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

Dissipative structure and diffusion phenomena for doubly dissipative elastic waves in two space dimensions

Wenhui Chen Institute of Applied Analysis, Faculty for Mathematics and Computer Science
Technical University Bergakademie Freiberg
Prüferstraße 9
09596 Freiberg
Germany
wenhui.chen.math@gmail.com
(Date: January 1, 2004)
Abstract.

In this paper we study the Cauchy problem for doubly dissipative elastic waves in two space dimensions, where the damping terms consist of two different friction or structural damping. We derive energy estimates and diffusion phenomena with different assumptions on initial data. Particularly, we find the dominant influence on diffusion phenomena by introducing a new threshold of diffusion structure.

Key words and phrases:
Dissipative elastic waves, friction, structural damping, energy estiamte, diffusion phenomenon.
1991 Mathematics Subject Classification:
Primary 35B40; Secondary 35L15

1. Introduction

In this paper we consider the following Cauchy problem for doubly dissipative elastic waves in two space dimensions:

{utta2Δu(b2a2)divu+(Δ)ρut+(Δ)θut=0,x2,t>0,(u,ut)(0,x)=(u0,u1)(x),x2,\left\{\begin{aligned} &u_{tt}-a^{2}\Delta u-\left(b^{2}-a^{2}\right)\nabla\operatorname{div}u+(-\Delta)^{\rho}u_{t}+(-\Delta)^{\theta}u_{t}=0,&&x\in\mathbb{R}^{2},\,\,t>0,\\ &(u,u_{t})(0,x)=(u_{0},u_{1})(x),&&x\in\mathbb{R}^{2},\end{aligned}\right. (1.1)

where the unknown u=u(t,x)2u=u(t,x)\in\mathbb{R}^{2} denotes the elastic displacement. The positive constants aa and bb in (1.1) are related to the Lamé constants and fulfill b>a>0b>a>0. Moreover, the parameters ρ\rho and θ\theta in (1.1) satisfy 0ρ<1/2<θ10\leq\rho<1/2<\theta\leq 1.

Let us recall some related works to our problem (1.1). Taking a=b=1a=b=1, ρ=0\rho=0 and θ=1\theta=1 in (1.1), then we immediately turn to doubly dissipative wave equation, where the damping terms consist of friction utu_{t} as well as viscoelastic damping Δut-\Delta u_{t}

{uttΔu+utΔut=0,xn,t>0,(u,ut)(0,x)=(u0,u1)(x),xn,\left\{\begin{aligned} &u_{tt}-\Delta u+u_{t}-\Delta u_{t}=0,&&x\in\mathbb{R}^{n},\,\,t>0,\\ &(u,u_{t})(0,x)=(u_{0},u_{1})(x),&&x\in\mathbb{R}^{n},\end{aligned}\right. (1.2)

with n1n\geq 1. The recent paper [15] derived asymptotic profiles of solutions to (1.2) in a framework of weighted L1L^{1} data. Precisely, the authors found that from asymptotic profiles of solutions point of view, friction utu_{t} is more dominant than viscoelastic damping Δut-\Delta u_{t} as tt\rightarrow\infty. Later in [13], the authors obtained higher-order asymptotic expansions of solutions to (1.2) and gave some lower bounds estimates to show the optimality of these expansions. For the other related works on (1.2), we refer the reader to the recent papers [16, 6, 18]. However, asymptotic profiles of solutions to general doubly dissipative wave equation, where the damping terms consist of friction or structural damping (i.e., taking a=b=1a=b=1 in (1.1)), are still open. This open problem is proposed in [15]. The main difficulty is to answer what is the dominant profile of solutions, due to the fact that the asymptotic profiles for wave equation with damping term (Δ)ρut(-\Delta)^{\rho}u_{t} for 0ρ<1/20\leq\rho<1/2, or with damping term (Δ)θut(-\Delta)^{\theta}u_{t} for 1/2<θ11/2<\theta\leq 1, are quite different. One may see, for example, [24, 22, 23, 9, 27, 28, 34, 11, 7, 14, 25, 26, 17, 33, 29, 8].

Let us come back to dissipative elastic waves. In recent years the Cauchy problem for dissipative elastic waves have aroused wide concern, which can be modeled by

{utta2Δu(b2a2)divu+𝒜ut=0,xn,t>0,(u,ut)(0,x)=(u0,u1)(x),xn,\left\{\begin{aligned} &u_{tt}-a^{2}\Delta u-\left(b^{2}-a^{2}\right)\nabla\operatorname{div}u+\mathcal{A}u_{t}=0,&&x\in\mathbb{R}^{n},\,\,t>0,\\ &(u,u_{t})(0,x)=(u_{0},u_{1})(x),&&x\in\mathbb{R}^{n},\end{aligned}\right. (1.3)

where b>a>0b>a>0 and the term 𝒜ut\mathcal{A}u_{t} describes several kinds of damping mechanisms.
In the case when

𝒜ut=ut,i.e., friction or external damping,\displaystyle\mathcal{A}u_{t}=u_{t},\,\,\,\,\text{i.e., \emph{friction} or \emph{external damping}},

the authors of [12] proved almost sharp energy estimates for n2n\geq 2 by using energy methods in the Fourier space and the Haraux-Komornik inequality, and then the recent paper [5] investigated propagation of singularities, sharp energy estimates and diffusion phenomenon for n=3n=3.
Furthermore, in the case when

𝒜ut=(Δ)θutwithθ(0,1],i.e., structural damping,\displaystyle\mathcal{A}u_{t}=(-\Delta)^{\theta}u_{t}\,\,\,\,\text{with}\,\,\,\,\theta\in(0,1],\,\,\,\,\text{i.e., \emph{structural damping}},

energy estimates are derived with different data spaces in [12] for n2n\geq 2, and in [30] for n=2n=2. Moreover, some qualitative properties of solutions, including smoothing effect, sharp energy estimate and diffusion phenomena (especially, double diffusion phenomena when θ(0,1/2)\theta\in(0,1/2)) are obtained for n=3n=3.
Finally, in the case when

𝒜ut=(a2Δ(b2a2)div)ut,i.e., Kelvin-Voigt damping,\displaystyle\mathcal{A}u_{t}=(-a^{2}\Delta-(b^{2}-a^{2})\nabla\operatorname{div})u_{t},\,\,\,\,\text{i.e., \emph{Kelvin-Voigt damping}},

by applying energy methods in the Fourier space, almost sharp energy estimates for n2n\geq 2 have been obtained in [35]. Then, sharp energy estimates, LpLqL^{p}-L^{q} estimates as well as asymptotic profiles of solutions are derived for n=2n=2 in [3]. Other studies on dissipative elastic waves can be found in literatures [1, 2]. Nevertheless, concerning about decay properties and diffusion phenomena for the Cauchy problem for doubly dissipative elastic waves it seems that we still do not have any previous research manuscripts. Moreover, this problem is strongly related to the open problem proposed in [15]. In this paper we give the answer to the two-dimensional case.

Let us point out that the study of the Cauchy problem (1.1) is not simply a generalization of elastic waves with friction or structural damping in [30, 5]. On one hand, because there exists two different damping terms (Δ)ρut(-\Delta)^{\rho}u_{t} and (Δ)θut(-\Delta)^{\theta}u_{t} with 0ρ<1/2<θ10\leq\rho<1/2<\theta\leq 1 in our problem (1.1), it is not clear which damping term has a dominant influence on dissipative structure. On the other hand, from the paper [5], the authors derived diffusion phenomena to elastic waves with the damping term (Δ)θut(-\Delta)^{\theta}u_{t} where θ[0,1/2)(1/2,1]\theta\in[0,1/2)\cup(1/2,1], which are described by the following so-called reference system.

  • In the case when θ=0\theta=0, the reference system consist of heat-type system with mass term as follows:

    V~t𝒟1ΔV~+𝒟2V~=0,\displaystyle\widetilde{V}_{t}-\mathcal{D}_{1}\Delta\widetilde{V}+\mathcal{D}_{2}\widetilde{V}=0,

    with real diagonal matrices 𝒟1\mathcal{D}_{1} and 𝒟2\mathcal{D}_{2}.

  • In the case when θ(0,1/2)\theta\in(0,1/2), the reference system consist of two different parabolic systems as follows:

    V~t+𝒟3(Δ)1θV~+𝒟4(Δ)θV~=0,\displaystyle\widetilde{V}_{t}+\mathcal{D}_{3}(-\Delta)^{1-\theta}\widetilde{V}+\mathcal{D}_{4}(-\Delta)^{\theta}\widetilde{V}=0,

    with real diagonal matrices 𝒟3\mathcal{D}_{3} and 𝒟4\mathcal{D}_{4}.

