{1}n_{y}=0\) on \(\partial \hat{\mathsf{s}}_{\text{\tiny b}}\), we obtain that \(\frac{\partial }{\partial \tilde{z}}\int _{\hat{\mathsf{s}}_{\text{\tiny b}}}\varSigma _{zz}^{0}\,\text{d}\hat{\mathbf{x}}'=0\). Next, with the form of \(\varSigma _{zz}^{0}\) given in (59), and accounting for the facts that (i) \(U_{z}^{0}\) in (56) does not depend on \(\tilde{z}\) and (ii) \(\int _{\hat{\mathsf{s}}_{\text{\tiny b}}}\hat{x}\,\text{d}\hat{\mathbf{x}}'= \int _{\hat{\mathsf{s}}_{\text{\tiny b}}}\hat{y}\,\text{d}\hat{\mathbf{x}}'=0\), we obtain that \(\frac{\partial ^{2} U_{z}^{1}}{\partial \tilde{z}^{2}}=0\). The field \(U_{z}^{1}\) could be a linear function of \(\tilde{z}\) but this is prevented by its \(\tilde{z}\)-periodicity hence
1.3 A.3 Proof of \(P_{2} :\frac{\partial \varOmega ^{1}}{\partial \tilde{z}}=0\) in (15)
We start by introducing the torsion force \({\mathcal{T}}^{0}\)
Using the forms of \(\varSigma _{xz}^{0}\) and \(\varSigma _{yz}^{0}\) in (59) we get that
(\(J_{\text{\tiny b}}\) is the torsion constant for a circular cross-section.) Now, we use \(\mathrm{div}_{{\hat{\mathbf{x}}}'}{\boldsymbol{\varSigma }^{1}}+\mathrm{div}_{ \tilde{z}}{\boldsymbol{\varSigma }^{0}}={\mathbf{0}}\), from \((\text{E}_{a})^{-1}\) and \((\text{E}_{z})^{-1}\) in (12), that we integrate over \(\hat{\mathsf{s}}_{\text{\tiny b}}\) after multiplication by \({\boldsymbol{a}}=\hat{y}{\mathbf{e}}_{x}-\hat{x}{\mathbf{e}}_{y}\); we get
In (63), we have applied (55) to the first integral involving \({\boldsymbol{\varSigma }^{1}}\). Indeed, it suffices to notice that \(\mathrm{div}_{{\hat{\mathbf{x}}}'}{\boldsymbol{\varSigma }^{1}}= \mathrm{div}_{{ \hat{\mathbf{x}}}'}{\boldsymbol{\varSigma }^{1}}|_{\hat{\mathbf{x}}'}\) where \({{\boldsymbol{\varSigma }}^{1}}|_{\hat{\mathbf{x}}'}\) is the 2x2 symmetric restriction of \({\boldsymbol{\varSigma }}^{1}\) to the indices \((x,y)\) and that \({\boldsymbol{\nabla}}_{\hat{\mathbf{x}}'}{\boldsymbol{a}}\) is antisymmetric. Since \({{\boldsymbol{\varSigma }}^{1}}|_{\hat{\mathbf{x}}'}{\mathbf{n}}={{\boldsymbol{\varSigma }}^{1}}{\mathbf{n}}={ \mathbf{0}}\) over \(\partial \hat{\mathsf{s}}_{\text{\tiny b}}\), we deduce that \(\int _{\partial \hat{\mathsf{s}}_{\text{\tiny b}}}{\boldsymbol{a}}\cdot ({ \boldsymbol{\varSigma }^{1}}{\mathbf{n}})\,\text{d}\hat{\mathbf{x}}'=0\). Now, recalling the definition (61) of \({\mathcal{T}}^{0}\) as well as (62), we find
Thus, \(\varOmega ^{1}\) is independent of \(\tilde{z}\) (a linear dependence w.r.t. \(\tilde{z}\) is prevented by its \(\tilde{z}\)-periodicity), and \({\mathcal{T}}^{0}\) reduces to
1.4 A.4 Proof of \(P_{3} : \varOmega ^{0}=0\) in (15)
Here, we shall see that the equation for torsion (64) can be complemented to get the wave equation for \(\varOmega ^{0}\); in addition that the associated boundary conditions impose \(\varOmega ^{0}=0\). (From (57) we already know that \(\varOmega ^{0}({\mathbf{x}})\) does not depend of \(\tilde{z}\).)