  • In the case when θ(1/2,1]\theta\in(1/2,1], the reference system consist of parabolic system and half-wave system as follows:

    V~t+𝒟5(Δ)θV~+i𝒟6(Δ)12V~=0,\displaystyle\widetilde{V}_{t}+\mathcal{D}_{5}(-\Delta)^{\theta}\widetilde{V}+i\mathcal{D}_{6}(-\Delta)^{\frac{1}{2}}\widetilde{V}=0,

    with real diagonal matrices 𝒟5\mathcal{D}_{5} and 𝒟6\mathcal{D}_{6}.

Hence, for different choices of damping terms, which mainly depend on the value of the parameter θ\theta in the damping term, the diffusion phenomena are quite different. In the Cauchy problem (1.1), the damping terms consist of (Δ)ρut(-\Delta)^{\rho}u_{t} with ρ[0,1/2)\rho\in[0,1/2), and (Δ)θut(-\Delta)^{\theta}u_{t} with θ(1/2,1]\theta\in(1/2,1]. Thus, it is not clear that the reference system is make up of what kind of evolution systems, and how do two different damping terms influence on diffusion structure. Furthermore, from [5] we know the threshold of diffusion structure is θ=1/2\theta=1/2 for elastic waves with structural damping. In other words, the structure of reference system will be changed from θ(0,1/2)\theta\in(0,1/2) to θ(1/2,1]\theta\in(1/2,1]. Then, the natural question is what is the threshold of diffusion structure for doubly dissipative elastic waves. Again, we give the answers for these questions in two dimensions.

Our main purpose of the present paper is to investigate dissipative structure and diffusion phenomena for doubly dissipative elastic waves with different assumptions on initial data. We find that the damping term (Δ)ρut(-\Delta)^{\rho}u_{t} with 0ρ<1/20\leq\rho<1/2 has the dominant influence on energy estimates (see Theorems 3.1 and 3.2). Furthermore, in the case when ρ+θ<1\rho+\theta<1, the damping terms (Δ)ρut(-\Delta)^{\rho}u_{t} and (Δ)θut(-\Delta)^{\theta}u_{t} with 0ρ<1/2<θ10\leq\rho<1/2<\theta\leq 1 have the influence on diffusion structure at the same time. However, in the case when ρ+θ1\rho+\theta\geq 1, the diffusion structure is determined by the damping term (Δ)ρut(-\Delta)^{\rho}u_{t} with 0ρ<1/20\leq\rho<1/2 only. Hence, one of our novelties is to derive a threshold ρ+θ=1\rho+\theta=1 of diffusion structure for doubly dissipative elastic waves.

This paper is organized as follows. In Section 2 we derive representation of solutions by applying WKB analysis and multistep diagonalization procedure. In Section 3 we obtain pointwise estimate in the Fourier space and energy estimates by using this representation. In Section 4 we derive diffusion phenomena with different assumptions on initial data. Finally, in Section 5 some concluding remarks complete the paper.


Notations: In this paper fgf\lesssim g means that there exists a positive constant CC such that fCgf\leq Cg. We write fgf\asymp g when gfgg\lesssim f\lesssim g Moreover, HsH^{s} and H˙s\dot{H}^{s} with s0s\geq 0, denote Bessel and Riesz potential spaces based on L2L^{2}, respectively. Furthermore, Ds\langle D\rangle^{s} and |D|s|D|^{s} stand for the pseudo-differential operators with symbols ξs\langle\xi\rangle^{s} and |ξ|s|\xi|^{s}, respectively, where ξ2:=1+|ξ|2\langle\xi\rangle^{2}:=1+|\xi|^{2}. We denote the identity matrix of dimensions k×kk\times k by Ik×kI_{k\times k}. We denote the diagonal matrix by

diag(eλjt)j=14:=diag(eλ1t,eλ2t,eλ3t,eλ4t).\displaystyle\operatorname{diag}\left(e^{-\lambda_{j}t}\right)_{j=1}^{4}:=\operatorname{diag}\left(e^{-\lambda_{1}t},e^{-\lambda_{2}t},e^{-\lambda_{3}t},e^{-\lambda_{4}t}\right).

The weighted spaces L1,γL^{1,\gamma} for γ[0,)\gamma\in[0,\infty) are defined by

L1,γ:={fL1:fL1,γ:=n(1+|x|)γ|f(x)|𝑑x<}.\displaystyle L^{1,\gamma}:=\left\{f\in L^{1}:\|f\|_{L^{1,\gamma}}:=\int_{\mathbb{R}^{n}}(1+|x|)^{\gamma}|f(x)|dx<\infty\right\}.

Finally, let us define the cut-off functions χint(ξ),χbdd(ξ),χext(ξ)𝒞\chi_{\operatorname{int}}(\xi),\chi_{\text{bdd}}(\xi),\chi_{\operatorname{ext}}(\xi)\in\mathcal{C}^{\infty} having their supports in the following zones:

𝒵int(ε)\displaystyle\mathcal{Z}_{\operatorname{int}}(\varepsilon) :={ξ2:|ξ|<ε1},\displaystyle:=\left\{\xi\in\mathbb{R}^{2}:|\xi|<\varepsilon\ll 1\right\},
𝒵bdd(ε,N)\displaystyle\mathcal{Z}_{\text{bdd}}(\varepsilon,N) :={ξ2:ε|ξ|N},\displaystyle:=\left\{\xi\in\mathbb{R}^{2}:\varepsilon\leq|\xi|\leq N\right\},
𝒵ext(N)\displaystyle\mathcal{Z}_{\operatorname{ext}}(N) :={ξ2:|ξ|>N1},\displaystyle:=\left\{\xi\in\mathbb{R}^{2}:|\xi|>N\gg 1\right\},

respectively, so that χint(ξ)+χbdd(ξ)+χext(ξ)=1\chi_{\operatorname{int}}(\xi)+\chi_{\text{bdd}}(\xi)+\chi_{\operatorname{ext}}(\xi)=1.

2. Asymptotic behavior of solutions in the Fourier space

In this section we will derive asymptotic behavior of solutions and representation of solutions in the Fourier space. Let us apply the partial Fourier transform with respect to spatial variable such that u^(t,ξ)=xξ(u(t,x))\hat{u}(t,\xi)=\mathcal{F}_{x\rightarrow\xi}(u(t,x)) to obtain

{u^tt+|ξ|2A(η)u^+(|ξ|2ρ+|ξ|2θ)u^t=0,ξ2,t>0,(u^,u^t)(0,ξ)=(u^0,u^1)(ξ),ξ2,\left\{\begin{aligned} &\hat{u}_{tt}+|\xi|^{2}A(\eta)\hat{u}+\left(|\xi|^{2\rho}+|\xi|^{2\theta}\right)\hat{u}_{t}=0,&&\xi\in\mathbb{R}^{2},\,\,t>0,\\ &(\hat{u},\hat{u}_{t})(0,\xi)=(\hat{u}_{0},\hat{u}_{1})(\xi),&&\xi\in\mathbb{R}^{2},\end{aligned}\right. (2.4)

where

A(η)=(a2+(b2a2)η12(b2a2)η1η2(b2a2)η1η2a2+(b2a2)η22)\displaystyle A(\eta)=\left({\begin{array}[]{*{20}c}a^{2}+\left(b^{2}-a^{2}\right)\eta_{1}^{2}&\left(b^{2}-a^{2}\right)\eta_{1}\eta_{2}\\ \left(b^{2}-a^{2}\right)\eta_{1}\eta_{2}&a^{2}+\left(b^{2}-a^{2}\right)\eta_{2}^{2}\\ \end{array}}\right)

with η=ξ/|ξ|𝕊1\eta=\xi/|\xi|\in\mathbb{S}^{1}. Similar as [30, 3], we introduce the matrix

M(η):=(η1η2η2η1),\begin{split}M(\eta):=\left({\begin{array}[]{*{20}c}\eta_{1}&\eta_{2}\\ \eta_{2}&-\eta_{1}\\ \end{array}}\right),\end{split}

and define a new variable W=W(t,ξ)W=W(t,\xi) such that

W:=(vt+i|ξ|diag(b,a)vvti|ξ|diag(b,a)v),W:=\left(\begin{aligned} &v_{t}+i|\xi|\text{diag}(b,a)v\\ &v_{t}-i|\xi|\text{diag}(b,a)v\end{aligned}\right),

where v:=M1(η)u^v:={M}^{-1}(\eta)\hat{u}. Moreover, we have U(t,x)=ξx1(W(t,ξ))U(t,x)=\mathcal{F}^{-1}_{\xi\rightarrow x}(W(t,\xi)). Next, the following first-order system can be derived:

{Wt+(12B0|ξ|2ρ+iB1|ξ|+12B0|ξ|2θ)W=0,ξ2,t>0,W(0,ξ)=W0(ξ),ξ2,\left\{\begin{aligned} &W_{t}+\left(\frac{1}{2}{B}_{0}|\xi|^{2\rho}+i{B}_{1}|\xi|+\frac{1}{2}{B}_{0}|\xi|^{2\theta}\right)W=0,&&\xi\in\mathbb{R}^{2},\,\,t>0,\\ &W(0,\xi)=W_{0}(\xi),&&\xi\in\mathbb{R}^{2},\end{aligned}\right. (2.5)

where the coefficient matrices B0{B}_{0} and B1{B}_{1} are respectively given by

B0=(1010010110100101)andB1=(b0000a0000b0000a).\begin{split}{B}_{0}=\left({\begin{array}[]{*{20}c}1&0&1&0\\ 0&1&0&1\\ 1&0&1&0\\ 0&1&0&1\\ \end{array}}\right)\,\,\,\,\text{and}\,\,\,\,{B}_{1}=\left({\begin{array}[]{*{20}c}-b&0&0&0\\ 0&-a&0&0\\ 0&0&b&0\\ 0&0&0&a\\ \end{array}}\right).\end{split} (2.6)

Let us point out that throughout this section, we will study representation of solutions U=U(t,x)U=U(t,x) to the following Cauchy problem by deriving representation of its partial Fourier transform W(t,ξ)=xξ(U(t,x))W(t,\xi)=\mathcal{F}_{x\rightarrow\xi}(U(t,x)):

{Ut+12B0(Δ)ρU+iB1(Δ)12U+12B0(Δ)θU=0,x2,t>0,U(0,x)=U0(x),x2,\left\{\begin{aligned} &U_{t}+\frac{1}{2}{B}_{0}(-\Delta)^{\rho}U+i{B}_{1}(-\Delta)^{\frac{1}{2}}U+\frac{1}{2}{B}_{0}(-\Delta)^{\theta}U=0,&&x\in\mathbb{R}^{2},\,\,t>0,\\ &U(0,x)=U_{0}(x),&&x\in\mathbb{R}^{2},\end{aligned}\right. (2.7)

where the coefficient matrices B0B_{0} and B1B_{1} are given in (2.6). Moreover, to derive qualitative properties of solutions to (1.1), we only need to study the solutions to (2.7).

With the aim of deriving representation of solutions, we may apply WKB analysis and multistep diagonalization procedure (see for example [31, 36, 20, 19, 21, 30, 4]). Before doing these, we should understand the influence of the parameter |ξ||\xi| on the asymptotic behavior of solutions to (2.5). Due to our assumption 02ρ<1<2θ20\leq 2\rho<1<2\theta\leq 2, we now discuss the influence of |ξ||\xi| by three parts. Specifically, we will apply diagonalization procedure for small frequencies ξ𝒵int(ε)\xi\in\mathcal{Z}_{\operatorname{int}}(\varepsilon) and large frequencies ξ𝒵ext(N)\xi\in\mathcal{Z}_{\operatorname{ext}}(N) in Subsections 2.1 and 2.2, respectively. Then, the contradiction argument will be applied to prove an exponential stability of solutions for bounded frequencies ξ𝒵bdd(ε,N)\xi\in\mathcal{Z}_{\text{bdd}}(\varepsilon,N) in Subsection 2.3.

2.1. Treatment for small frequencies

In the case when ξ𝒵int(ε)\xi\in\mathcal{Z}_{\operatorname{int}}(\varepsilon), it is clear that the matrix 12|ξ|2ρB0\frac{1}{2}|\xi|^{2\rho}B_{0} has a dominant influence comparing with the matrices i|ξ|B1i|\xi|B_{1} and 12|ξ|2θB0\frac{1}{2}|\xi|^{2\theta}B_{0}. For this reason, by defining

T1:=(1010010110100101),\displaystyle T_{1}:=\left({\begin{array}[]{*{20}c}-1&0&1&0\\ 0&-1&0&1\\ 1&0&1&0\\ 0&1&0&1\\ \end{array}}\right), (2.12)

we introduce W(1):=T11WW^{(1)}:=T_{1}^{-1}W. Then, we may derive

Wt(1)+Λ1(|ξ|)W(1)+(B0(1)(|ξ|)+B1(1)(|ξ|))W(1)=0,\displaystyle W^{(1)}_{t}+\Lambda_{1}(|\xi|)W^{(1)}+\left(B_{0}^{(1)}(|\xi|)+B_{1}^{(1)}(|\xi|)\right)W^{(1)}=0,

where

Λ1(|ξ|)\displaystyle\Lambda_{1}(|\xi|) =|ξ|2ρdiag(0,0,1,1)=𝒪(|ξ|2ρ),\displaystyle=|\xi|^{2\rho}\operatorname{diag}(0,0,1,1)=\mathcal{O}\left(|\xi|^{2\rho}\right),
B0(1)(|ξ|)\displaystyle B_{0}^{(1)}(|\xi|) =|ξ|2θdiag(0,0,1,1)=𝒪(|ξ|2θ),\displaystyle=|\xi|^{2\theta}\operatorname{diag}(0,0,1,1)=\mathcal{O}\left(|\xi|^{2\theta}\right),
B1(1)(|ξ|)\displaystyle B_{1}^{(1)}(|\xi|) =i|ξ|T11B1T1=𝒪(|ξ|).\displaystyle=i|\xi|T_{1}^{-1}B_{1}T_{1}=\mathcal{O}(|\xi|).

Here

T11B1T1=(00b0000ab0000a00).\displaystyle T_{1}^{-1}B_{1}T_{1}=\left({\begin{array}[]{*{20}c}0&0&b&0\\ 0&0&0&a\\ b&0&0&0\\ 0&a&0&0\\ \end{array}}\right).

In the second step we introduce W(2):=T21W(1)W^{(2)}:=T_{2}^{-1}W^{(1)}, where

T2:=I4×4+N2(|ξ|)withN2(|ξ|):=i|ξ|12ρ(00b0000ab0000a00).\displaystyle T_{2}:=I_{4\times 4}+N_{2}(|\xi|)\quad\text{with}\quad N_{2}(|\xi|):=i|\xi|^{1-2\rho}\left({\begin{array}[]{*{20}c}0&0&b&0\\ 0&0&0&a\\ -b&0&0&0\\ 0&-a&0&0\\ \end{array}}\right). (2.17)

The following first-order system comes:

Wt(2)+Λ1(|ξ|)W(2)+R2(|ξ|)W(2)=0,\displaystyle W_{t}^{(2)}+\Lambda_{1}(|\xi|)W^{(2)}+R_{2}(|\xi|)W^{(2)}=0,

where

R2=T21B1(1)(|ξ|)N2(|ξ|)=𝒪(|ξ|22ρ)+T21B0(1)(|ξ|)T2=𝒪(|ξ|2θ)=𝒪(|ξ|min{22ρ;2θ}).\displaystyle R_{2}=\underbrace{T_{2}^{-1}B_{1}^{(1)}(|\xi|)N_{2}(|\xi|)}_{=\mathcal{O}(|\xi|^{2-2\rho})}+\underbrace{T_{2}^{-1}B_{0}^{(1)}(|\xi|)T_{2}}_{=\mathcal{O}(|\xi|^{2\theta})}=\mathcal{O}\left(|\xi|^{\min\{2-2\rho;2\theta\}}\right).

To understand the dominant term in the remainder R2(|ξ|)R_{2}(|\xi|), we distinguish between three cases.

Case 2.1.1: ρ+θ<1\rho+\theta<1.
In this case the matrix T21B0(1)(|ξ|)T2T_{2}^{-1}B_{0}^{(1)}(|\xi|)T_{2} has a dominant influence. We find that this matrix can be rewritten by the following way:

T21B0(1)(|ξ|)T2=B0(1)(|ξ|)+T21[B0(1)(|ξ|),N2(|ξ|)].\displaystyle T_{2}^{-1}B_{0}^{(1)}(|\xi|)T_{2}=B_{0}^{(1)}(|\xi|)+T_{2}^{-1}\left[B_{0}^{(1)}(|\xi|),N_{2}(|\xi|)\right].

Thus, setting W(3):=W(2)W^{(3)}:=W^{(2)} implies

Wt(3)+(Λ1(|ξ|)+Λ2(|ξ|))W(3)+R3(|ξ|)W(3)=0,\displaystyle W^{(3)}_{t}+(\Lambda_{1}(|\xi|)+\Lambda_{2}(|\xi|))W^{(3)}+R_{3}(|\xi|)W^{(3)}=0,

where Λ2(|ξ|)=B0(1)(|ξ|)=𝒪(|ξ|2θ)\Lambda_{2}(|\xi|)=B_{0}^{(1)}(|\xi|)=\mathcal{O}\left(|\xi|^{2\theta}\right) and

R3(|ξ|)=T21B1(1)(|ξ|)N2(|ξ|)=𝒪(|ξ|22ρ)+T21[Λ2(|ξ|),N2(|ξ|)]=𝒪(|ξ|1+2θ2ρ)=𝒪(|ξ|22ρ).\displaystyle R_{3}(|\xi|)=\underbrace{T_{2}^{-1}B_{1}^{(1)}(|\xi|)N_{2}(|\xi|)}_{=\mathcal{O}(|\xi|^{2-2\rho})}+\underbrace{T_{2}^{-1}[\Lambda_{2}(|\xi|),N_{2}(|\xi|)]}_{=\mathcal{O}(|\xi|^{1+2\theta-2\rho})}=\mathcal{O}\left(|\xi|^{2-2\rho}\right).