1.4.1 A.4.1 The Wave Equation for Torsion
We start with the derivation of the equation satisfied by the macroscopic rotation \(\varOmega ^{0}\). We use the relation \(\mathrm{div}_{{\hat{\mathbf{x}}}'}{\boldsymbol{\varSigma }^{2}}+\mathrm{div}_{ \tilde{z}}{\boldsymbol{\varSigma }^{1}}+{\mathrm{div}}_{{\mathbf{x}}}{\boldsymbol{\varSigma }^{0}}+ \rho _{\text{\tiny b}}\omega ^{2}{\mathbf{V}}^{0}={\mathbf{0}}\) from \((\text{E}_{a})^{0}\) and \((\text{E}_{z})^{0}\) in (12), that we multiply by \({\boldsymbol{a}}=\hat{y}{\mathbf{e}}_{x}-\hat{x}{\mathbf{e}}_{y}\) and integrate over \((\hat{\mathbf{x}}',\tilde{z})\in \hat{\mathsf{s}}_{\text{\tiny b}}\times (0, \,\tilde{h}_{\text{\tiny F}})\). We get
where we have accounted for \(\int _{\hat{\mathsf{s}}_{\text{\tiny b}}}{\boldsymbol{a}}\cdot \mathrm{div}_{{ \hat{\mathbf{x}}}'}{\boldsymbol{\varSigma }^{2}}\,\text{d}\hat{\mathbf{x}}'=0\) as in (63) and for the \(\tilde{z}\)-periodicity of \({\boldsymbol{\varSigma }}^{1}\). Next, owing to the form of \({\mathbf{V}}^{0}\) in (56) (with \(\varOmega ^{0}=\varOmega ^{0}({\mathbf{x}})\) from (57)) and the form of \({\boldsymbol{\varSigma }}^{0}\) in (59) (with \(\frac{\partial }{\partial \tilde{z}}\varOmega ^{1}=0\) from (64)), we obtain
with \(J_{\text{\tiny b}}\) defined in (62).
1.4.2 A.4.2 The Corresponding Boundary Conditions
To conclude that \(\varOmega ^{0}=0\), we have to specify the boundary conditions at \(z=0\) and \(h_{\text{\tiny b}}\) by analyzing the torsional behavior near the extremities.
At \(z=h_{\text{\tiny b}}\)
In the intermediate region located near \(z=h_{\text{\tiny b}}\), we introduce the torsion force \({\widetilde{{\mathcal{T}}}^{0}}_{|\,h_{\text{\tiny b}}}({\mathbf{x}}', \tilde{z})\). Repeating the same procedure than in A.3 in the region of the beam, we find that
hence \({\widetilde{{\mathcal{T}}}^{0}}_{|\,h_{\text{\tiny b}}}({\mathbf{x}}')\) does not depend on \(\tilde{z}\). To determine its value, we go to the microscopic scale near \(z=h_{\text{\tiny b}}\) with the expansions (27). Multiplying \(\mathrm{div}_{\hat{\mathbf{x}}} \widehat{\boldsymbol{\varSigma }}^{0}|_{h_{\text{\tiny b}}}={\mathbf{0}}\) by \({\boldsymbol{a}}=\hat{y}{\mathbf{e}}_{x}-\hat{x}{\mathbf{e}}_{y}\), integrating over \(\widehat{\mathsf{X}}_{h_{\text{\tiny b}}}=\left \{ \hat{\mathbf{x}}'\in \hat{\mathsf{s}}_{\text{\tiny b}},\hat{z}\in (-\infty ,0)\right \} \) and making use of the matching conditions and traction free boundary conditions, we deduce that
Finally, using the matching condition between the intermediate region and the inner region of the beams we get
At \(z=0\)
In the intermediate region located near \(z=0\), in virtue of (13) which holds also for the intermediate expansions (23), we have \(\varepsilon ^{\hat{\mathbf{x}}'}\big ({\widetilde{{\mathbf{V}}}^{0}}_{|0}\big )={ \mathbf{0}}\), hence \({\widetilde{V_{x}^{0}}}_{|0}={\widetilde{U_{x}^{0}}}_{|0}({\mathbf{x}}', \tilde{z})+{\widetilde{\varOmega }^{0}}_{|0}({\mathbf{x}}')\hat{y}\). The goal is to show that \({\widetilde{\varOmega }^{0}}_{|0}({\mathbf{x}}')=0\). For that, we go at the microscopic