Because 2θ>12\theta>1, the term T21B1(1)(|ξ|)N2(|ξ|)T_{2}^{-1}B_{1}^{(1)}(|\xi|)N_{2}(|\xi|) has a dominant influence in comparison with the term T21[Λ2(|ξ|),N2(|ξ|)]T_{2}^{-1}[\Lambda_{2}(|\xi|),N_{2}(|\xi|)] in the remainder R3(|ξ|)R_{3}(|\xi|). We observe that

T21B1(1)(|ξ|)N2(|ξ|)\displaystyle T_{2}^{-1}B_{1}^{(1)}(|\xi|)N_{2}(|\xi|) =B1(1)(|ξ|)N2(|ξ|)N2(|ξ|)T21B1(1)(|ξ|)N2(|ξ|),\displaystyle=B_{1}^{(1)}(|\xi|)N_{2}(|\xi|)-N_{2}(|\xi|)T_{2}^{-1}B_{1}^{(1)}(|\xi|)N_{2}(|\xi|),
B1(1)(|ξ|)N2(|ξ|)\displaystyle B_{1}^{(1)}(|\xi|)N_{2}(|\xi|) =|ξ|22ρdiag(b2,a2,b2,a2).\displaystyle=|\xi|^{2-2\rho}\operatorname{diag}\left(b^{2},a^{2},-b^{2},-a^{2}\right).

So, by taking W(4):=W(3)W^{(4)}:=W^{(3)} we have

Wt(4)+(Λ1(|ξ|)+Λ2(|ξ|)+Λ3(|ξ|))W(4)+R4(|ξ|)W(4)=0,\displaystyle W^{(4)}_{t}+(\Lambda_{1}(|\xi|)+\Lambda_{2}(|\xi|)+\Lambda_{3}(|\xi|))W^{(4)}+R_{4}(|\xi|)W^{(4)}=0,

where Λ3(|ξ|)=B1(1)(|ξ|)N2(|ξ|)=𝒪(|ξ|22ρ)\Lambda_{3}(|\xi|)=B_{1}^{(1)}(|\xi|)N_{2}(|\xi|)=\mathcal{O}\left(|\xi|^{2-2\rho}\right) and

R4(|ξ|)=N2(|ξ|)T21B1(1)(|ξ|)N2(|ξ|)+T21[Λ2(|ξ|),N2(|ξ|)]=𝒪(|ξ|1+2θ2ρ).\displaystyle R_{4}(|\xi|)=-N_{2}(|\xi|)T_{2}^{-1}B_{1}^{(1)}(|\xi|)N_{2}(|\xi|)+T_{2}^{-1}[\Lambda_{2}(|\xi|),N_{2}(|\xi|)]=\mathcal{O}\left(|\xi|^{1+2\theta-2\rho}\right).

Up to now, we have derived pairwise distinct eigenvalues and R4(|ξ|)=𝒪(|ξ|1+2θ2ρ)R_{4}(|\xi|)=\mathcal{O}\left(|\xi|^{1+2\theta-2\rho}\right).

Case 2.1.2: ρ+θ=1\rho+\theta=1.
In this case the matrices T21B0(1)(|ξ|)T2T_{2}^{-1}B_{0}^{(1)}(|\xi|)T_{2} and T21B1(1)(|ξ|)N2(|ξ|)T_{2}^{-1}B_{1}^{(1)}(|\xi|)N_{2}(|\xi|) have the same influence. For this reason, we set

Λ2(|ξ|)=B1(1)(|ξ|)N2(|ξ|)+B0(1)(|ξ|)=|ξ|22ρdiag(b2,a2,1b2,1a2)=𝒪(|ξ|22ρ).\displaystyle\Lambda_{2}(|\xi|)=B_{1}^{(1)}(|\xi|)N_{2}(|\xi|)+B_{0}^{(1)}(|\xi|)=|\xi|^{2-2\rho}\operatorname{diag}\left(b^{2},a^{2},1-b^{2},1-a^{2}\right)=\mathcal{O}\left(|\xi|^{2-2\rho}\right).

Then, taking W(3):=W(2)W^{(3)}:=W^{(2)} again we derive

Wt(3)+(Λ1(|ξ|)+Λ2(|ξ|))W(3)+R3(|ξ|)W(3)=0,\displaystyle W_{t}^{(3)}+(\Lambda_{1}(|\xi|)+\Lambda_{2}(|\xi|))W^{(3)}+R_{3}(|\xi|)W^{(3)}=0,

where

R3(|ξ|)=N2(|ξ|)T21B1(1)(|ξ|)N2(|ξ|)+T21[B0(1)(|ξ|),N2(|ξ|)]=𝒪(|ξ|34ρ).\displaystyle R_{3}(|\xi|)=-N_{2}(|\xi|)T_{2}^{-1}B_{1}^{(1)}(|\xi|)N_{2}(|\xi|)+T_{2}^{-1}\left[B_{0}^{(1)}(|\xi|),N_{2}(|\xi|)\right]=\mathcal{O}\left(|\xi|^{3-4\rho}\right).

Up to now, we have derived pairwise distinct eigenvalues and R3(|ξ|)=𝒪(|ξ|34ρ)R_{3}(|\xi|)=\mathcal{O}\left(|\xi|^{3-4\rho}\right).

Case 2.1.3: ρ+θ>1\rho+\theta>1.
In this case the matrix T21B1(1)(|ξ|)N2(|ξ|)T_{2}^{-1}B_{1}^{(1)}(|\xi|)N_{2}(|\xi|) has a dominant influence. Following the idea from Case 2.1.1 and setting W(3):=W(2)W^{(3)}:=W^{(2)} again, we may derive

Wt(3)+(Λ1(|ξ|)+Λ2(|ξ|))W(3)+R3(|ξ|)W(3)=0,\displaystyle W_{t}^{(3)}+(\Lambda_{1}(|\xi|)+\Lambda_{2}(|\xi|))W^{(3)}+R_{3}(|\xi|)W^{(3)}=0,

where Λ2(|ξ|)=B1(1)(|ξ|)N2(|ξ|)=𝒪(|ξ|22ρ)\Lambda_{2}(|\xi|)=B_{1}^{(1)}(|\xi|)N_{2}(|\xi|)=\mathcal{O}(|\xi|^{2-2\rho}) and

R3(|ξ|)=N2(|ξ|)T21B1(1)(|ξ|)N2(|ξ|)+T21B0(1)(|ξ|)T2=𝒪(|ξ|min{34ρ;2θ}).\displaystyle R_{3}(|\xi|)=-N_{2}(|\xi|)T_{2}^{-1}B_{1}^{(1)}(|\xi|)N_{2}(|\xi|)+T_{2}^{-1}B_{0}^{(1)}(|\xi|)T_{2}=\mathcal{O}\left(|\xi|^{\min\{3-4\rho;2\theta\}}\right).

Up to now, we have derived pairwise distinct eigenvalues and R3(|ξ|)=𝒪(|ξ|min{34ρ;2θ})R_{3}(|\xi|)=\mathcal{O}\left(|\xi|^{\min\{3-4\rho;2\theta\}}\right).

Summarizing above diagonalization procedure, according to [20] we obtain the next proposition, which tells us the asymptotic behavior of eigenvalues and representation of solutions.

Proposition 2.1.

The eigenvalues λj=λj(|ξ|)\lambda_{j}=\lambda_{j}(|\xi|) of the coefficient matrix

B(|ξ|;ρ,θ)=12(|ξ|2ρ+|ξ|2θ)B0+i|ξ|B1\displaystyle B(|\xi|;\rho,\theta)=\frac{1}{2}\left(|\xi|^{2\rho}+|\xi|^{2\theta}\right)B_{0}+i|\xi|B_{1}

from (2.5) behave for |ξ|<ε1|\xi|<\varepsilon\ll 1 as

  • if ρ+θ<1\rho+\theta<1, then

    λ1(|ξ|)=b2|ξ|22ρ+𝒪(|ξ|1+2θ2ρ),\displaystyle\lambda_{1}(|\xi|)=b^{2}|\xi|^{2-2\rho}+\mathcal{O}\left(|\xi|^{1+2\theta-2\rho}\right),
    λ2(|ξ|)=a2|ξ|22ρ+𝒪(|ξ|1+2θ2ρ),\displaystyle\lambda_{2}(|\xi|)=a^{2}|\xi|^{2-2\rho}+\mathcal{O}\left(|\xi|^{1+2\theta-2\rho}\right),
    λ3(|ξ|)=|ξ|2ρ+|ξ|2θb2|ξ|22ρ+𝒪(|ξ|1+2θ2ρ),\displaystyle\lambda_{3}(|\xi|)=|\xi|^{2\rho}+|\xi|^{2\theta}-b^{2}|\xi|^{2-2\rho}+\mathcal{O}\left(|\xi|^{1+2\theta-2\rho}\right),
    λ4(|ξ|)=|ξ|2ρ+|ξ|2θa2|ξ|22ρ+𝒪(|ξ|1+2θ2ρ);\displaystyle\lambda_{4}(|\xi|)=|\xi|^{2\rho}+|\xi|^{2\theta}-a^{2}|\xi|^{2-2\rho}+\mathcal{O}\left(|\xi|^{1+2\theta-2\rho}\right);
  • if ρ+θ=1\rho+\theta=1, then

    λ1(|ξ|)=b2|ξ|22ρ+𝒪(|ξ|34ρ),\displaystyle\lambda_{1}(|\xi|)=b^{2}|\xi|^{2-2\rho}+\mathcal{O}\left(|\xi|^{3-4\rho}\right),
    λ2(|ξ|)=a2|ξ|22ρ+𝒪(|ξ|34ρ),\displaystyle\lambda_{2}(|\xi|)=a^{2}|\xi|^{2-2\rho}+\mathcal{O}\left(|\xi|^{3-4\rho}\right),
    λ3(|ξ|)=|ξ|2ρ+(1b2)|ξ|22ρ+𝒪(|ξ|34ρ),\displaystyle\lambda_{3}(|\xi|)=|\xi|^{2\rho}+\left(1-b^{2}\right)|\xi|^{2-2\rho}+\mathcal{O}\left(|\xi|^{3-4\rho}\right),
    λ4(|ξ|)=|ξ|2ρ+(1a2)|ξ|22ρ+𝒪(|ξ|34ρ);\displaystyle\lambda_{4}(|\xi|)=|\xi|^{2\rho}+\left(1-a^{2}\right)|\xi|^{2-2\rho}+\mathcal{O}\left(|\xi|^{3-4\rho}\right);
  • if ρ+θ>1\rho+\theta>1, then

    λ1(|ξ|)=b2|ξ|22ρ+𝒪(|ξ|min{34ρ;2θ}),\displaystyle\lambda_{1}(|\xi|)=b^{2}|\xi|^{2-2\rho}+\mathcal{O}\left(|\xi|^{\min\{3-4\rho;2\theta\}}\right),
    λ2(|ξ|)=a2|ξ|22ρ+𝒪(|ξ|min{34ρ;2θ}),\displaystyle\lambda_{2}(|\xi|)=a^{2}|\xi|^{2-2\rho}+\mathcal{O}\left(|\xi|^{\min\{3-4\rho;2\theta\}}\right),
    λ3(|ξ|)=|ξ|2ρb2|ξ|22ρ+𝒪(|ξ|min{34ρ;2θ}),\displaystyle\lambda_{3}(|\xi|)=|\xi|^{2\rho}-b^{2}|\xi|^{2-2\rho}+\mathcal{O}\left(|\xi|^{\min\{3-4\rho;2\theta\}}\right),
    λ4(|ξ|)=|ξ|2ρa2|ξ|22ρ+𝒪(|ξ|min{34ρ;2θ}).\displaystyle\lambda_{4}(|\xi|)=|\xi|^{2\rho}-a^{2}|\xi|^{2-2\rho}+\mathcal{O}\left(|\xi|^{\min\{3-4\rho;2\theta\}}\right).

    Furthermore, the solution to the Cauchy problem (2.5) has in 𝒵int(ε)\mathcal{Z}_{\operatorname{int}}(\varepsilon) the representation

    W(t,ξ)=Tint1(|ξ|)diag(eλj(|ξ|)t)j=14Tint(|ξ|)W0(ξ),\displaystyle W(t,\xi)=T_{\operatorname{int}}^{-1}(|\xi|)\operatorname{diag}\left(e^{-\lambda_{j}(|\xi|)t}\right)_{j=1}^{4}T_{\operatorname{int}}(|\xi|)W_{0}(\xi),

    where Tint(|ξ|)=(I4×4+N2(|ξ|))1T11T_{\operatorname{int}}(|\xi|)=(I_{4\times 4}+N_{2}(|\xi|))^{-1}T_{1}^{-1} with a matrix N2(|ξ|)=𝒪(|ξ|12ρ)N_{2}(|\xi|)=\mathcal{O}\left(|\xi|^{1-2\rho}\right) for |ξ|0|\xi|\rightarrow 0. Here the matrix T1T_{1} is defined in (2.12).

2.2. Treatment for large frequencies

We observe that the symmetric of the system (2.5) with respective to the parameters ρ\rho and θ\theta. Thus, by similar procedure we can obtain pairwise distinct eigenvalues. Before stating our result for large frequencies, we define

T3:=I4×4+N3(|ξ|)withN3(|ξ|)=i|ξ|12θ(00b0000ab0000a00).\displaystyle T_{3}:=I_{4\times 4}+N_{3}(|\xi|)\quad\text{with}\quad N_{3}(|\xi|)=i|\xi|^{1-2\theta}\left({\begin{array}[]{*{20}c}0&0&b&0\\ 0&0&0&a\\ -b&0&0&0\\ 0&-a&0&0\\ \end{array}}\right).

Then, following the similar procedure as the case for small frequencies and according to the thesis [20] we obtain the next proposition.

Proposition 2.2.

The eigenvalues μj=μj(|ξ|)\mu_{j}=\mu_{j}(|\xi|) of the coefficient matrix

B(|ξ|;ρ,θ)=12(|ξ|2ρ+|ξ|2θ)B0+i|ξ|B1\displaystyle B(|\xi|;\rho,\theta)=\frac{1}{2}\left(|\xi|^{2\rho}+|\xi|^{2\theta}\right)B_{0}+i|\xi|B_{1}

from (2.5) behave for |ξ|>N1|\xi|>N\gg 1 as

  • if ρ+θ<1\rho+\theta<1, then

    μ1(|ξ|)=b2|ξ|22θ+𝒪(|ξ|min{34θ;2ρ}),\displaystyle\mu_{1}(|\xi|)=b^{2}|\xi|^{2-2\theta}+\mathcal{O}\left(|\xi|^{\min\{3-4\theta;2\rho\}}\right),
    μ2(|ξ|)=a2|ξ|22θ+𝒪(|ξ|min{34θ;2ρ}),\displaystyle\mu_{2}(|\xi|)=a^{2}|\xi|^{2-2\theta}+\mathcal{O}\left(|\xi|^{\min\{3-4\theta;2\rho\}}\right),
    μ3(|ξ|)=|ξ|2θb2|ξ|22θ+𝒪(|ξ|min{34θ;2ρ}),\displaystyle\mu_{3}(|\xi|)=|\xi|^{2\theta}-b^{2}|\xi|^{2-2\theta}+\mathcal{O}\left(|\xi|^{\min\{3-4\theta;2\rho\}}\right),
    μ4(|ξ|)=|ξ|2θa2|ξ|22θ+𝒪(|ξ|min{34θ;2ρ}).\displaystyle\mu_{4}(|\xi|)=|\xi|^{2\theta}-a^{2}|\xi|^{2-2\theta}+\mathcal{O}\left(|\xi|^{\min\{3-4\theta;2\rho\}}\right).
  • if ρ+θ=1\rho+\theta=1, then

    μ1(|ξ|)=b2|ξ|22θ+𝒪(|ξ|34θ),\displaystyle\mu_{1}(|\xi|)=b^{2}|\xi|^{2-2\theta}+\mathcal{O}\left(|\xi|^{3-4\theta}\right),
    μ2(|ξ|)=a2|ξ|22θ+𝒪(|ξ|34θ),\displaystyle\mu_{2}(|\xi|)=a^{2}|\xi|^{2-2\theta}+\mathcal{O}\left(|\xi|^{3-4\theta}\right),
    μ3(|ξ|)=|ξ|2θ+(1b2)|ξ|22θ+𝒪(|ξ|34θ),\displaystyle\mu_{3}(|\xi|)=|\xi|^{2\theta}+\left(1-b^{2}\right)|\xi|^{2-2\theta}+\mathcal{O}\left(|\xi|^{3-4\theta}\right),
    μ4(|ξ|)=|ξ|2θ+(1a2)|ξ|22θ+𝒪(|ξ|34θ);\displaystyle\mu_{4}(|\xi|)=|\xi|^{2\theta}+\left(1-a^{2}\right)|\xi|^{2-2\theta}+\mathcal{O}\left(|\xi|^{3-4\theta}\right);
  • if ρ+θ>1\rho+\theta>1, then