scale near \(z=0\), where we recall (see section (3.2.2)) that since \(\varepsilon ^{\hat{\mathbf{x}}'}\big ({\widehat{\mathbf{V}}^{0}}_{|0})={\mathbf{0}}\), \({\widehat{\mathbf{V}}^{0}}_{|0}\) is a rigid body motion which can be reduced to a translation thanks to the periodic conditions acting at this scale. Therefore we get \({\widehat{V^{0}_{x}}}_{|0}={\widehat{U^{0}_{x}}}_{|0}({\mathbf{x}}')\). It now suffices to apply the micro-meso matching conditions, i.e., \({\widetilde{V^{0}_{x}}}_{|0}({\mathbf{x}}',0)=\lim _{\hat{z}\rightarrow + \infty } {\widehat{V^{0}_{x}}}_{|0}\), to show that \({\widetilde{\varOmega }^{0}}_{|0}({\mathbf{x}}')=0\). Finally, using again matching conditions between the intermediate region and the inner region of the beams we deduce that
Given the wave equation (65) and the corresponding boundary conditions (68)-(69), we deduce that outside the set of torsional resonant frequencies which correspond to \(\omega _{n}=\sqrt{\frac{\mu _{\text{\tiny b}}}{\rho _{\text{\tiny b}}}} \frac{n\pi }{2h_{\text{\tiny b}}}\) with \(n\in {Z}^{*}\), we have
1.5 A.5 Proof of \(P_{4} : \frac{\partial U_{x}^{0}}{\partial z}=\frac{\partial U_{y}^{0}}{\partial z}=0\) in (15)
1.5.1 A.5.1 Determination of \({\mathbf{V}}^{3}\) and \({\boldsymbol{\varSigma }}^{1}\)
The first step is to solve the problem on \(({\boldsymbol{\varSigma }}^{1},{\boldsymbol{V}}^{3})\) which reads, from \((\text{E}_{a})^{-1}\) and \((\text{E}_{z})^{-1}\) in (12) and from \((\text{C'})^{1}\) in (13),
We short-circuit the proof by assuming, or anticipating, that
which can be verified a posteriori. As a result, from (71) and with \(\text{tr}\big ({\boldsymbol{\varSigma }}^{1}\big )=\varSigma _{zz}^{1}\), we have \(\varSigma _{zz}^{1}=E_{\text{\tiny b}}\left ( \frac{\partial V_{z}^{2}}{\partial \tilde{z}}+ \frac{\partial V_{z}^{1}}{\partial z}\right )\). Next, using \(V_{z}^{2}\) in (58), \(V_{z}^{1}\) in (57) and with \(V_{z}^{0}= U_{z}^{0}({\mathbf{x}})\) independent of \(\tilde{z}\) from (56), we obtain
Let us consider now in (71) (second line) the strains associated to the 2D vanishing stress components \((\varSigma _{xx}^{1},\varSigma _{xy}^{1},\varSigma _{yy}^{1})\). We obtain
and
We find that there exists \((U_{x}^{3}({\mathbf{x}},\tilde{z}),U_{y}^{3}({\mathbf{x}},\tilde{z}),\varOmega ^{3}({ \mathbf{x}},\tilde{z}))\) such that the displacements \(( V_{x}^{3}, V_{y}^{3})\) read
The remaining part of the solution, \((\varSigma _{xz}^{1},\varSigma _{yz}^{1})\) and \(V_{z}^{3}\), is more demanding. From (71) with \(\varSigma _{zz}^{0}\) in (59) and using that \(\frac{\partial }{\partial \tilde{z}}U_{z}^{0}= \frac{\partial }{\partial \tilde{z}}U_{z}^{1}=0\) from (56) and (60), these fields satisfy
(We have used the constitutive equation \((\text{C})^{1}\) instead of \((\text{C'})^{1}\) in the second line.) Now, we rearrange the constitutive behavior using (i) \((V_{x}^{2},V_{y}^{2})\) in (58) (with \(\varOmega ^{0}=0\) from (70)), (ii) \((V_{x}^{1},V_{y}^{1},V_{z}^{1})\) in (56)-(57) and (iii) \(\frac{\partial }{\partial \tilde{z}}U_{z}^{0}=0\) from 56, resulting in