    μ1(|ξ|)=b2|ξ|22θ+𝒪(|ξ|1+2ρ2θ),\displaystyle\mu_{1}(|\xi|)=b^{2}|\xi|^{2-2\theta}+\mathcal{O}\left(|\xi|^{1+2\rho-2\theta}\right),
    μ2(|ξ|)=a2|ξ|22θ+𝒪(|ξ|1+2ρ2θ),\displaystyle\mu_{2}(|\xi|)=a^{2}|\xi|^{2-2\theta}+\mathcal{O}\left(|\xi|^{1+2\rho-2\theta}\right),
    μ3(|ξ|)=|ξ|2θ+|ξ|2ρb2|ξ|22θ+𝒪(|ξ|1+2ρ2θ),\displaystyle\mu_{3}(|\xi|)=|\xi|^{2\theta}+|\xi|^{2\rho}-b^{2}|\xi|^{2-2\theta}+\mathcal{O}\left(|\xi|^{1+2\rho-2\theta}\right),
    μ4(|ξ|)=|ξ|2θ+|ξ|2ρa2|ξ|22θ+𝒪(|ξ|1+2ρ2θ);\displaystyle\mu_{4}(|\xi|)=|\xi|^{2\theta}+|\xi|^{2\rho}-a^{2}|\xi|^{2-2\theta}+\mathcal{O}\left(|\xi|^{1+2\rho-2\theta}\right);

    Furthermore, the solution to the Cauchy problem (2.5) has in 𝒵ext(N)\mathcal{Z}_{\operatorname{ext}}(N) the representation

    W(t,ξ)=Text1(|ξ|)diag(eμj(|ξ|)t)j=14Text(|ξ|)W0(ξ),\displaystyle W(t,\xi)=T_{\operatorname{ext}}^{-1}(|\xi|)\operatorname{diag}\left(e^{-\mu_{j}(|\xi|)t}\right)_{j=1}^{4}T_{\operatorname{ext}}(|\xi|)W_{0}(\xi),

    where Text(|ξ|)=(I4×4+N3(|ξ|))1T11T_{\operatorname{ext}}(|\xi|)=(I_{4\times 4}+N_{3}(|\xi|))^{-1}T_{1}^{-1} with a matrix N3(|ξ|)=𝒪(|ξ|12θ)N_{3}(|\xi|)=\mathcal{O}\left(|\xi|^{1-2\theta}\right) for |ξ||\xi|\rightarrow\infty. Here the matrix T1T_{1} is defined in (2.12).

2.3. Treatment for bounded frequencies

Finally, we only need to derive an exponential decay of solutions to (2.5) for bounded frequencies to guarantee the exponential stability of solutions.

Proposition 2.3.

The solution W=W(t,ξ)W=W(t,\xi) to the Cauchy problem (2.5) with 0ρ<1/2<θ10\leq\rho<1/2<\theta\leq 1 fulfills the following exponential decay estimate:

|W(t,ξ)|ect|W0(ξ)|,\displaystyle|W(t,\xi)|\lesssim e^{-ct}|W_{0}(\xi)|,

for (t,ξ)(0,)×𝒵bdd(ε,N)(t,\xi)\in(0,\infty)\times\mathcal{Z}_{\text{bdd}}(\varepsilon,N), where cc is a positive constant.

Proof.

Let us recall that

B(|ξ|;ρ,θ)=12(|ξ|2ρ+|ξ|2θ)B0+i|ξ|B1.\displaystyle B(|\xi|;\rho,\theta)=\frac{1}{2}\left(|\xi|^{2\rho}+|\xi|^{2\theta}\right)B_{0}+i|\xi|B_{1}.

It is clear that the eigenvalues of B(|ξ|;ρ,θ)B(|\xi|;\rho,\theta) satisfy

0\displaystyle 0 =det(B(|ξ|;ρ,θ)λI4×4)\displaystyle=\det(B(|\xi|;\rho,\theta)-\lambda I_{4\times 4})
=λ42(|ξ|2ρ+|ξ|2θ)λ3+((|ξ|2ρ+|ξ|2θ)2+(a2+b2)|ξ|2)λ2\displaystyle=\lambda^{4}-2\left(|\xi|^{2\rho}+|\xi|^{2\theta}\right)\lambda^{3}+\left(\left(|\xi|^{2\rho}+|\xi|^{2\theta}\right)^{2}+\left(a^{2}+b^{2}\right)|\xi|^{2}\right)\lambda^{2}
(a2+b2)|ξ|2(|ξ|2ρ+|ξ|2θ)λ+a2b2|ξ|4.\displaystyle\quad-\left(a^{2}+b^{2}\right)|\xi|^{2}\left(|\xi|^{2\rho}+|\xi|^{2\theta}\right)\lambda+a^{2}b^{2}|\xi|^{4}.

Now, we assume there exists an eigenvalue λ=id\lambda=id with d\{0}d\in\mathbb{R}\backslash\{0\}. Therefore, the real number dd should satisfy the equations

{d4((|ξ|2ρ+|ξ|2θ)2+(a2+b2)|ξ|2)d2+a2b2|ξ|4=0,id(|ξ|2ρ+|ξ|2θ)(2d2(a2+b2)|ξ|2)=0.\left\{\begin{aligned} &d^{4}-\left(\left(|\xi|^{2\rho}+|\xi|^{2\theta}\right)^{2}+\left(a^{2}+b^{2}\right)|\xi|^{2}\right)d^{2}+a^{2}b^{2}|\xi|^{4}=0,\\ &id\left(|\xi|^{2\rho}+|\xi|^{2\theta}\right)\left(2d^{2}-\left(a^{2}+b^{2}\right)|\xi|^{2}\right)=0.\end{aligned}\right. (2.18)

Due to the facts that d0d\neq 0 and ξ𝒵bdd(ε,N)\xi\in\mathcal{Z}_{\text{bdd}}(\varepsilon,N), the equations (2.18) leads to

(b2a2)2|ξ|2=2(a2+b2)(|ξ|2ρ+|ξ|2θ)2.\displaystyle-\left(b^{2}-a^{2}\right)^{2}|\xi|^{2}=2\left(a^{2}+b^{2}\right)\left(|\xi|^{2\rho}+|\xi|^{2\theta}\right)^{2}.

From our assumption b>a>0b>a>0, we immediately find a contradiction. Thus, there not exists pure imaginary eigenvalue of B(|ξ|;ρ,θ)B(|\xi|;\rho,\theta) for any 0ρ<1/2<θ10\leq\rho<1/2<\theta\leq 1 and ξ𝒵bdd(ε,N)\xi\in\mathcal{Z}_{\text{bdd}}(\varepsilon,N). Lastly, by using the compactness of the bounded zone 𝒵bdd(ε,N)\mathcal{Z}_{\text{bdd}}(\varepsilon,N) and the continuity of the eigenvalues, the proof is complete. ∎

3. Energy estimates

The aim of the section is to study the dissipative structure and sharp energy estimates to doubly dissipative elastic waves, where initial data belongs to Bessel potential space with additional LmL^{m} regularity (m[1,2]m\in[1,2]) or with additional weighted L1L^{1} regularity.

The crucial point of sharp energy estimates is to derive the sharp pointwise estimate. By summarizing the results in Propositions 2.1, 2.2 and 2.3, we obtain the result on the sharp pointwise estimate of solutions to (2.5).

Proposition 3.1.

The solution W=W(t,ξ)W=W(t,\xi) to the Cauchy problem (2.5) with 0ρ<1/2<θ10\leq\rho<1/2<\theta\leq 1 satisfies the following pointwise estimates for any ξ2\xi\in\mathbb{R}^{2} and t0t\geq 0:

|W(t,ξ)|ecη(|ξ|)t|W0(ξ)|,\displaystyle|W(t,\xi)|\lesssim e^{-c\eta(|\xi|)t}|W_{0}(\xi)|,

where η(|ξ|):=|ξ|22ρ1+|ξ|2θ2ρ\eta(|\xi|):=\frac{|\xi|^{2-2\rho}}{1+|\xi|^{2\theta-2\rho}} and cc is positive constant.

Remark 3.1.

The pointwise estimate in Proposition 3.1 gives the characterization of the dissipative structure of doubly dissipative elastic waves. We now compare the dissipative structure of doubly dissipative elastic waves and elastic waves with friction or structural damping in [30, 5]. For one thing, as |ξ|0|\xi|\rightarrow 0, the dissipative structure of doubly dissipative elastic waves is the same as elastic waves with friction or structural damping (Δ)ρut(-\Delta)^{\rho}u_{t} for ρ[0,1/2)\rho\in[0,1/2), that is η(|ξ|)|ξ|22ρ\eta(|\xi|)\asymp|\xi|^{2-2\rho} for |ξ|0|\xi|\rightarrow 0. For another, as |ξ||\xi|\rightarrow\infty, the dissipative structure of doubly dissipative elastic waves is the same as elastic waves with structural damping (Δ)θut(-\Delta)^{\theta}u_{t} for θ(1/2,1]\theta\in(1/2,1], that is η(|ξ|)|ξ|22θ\eta(|\xi|)\asymp|\xi|^{2-2\theta} for |ξ||\xi|\rightarrow\infty.