where
Basically, the idea is to work on the auxiliary displacement variable \(V'\) which has absorbed \(V_{z}'\) associated to a gradient. Thus, it is now sufficient to solve the problem on \(({\boldsymbol{\varSigma }}', V')\) which satisfy (75) along with \(\mathrm{div}_{{\hat{\mathbf{x}}}'}{\boldsymbol{\varSigma }}'=E_{\text{\tiny b}} \frac{\partial ^{3} }{\partial \tilde{z}^{3} }\left (U_{x}^{0}\hat{x}+U_{y}^{0} \hat{y}\right )\) and \({\boldsymbol{\varSigma }}'{\boldsymbol{n}}=0\); by linearity, we can set
and solve
The above problems can be solved in polar coordinates with \(\hat{x}=\hat{r}\cos \theta \), \(\hat{y}=\hat{r}\sin \theta \) by looking for a solution of the form \(v^{x}= g(\hat{r})\cos \theta \) and \(v^{y}= g(\hat{r})\sin \theta \), with \(g\) satisfying \(g''+\frac{g'}{\hat{r}}-\frac{g}{\hat{r}^{2}}=2\hat{r}\), along with the boundary condition \((2 g'(\hat{r}_{\text{\tiny b}})+\nu _{\text{\tiny b}}\hat{r}_{\text{\tiny b}}^{2})=0\). It follows that
Eventually the solution reads
with \(\hat{r}=\sqrt{\hat{x}^{2}+\hat{y}^{2}}\) and with \((V'_{z},S)\) defined in (76).
1.5.2 A.5.2 Macroscopic Equation for the Flexural Motions \(U_{a}^{1}\)
Now, we want to derive the relations equivalent to (19), (20) and (21). To do so, we iterate the procedure conducted in §3.1.5. First, we obtain \(M^{1}_{a}\), \(a=x,y\), by integrating \(\varSigma _{zz}^{1}\) (whose form is in (73)) over \(\hat{\mathbf{x}}'\in \hat{\mathsf{s}}_{\text{\tiny b}}\) after multiplication by \(\hat{x}\) then by \(\hat{y}\). Next, the equilibrium is obtained starting with \((\text{E}_{a})^{1}\) in (12): \(\frac{\partial }{\partial \hat{x}}\varSigma _{xa}^{3}+ \frac{\partial }{\partial \hat{y}}\varSigma _{ya}^{3}+ \frac{\partial }{\partial \tilde{z}}\varSigma _{az}^{2}+ \frac{\partial }{\partial z}\varSigma _{az}^{1}+\rho _{\text{\tiny b}} \omega ^{2} V_{a}^{1}=0\), \(a=x,y\) (since \(\varSigma ^{1}_{xa}=\varSigma ^{1}_{ya}=0\) from (72)), that we integrate over \(\hat{\mathbf{x}}'\in \hat{\mathsf{s}}_{\text{\tiny b}}\); then we use \(\varSigma _{az}^{1}\) in (77) and \(V^{1}_{a}\) in (56). Eventually, the relations between bending moments and shear forces are written from \((\text{E}_{z})^{0}\) in (12): \(\frac{\partial }{\partial \hat{x}}\varSigma _{xz}^{2}+ \frac{\partial }{\partial \hat{y}}\varSigma _{yz}^{2}+ \frac{\partial }{\partial \tilde{z}}\varSigma _{zz}^{1}+ \frac{\partial }{\partial z}\varSigma _{zz}^{0}+\rho _{\text{\tiny b}} \omega ^{2} V_{z}^{0}=0\) (since \(\varSigma _{xz}^{0}=\varSigma _{yz}^{0}=0\) from (16)), that we integrate over \(\hat{\mathbf{x}}'\in \hat{\mathsf{s}}_{\text{\tiny b}}\) after multiplication by \(\hat{a}=\hat{x},\hat{y}\); then, we use \(\varSigma _{zz}^{0}\) in (16) and \(V_{z}^{0}\) in (56). We obtain
with \(A_{\text{\tiny b}}=-\frac{1}{2}(1+\nu _{\text{\tiny b}})\pi r_{\text{\tiny b}}^{4}\). The above system leads to
with \(\kappa ^{4}= \frac{\rho _{\text{\tiny b}}\varphi }{E_{\text{\tiny b}}I_{\text{\tiny b}}} \omega ^{2}\) and \(B_{\text{\tiny b}}= \frac{\mu _{\text{\tiny b}}A_{\text{\tiny b}}}{E_{\text{\tiny b}}I_{\text{\tiny b}}}-3=-4\). If \(U^{0}_{a}\) depends on \(z\) (and we want to show that it is not the case), then \(U^{0}_{a}\) solution of (22) reads \(U^{0}_{a}({\mathbf{x}},\tilde{z})=A_{a}({\mathbf{x}})\sin \kappa \tilde{z}+B_{a}({ \mathbf{x}})\cos \kappa \tilde{z}\). Integrating (89) over \(\tilde{z}\in (0,2\pi /\kappa )\) after multiplication by \(\cos \kappa \tilde{z}\) (resp. by \(\sin \kappa \tilde{z}\)) and integrating by parts four times provides \(\frac{\partial }{\partial z}A_{a}=0\) (resp. \(\frac{\partial }{\partial z}B_{a}=0\)), which allows us to conclude that
Appendix B: The Case of a Two-Dimensional Array of Plates
We provide in this section the effective model for the alternative situation of a periodic array of identical plates supported by a soil substrate. We transpose the notations of the beams case, assuming that the plates have a height \(h_{\text{\tiny b}}\), a thickness \(2r_{\text{\tiny b}}\) and that the periodicity of the array is \(\ell \). We denote \((\lambda _{\text{\tiny b}},\mu _{\text{\tiny b}},\rho _{\text{\tiny b}})\) the material parameters of the plates and \((\lambda _{\text{\tiny s}},\mu _{\text{\tiny s}},\rho _{\text{\tiny s}})\) those of the substrate. The same scaling as for the beams case is used, see (7). For in-plane incident wave, the resulting displacement fields remain in-plane with
Following the same asymptotic procedure as for the beams, we derive an effective model for the array of plates. We provide below the main results of the derivation. With \({\mathbf{x}}=(x,z)\), the equivalent of (14) in the region of the plates and far from their extremities reads
see also forthcoming Remark (87). From \((\text{E}_{a})^{-2}\), \((\text{E}_{z})^{-2}\) and from \((\text{C})^{0}\) in (12), we have \(\frac{\partial }{\partial \hat{x}}\varSigma _{xx}^{0}= \frac{\partial }{\partial \hat{x}}\varSigma _{xz}^{0}=0\) with \({\boldsymbol{\varSigma }}^{0}={\mathsf{A}}\,{\boldsymbol{\varepsilon }}^{\hat{x}}\big ({\mathbf{V}}^{2} \big )+{\mathsf{A}}\,{\boldsymbol{\varepsilon }}^{\tilde{z}}\big ({\mathbf{V}}^{1}\big )+{ \mathsf{A}}\,{\boldsymbol{\varepsilon }}^{{\mathbf{x}}}\big ({\mathbf{V}}^{0}\big )\), and the system on \(({\boldsymbol{\varSigma }}^{0},{\mathbf{V}}^{2})\) has again explicit solutions which reads
and the corresponding stress tensor \({\boldsymbol{\varSigma }}^{0}\) by
which is the two-dimensional version of (58)-(59). The axial stress is \(N_{zz}^{0}= \int _{\hat{\mathsf{s}}_{\text{\tiny b}}}\varSigma _{zz}^{0} \text{d}\hat{x}\) with now \(\hat{\mathsf{s}}_{\text{\tiny b}}=\{\hat{x}\in \left (-\hat{r}_{\text{\tiny b}},\hat{r}_{\text{\tiny b}}\right )\}\) and the equilibrium is deduced from the two-dimensional version of \((\text{E}_{z})^{0}\) in (12) which reads \(\frac{\partial }{\partial \hat{x}}\varSigma _{xz}^{2}+ \frac{\partial }{\partial \tilde{z}}\varSigma _{zz}^{1}+ \frac{\partial }{\partial z}\varSigma _{zz}^{0}+\rho _{\text{\tiny b}} \omega ^{2}U_{z}^{0}=0\)