Now, we state our main result on energy estimates.

Theorem 3.1.

Let us consider the Cauchy problem (2.7) with 0ρ<1/2<θ10\leq\rho<1/2<\theta\leq 1 and U0HsLmU_{0}\in H^{s}\cap L^{m}, where s0s\geq 0 and m[1,2]m\in[1,2]. Then, the following estimates hold:

U(t,)H˙s(1+t)s22ρ2mm(22ρ)U0HsLm.\displaystyle\|U(t,\cdot)\|_{\dot{H}^{s}}\lesssim(1+t)^{-\frac{s}{2-2\rho}-\frac{2-m}{m(2-2\rho)}}\|U_{0}\big{\|}_{H^{s}\cap L^{m}}.
Remark 3.2.

According to Proposition 2.1 and sharp pointwise estimate in Proposition 3.1, the energy estimates in Theorem 3.1 are sharp for initial data U0HsLmU_{0}\in H^{s}\cap L^{m}, where s0s\geq 0 and m[1,2]m\in[1,2].

Remark 3.3.

We remark that the energy estimates for doubly dissipative elastic waves (1.1) in Theorem 3.1 are the same as damped elastic waves with damping term (Δ)ρut(-\Delta)^{\rho}u_{t} for ρ[0,1/2)\rho\in[0,1/2) in Theorems 7.2 and 7.3 in [30].

Remark 3.4.

From energy estimates in Theorem 3.1, we observe that the decay rate is only determined by the damping term (Δ)ρut(-\Delta)^{\rho}u_{t} with ρ[0,1/2)\rho\in[0,1/2) in (1.1). For the other damping term (Δ)θut(-\Delta)^{\theta}u_{t} with θ(1/2,1]\theta\in(1/2,1], there is no any influence for the energy estimates. The main reason is that the decay rate for energy estimates of (1.1) is mainly determined by dissipative structure for small frequencies. However, for the dissipative structure for small frequencies (see Proposition 2.1), the dominant influence of eigenvalues are determined by |ξ|22ρ|\xi|^{2-2\rho}. Although the parameter θ\theta in the damping term (Δ)θut(-\Delta)^{\theta}u_{t} has a great influence on the asymptotic behavior of eigenvalues for large frequencies, the solutions satisfies an exponential decay for large frequencies providing that we assume suitable regularity for initial data.

Proof.

To begin with, by using Proposition 3.1, we calculate

W(t,)H˙s\displaystyle\|W(t,\cdot)\|_{\dot{H}^{s}} χint(ξ)|ξ|sec|ξ|22ρtW0(ξ)L2+ectχbdd(ξ)|ξ|sW0(ξ)L2\displaystyle\lesssim\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s}e^{-c|\xi|^{2-2\rho}t}W_{0}(\xi)\right\|_{L^{2}}+e^{-ct}\left\|\chi_{\text{bdd}}(\xi)|\xi|^{s}W_{0}(\xi)\right\|_{L^{2}}
+χext(ξ)|ξ|sec|ξ|22θtW0(ξ)L2\displaystyle\quad+\left\|\chi_{\operatorname{ext}}(\xi)|\xi|^{s}e^{-c|\xi|^{2-2\theta}t}W_{0}(\xi)\right\|_{L^{2}}
χint(ξ)|ξ|sec|ξ|22ρtW0(ξ)L2+ectU0Hs.\displaystyle\lesssim\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s}e^{-c|\xi|^{2-2\rho}t}W_{0}(\xi)\right\|_{L^{2}}+e^{-ct}\|U_{0}\|_{H^{s}}.

Next, we divide the proof into two cases. For the case when m=2m=2 in Theorem 3.1, we have

χint(ξ)|ξ|sec|ξ|22ρtW0(ξ)L2\displaystyle\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s}e^{-c|\xi|^{2-2\rho}t}W_{0}(\xi)\right\|_{L^{2}} supξ𝒵int(ε)||ξ|sec|ξ|22ρt|W0L2\displaystyle\lesssim\sup\limits_{\xi\in\mathcal{Z}_{\operatorname{int}}(\varepsilon)}\left||\xi|^{s}e^{-c|\xi|^{2-2\rho}t}\right|\|W_{0}\|_{L^{2}}
(1+t)s22ρU0L2.\displaystyle\lesssim(1+t)^{-\frac{s}{2-2\rho}}\|U_{0}\|_{L^{2}}.

For the case when m[1,2)m\in[1,2) in Theorem 3.1, the applications of Hölder’s inequality and the Hausdorff-Young inequality yield

χint(ξ)|ξ|sec|ξ|22ρtW0(ξ)L2\displaystyle\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s}e^{-c|\xi|^{2-2\rho}t}W_{0}(\xi)\right\|_{L^{2}} χint(ξ)|ξ|sec|ξ|22ρtL2m2mU0Lm\displaystyle\lesssim\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s}e^{-c|\xi|^{2-2\rho}t}\right\|_{L^{\frac{2m}{2-m}}}\|U_{0}\|_{L^{m}}
(1+t)s22ρ2mm(22ρ)U0Lm.\displaystyle\lesssim(1+t)^{-\frac{s}{2-2\rho}-\frac{2-m}{m(2-2\rho)}}\|U_{0}\|_{L^{m}}.

Finally, by applying the Parseval-Plancherel theorem, we immediately complete the proof. ∎

Furthermore, we discuss energy estimates in a framework of weighted L1L^{1} data. Before stating our result, we recall the Lemma 2.1 in the paper [10].

Lemma 3.1.

Let fL1,γf\in L^{1,\gamma} with γ(0,1]\gamma\in(0,1]. Then, the following estimate holds:

|f^(ξ)|Cγ|ξ|γfL1,γ+|nf(x)𝑑x|,\displaystyle\left|\hat{f}(\xi)\right|\leq C_{\gamma}|\xi|^{\gamma}\|f\|_{L^{1,\gamma}}+\left|\int_{\mathbb{R}^{n}}f(x)dx\right|,

with a positive constant Cγ>0C_{\gamma}>0.

Theorem 3.2.

Let us consider the Cauchy problem (2.7) with 0ρ<1/2<θ10\leq\rho<1/2<\theta\leq 1 and U0HsL1,γU_{0}\in H^{s}\cap L^{1,\gamma}, where s0s\geq 0 and γ(0,1]\gamma\in(0,1]. Then, the following estimates hold:

U(t,)H˙s\displaystyle\|U(t,\cdot)\|_{\dot{H}^{s}} (1+t)s+γ22ρ122ρU0HsL1,γ+(1+t)s22ρ122ρ|2U0(x)𝑑x|.\displaystyle\lesssim(1+t)^{-\frac{s+\gamma}{2-2\rho}-\frac{1}{2-2\rho}}\|U_{0}\|_{H^{s}\cap L^{1,\gamma}}+(1+t)^{-\frac{s}{2-2\rho}-\frac{1}{2-2\rho}}\left|\int_{\mathbb{R}^{2}}U_{0}(x)dx\right|.
Remark 3.5.

We remark that if we take initial data satisfying |2U0(x)𝑑x|=0\left|\int_{\mathbb{R}^{2}}U_{0}(x)dx\right|=0 in Theorem 3.2, then we can observe that the decay rates given in Theorem 3.1 when m=1m=1 can be improved by (1+t)γ22ρ(1+t)^{-\frac{\gamma}{2-2\rho}} for γ(0,1]\gamma\in(0,1].

Proof.

To prove Theorem 3.2, we only need to modify the estimate for small frequencies. By using Lemma 3.1, we have

|W0(ξ)||ξ|γU0L1,γ+|2U0(x)𝑑x|.\displaystyle|W_{0}(\xi)|\lesssim|\xi|^{\gamma}\|U_{0}\|_{L^{1,\gamma}}+\left|\int_{\mathbb{R}^{2}}U_{0}(x)dx\right|.

Then, we derive

χint(ξ)|ξ|sec|ξ|22ρtW0(ξ)L2\displaystyle\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s}e^{-c|\xi|^{2-2\rho}t}W_{0}(\xi)\right\|_{L^{2}} χint(ξ)|ξ|s+γec|ξ|22ρtL2U0L1,γ\displaystyle\lesssim\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s+\gamma}e^{-c|\xi|^{2-2\rho}t}\right\|_{L^{2}}\|U_{0}\|_{L^{1,\gamma}}
+χint(ξ)|ξ|sec|ξ|22ρtL2|2U0(x)𝑑x|.\displaystyle\quad+\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s}e^{-c|\xi|^{2-2\rho}t}\right\|_{L^{2}}\left|\int_{\mathbb{R}^{2}}U_{0}(x)dx\right|.