being the equivalent of (17). For the flexural motions, (i) we obtain the bending moment \(M_{x}^{0}\) by integrating \(\varSigma _{zz}^{0}\), in (80), over \(\hat{x}\in \hat{\mathsf{s}}_{\text{\tiny b}}\) after multiplication by \(\hat{x}\), (ii) next, from (80), we use that \(\varSigma _{xz}^{0}=0\), (iii) eventually we integrate \(\frac{\partial }{\partial \hat{x}}\varSigma _{xx}^{2}+ \frac{\partial }{\partial \tilde{z}}\varSigma _{xz}^{1}+\rho _{\text{\tiny b}}\omega ^{2} V_{x}^{0}=0\) over \(\hat{x}\in \hat{\mathsf{s}}_{\text{\tiny b}}\). We get
The relation between the bending moment and the shear force is obtained from the two-dimensional version of \((\text{E}_{z})^{-1}\) in (12), which reads \(\frac{\partial }{\partial \hat{x}}\varSigma _{xz}^{1}+ \frac{\partial }{\partial \tilde{z}}\varSigma _{zz}^{0}=0\) that we integrate over \(\hat{x}\in \hat{\mathsf{s}}_{\text{\tiny b}}\) after multiplication by \(\hat{x}\). It results that
as in (21), which along with the above relation provides the wave equation for the flexural motion \(U_{x}\) in the form
At the intermediate scale near the ends, the displacement \(\widetilde{U}{_{x}^{0}}(x,\tilde{z})\) is not \(\tilde{z}\)-periodic anymore. We obtain the same flexural equation as in (82)-(84), namely
with \(\tilde{z}\in (0,+\infty )\) if \(z^{*}=0\) and \(\tilde{z}\in (-\infty ,0)\) if \(z^{*}=h_{\text{\tiny b}}\). When \(|\tilde{z}|\rightarrow \infty \), the fields \(( \widetilde{U}{_{x}^{0}}_{|\,z^{*}},\widetilde{M}{_{x}^{0}}_{|\,z^{*}}, \widetilde{T}{_{x}^{1}}_{|\,z^{*}})\) match their \(\tilde{z}\)-periodic counterparts \(( {U}{_{x}^{0}},{M}{_{x}^{0}}, {T}{_{x}^{1}})\) defined inside the plates. Near \(\tilde{z}=0\), the boundary conditions are obtained in the same way as the matchings developed in §3.2 for beams. Doing so, we obtain
Similarly, we get the boundary conditions for the longitudinal wave equation (81)
As in §3.3, we construct the problem at the dominant order for the displacement \(U_{z}({\mathbf{x}})=U_{z}^{0}({\mathbf{x}})\) and \(U_{x}({\mathbf{x}})=\left (\widetilde{U}{_{x}^{0}}_{|0}(x,\tilde{z}),U_{x}(x, \tilde{z}),\widetilde{U}{_{x}^{0}}_{|\,h_{\text{\tiny b}}}(x,\tilde{z}) \right )\) accounting for \(\tilde{z}=z/\eta \) and \(r_{\text{\tiny b}}=\eta ^{2}\hat{r}_{\text{\tiny b}}\). Gathering the results in the inner region of the plates and the intermediate region near the extremities, we get the final set of homogenized equations for \(z\in (0,h_{\text{\tiny b}})\) as follows. In the substrate, the displacement field is governed by the classical balance of linear momentum equation and constitutive equation of the substrate. The array of the plates is replaced by frequency dependent boundary conditions which encapsulate both the longitudinal and flexural resonant motions of the plates
and
with
Eventually, as in the three-dimensional case, once the effective problem (85) in the substrate has been resolved, the solution in the region of the plates can be post-processed, as \(U_{x}\) and \(U_{z}\) in the region of the plates are solutions to
Solving these two problems for which the inputs are the prescribed displacements \({\mathbf{u}}(x,0)\) at the junction with the substrate gives
Remark 1
The expressions of the displacements in (79) have been obtained using
being the equivalent of (15). A priori we have \(V_{x}^{0}=U_{x}^{0}({\mathbf{x}},\tilde{z})\), \(V_{z}^{0}= U_{z}^{0}({\mathbf{x}})\) and \(V_{x}^{1}=U_{x}^{1}({\mathbf{x}},\tilde{z})\), \(V_{z}^{1}= U_{z}^{1}({\mathbf{x}},\tilde{z})- \frac{\partial }{\partial \tilde{z}}U_{x}^{0}\,\hat{x}\), eventually \(\varSigma _{zz}^{0}= E_{\text{\tiny b}}^{*}\left ({ \frac{\partial U_{z}^{1}}{\partial \tilde{z}}}+ \frac{\partial U_{z}^{0}}{\partial z}- \frac{\partial ^{2} U_{x}^{0}}{\partial \tilde{z}^{2}}\hat{x}\right )\). The proof of \(p_{1}\) is similar to that of \(P_{1}\) provided in A.2 using the two-dimensional version of \((\text{E}_{z})^{-1}\) in (12), \(\frac{\partial }{\partial \hat{x}}\varSigma _{xz}^{1}+ \frac{\partial }{\partial \tilde{z}}\varSigma _{zz}^{0}=0\), that we integrate over \(\hat{x}\in \hat{\mathsf{s}}_{\text{\tiny b}}\) and we get \(\frac{\partial ^{2} U_{z}^{1}}{\partial \tilde{z}^{2}}=0\) hence \(p_{1}\) as a linear dependence of \(U_{z}^{1}\) on \(\tilde{z}\) is prevented by its \(\tilde{z}\)-periodicity. The proof of \(p_{4}\) – We start with the two-dimensional version of (71) (hence \(\mathrm{div}_{{\hat{\mathbf{x}}}'}\to {\mathrm{div}}_{\hat{x}}\) and (\(\varSigma ^{1}_{xx},\varSigma _{xz}^{1},\varSigma _{zz}^{1}\)) given by \((\text{C})^{1}\) in (12)). The problem can be solved on \((\varSigma _{xx}^{1},\varSigma _{zz}^{1},V^{3}_{x})\) then on \((\varSigma ^{1}_{xz},V_{z}^{3})\) resulting in
and
where \(\alpha = \frac{3\lambda _{\text{\tiny b}}+4\mu _{\text{\tiny b}}}{6(\lambda _{\text{\tiny b}}+2\mu _{\text{\tiny b}})}\) and \(\beta = \frac{2(\lambda _{\text{\tiny b}}+\mu _{\text{\tiny b}})}{\lambda _{\text{\tiny b}}+2\mu _{\text{\tiny b}}}\). To conclude the proof of \(P_{4}\), we use the procedure conducted in A.5.2. First, we obtain \(M^{1}_{x}\) by integrating \(\hat{x}\varSigma _{zz}^{1}\) over \(\hat{x}\in \hat{\mathsf{s}}_{\text{\tiny b}}\). Next, the equilibrium is obtained integrating \(\frac{\partial }{\partial \hat{x}}\varSigma _{xx}^{3}+ \frac{\partial }{\partial \tilde{z}}\varSigma _{xz}^{2}+ \frac{\partial }{\partial z}\varSigma _{xz}^{1}+\rho _{\text{\tiny b}} \omega ^{2} V_{x}^{1}=0\) over \(\hat{x}\in \hat{\mathsf{s}}_{\text{\tiny b}}\). Eventually, we use \(\frac{\partial }{\partial \hat{x}}\varSigma _{xz}^{2}+ \frac{\partial }{\partial \tilde{z}}\varSigma _{zz}^{1}+ \frac{\partial }{\partial z}\varSigma _{zz}^{0}+\rho _{\text{\tiny b}} \omega ^{2} V_{z}^{0}=0\) that we integrate over \(\hat{x}\in \hat{\mathsf{s}}_{\text{\tiny b}}\) after multiplication by \(\hat{x}\). We obtain
which leads to
as in (89). The conclusion is identical: \(\frac{\partial U_{x}^{0}}{\partial z}=0\).
Rights and permissions
About this article
Cite this article
Marigo, JJ., Pham, K., Maurel, A. et al. Effective Model for Elastic Waves in a Substrate Supporting an Array of Plates/Beams with Flexural and Longitudinal Resonances. J Elast 146, 143–177 (2021). https://doi.org/10.1007/s10659-021-09854-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10659-021-09854-4