Then, combining with the proof of Theorem 3.1, we complete the proof. ∎

4. Diffusion phenomena

Our main purpose in this section is to obtain diffusion phenomena for doubly dissipative elastic waves. According to Theorems 3.1 and 3.2, we observe that the decay rate of energy estimates is determined by small frequencies (see Remark 3.4). However, we may obtain an exponential decay estimates with suitable regularity on initial data for bounded frequencies and large frequencies. For this reason, we will interpret diffusion phenomena by the solutions localized in small frequency zone in this section.

To do this, we first introduce the corresponding reference systems for the cases ρ+θ<1\rho+\theta<1, ρ+θ=1\rho+\theta=1 and ρ+θ>1\rho+\theta>1, respectively. Firstly, we introduce the matrices

M1:=(b20000a20000b20000a2)andM2:=(0000000000100001).\displaystyle M_{1}:=\left({\begin{array}[]{*{20}c}b^{2}&0&0&0\\ 0&a^{2}&0&0\\ 0&0&-b^{2}&0\\ 0&0&0&-a^{2}\\ \end{array}}\right)\,\,\,\,\text{and}\,\,\,\,M_{2}:=\left({\begin{array}[]{*{20}c}0&0&0&0\\ 0&0&0&0\\ 0&0&1&0\\ 0&0&0&1\\ \end{array}}\right).

Motivated by the principle part of eigenvalues in Proposition 2.1, we define the different reference systems between the following three cases.

  • In the case ρ+θ<1\rho+\theta<1, we define U~=U~(t,x;ρ,θ)\widetilde{U}=\widetilde{U}(t,x;\rho,\theta) is the solution to the following evolution system:

    {U~t+M1(Δ)1ρU~+M2(Δ)ρU~+M2(Δ)θU~=0,x2,t>0,U~(0,x)=T11U0(x),x2.\left\{\begin{aligned} &\widetilde{U}_{t}+M_{1}(-\Delta)^{1-\rho}\widetilde{U}+M_{2}(-\Delta)^{\rho}\widetilde{U}+M_{2}(-\Delta)^{\theta}\widetilde{U}=0,&&x\in\mathbb{R}^{2},\,\,t>0,\\ &\widetilde{U}(0,x)=T_{1}^{-1}U_{0}(x),&&x\in\mathbb{R}^{2}.\end{aligned}\right. (4.19)
  • In the case ρ+θ=1\rho+\theta=1, we define U~=U~(t,x;ρ,θ)\widetilde{U}=\widetilde{U}(t,x;\rho,\theta) is the solution to the following evolution system:

    {U~t+(M1+M2)(Δ)1ρU~+M2(Δ)ρU~=0,x2,t>0,U~(0,x)=T11U0(x),x2.\left\{\begin{aligned} &\widetilde{U}_{t}+(M_{1}+M_{2})(-\Delta)^{1-\rho}\widetilde{U}+M_{2}(-\Delta)^{\rho}\widetilde{U}=0,&&x\in\mathbb{R}^{2},\,\,t>0,\\ &\widetilde{U}(0,x)=T_{1}^{-1}U_{0}(x),&&x\in\mathbb{R}^{2}.\end{aligned}\right. (4.20)
  • In the case ρ+θ>1\rho+\theta>1, we define U~=U~(t,x;ρ,θ)\widetilde{U}=\widetilde{U}(t,x;\rho,\theta) is the solution to the following evolution system:

    {U~t+M1(Δ)1ρU~+M2(Δ)ρU~=0,x2,t>0,U~(0,x)=T11U0(x),x2.\left\{\begin{aligned} &\widetilde{U}_{t}+M_{1}(-\Delta)^{1-\rho}\widetilde{U}+M_{2}(-\Delta)^{\rho}\widetilde{U}=0,&&x\in\mathbb{R}^{2},\,\,t>0,\\ &\widetilde{U}(0,x)=T_{1}^{-1}U_{0}(x),&&x\in\mathbb{R}^{2}.\end{aligned}\right. (4.21)

Here the matrix T1T_{1} is defined in (2.12).

Let us now give some explanation for these reference system.

In the case when ρ+θ<1\rho+\theta<1, for the evolution system (4.19), we find that the reference system is made up of three different parabolic systems. We may interpret this new effect as triple diffusion phenomena. This effect is shown firstly in [4] for thermoelastic plate equations with structural damping. In this case, the damping term (Δ)θut(-\Delta)^{\theta}u_{t} with θ(1/2,1]\theta\in(1/2,1] in (1.1) really has influence on the diffusion structure. But this effect does not appear in the other case ρ+θ1\rho+\theta\geq 1.

However, we find that when ρ+θ1\rho+\theta\geq 1, the reference system (4.19) is changed into (4.20) and (4.21). Obviously, these reference systems are only made up of two different parabolic systems, whose structures are similar as reference system for elastic waves with damping term (Δ)ρu(-\Delta)^{\rho}u for ρ[0,1/2)\rho\in[0,1/2). We may interpret this effect as double diffusion phenomena (one may see the pioneering paper [7]).

From the above discussions, we observe a new threshold of diffusion structure for doubly dissipative elastic waves, that is ρ+θ=1\rho+\theta=1. In other words, the structure of the reference system will be changed with the parameters changing from ρ+θ<1\rho+\theta<1 to ρ+θ1\rho+\theta\geq 1.

Let us begin to state our main theorems on diffusion phenomena.

Theorem 4.1.

Let us consider the Cauchy problem (2.7) with 0ρ<1/2<θ10\leq\rho<1/2<\theta\leq 1 and U0LmU_{0}\in L^{m} with m[1,2]m\in[1,2]. Then, the following refinement estimates hold:

χint(D)(U(t,)T1U~(t,;ρ,θ))H˙s(1+t)s22ρ2mm(22ρ)q(ρ,θ)U0Lm,\displaystyle\left\|\chi_{\operatorname{int}}(D)\left(U(t,\cdot)-T_{1}\widetilde{U}(t,\cdot;\rho,\theta)\right)\right\|_{\dot{H}^{s}}\lesssim(1+t)^{-\frac{s}{2-2\rho}-\frac{2-m}{m(2-2\rho)}-q(\rho,\theta)}\|U_{0}\|_{L^{m}},

where the function q=q(ρ,θ)q=q(\rho,\theta) is defined by

q(ρ,θ):={2θ122ρ,if ρ+θ<1,12ρ22ρ,if ρ+θ1,q(\rho,\theta):=\left\{\begin{aligned} &\frac{2\theta-1}{2-2\rho},&&\text{if }\rho+\theta<1,\\ &\frac{1-2\rho}{2-2\rho},&&\text{if }\rho+\theta\geq 1,\\ \end{aligned}\right. (4.22)

the matrix T1T_{1} is defined in (2.12).

Proof.

Here we only prove the case when ρ+θ<1\rho+\theta<1. For the other case when ρ+θ1\rho+\theta\geq 1, its proof is similar as the following discussion. Thus, we omit it.

First of all, let us apply the partial Fourier transform with respect to spatial variable such that W~(t,ξ;ρ,θ)=xξ(U~(t,x;ρ,θ))\widetilde{W}(t,\xi;\rho,\theta)=\mathcal{F}_{x\rightarrow\xi}(\widetilde{U}(t,x;\rho,\theta)) to get

W~(t,ξ;ρ,θ)=diag(eλ~j(|ξ|)t)j=14T11W0(ξ),\displaystyle\widetilde{W}(t,\xi;\rho,\theta)=\operatorname{diag}\left(e^{-\tilde{\lambda}_{j}(|\xi|)t}\right)_{j=1}^{4}T_{1}^{-1}W_{0}(\xi),

where

λ~1(|ξ|)=b2|ξ|22ρ,\displaystyle\tilde{\lambda}_{1}(|\xi|)=b^{2}|\xi|^{2-2\rho}, λ~2(|ξ|)=a2|ξ|22ρ,\displaystyle\tilde{\lambda}_{2}(|\xi|)=a^{2}|\xi|^{2-2\rho},
λ~3(|ξ|)=|ξ|2ρ+|ξ|2θb2|ξ|22ρ,\displaystyle\tilde{\lambda}_{3}(|\xi|)=|\xi|^{2\rho}+|\xi|^{2\theta}-b^{2}|\xi|^{2-2\rho}, λ~4(|ξ|)=|ξ|2ρ+|ξ|2θa2|ξ|22ρ.\displaystyle\tilde{\lambda}_{4}(|\xi|)=|\xi|^{2\rho}+|\xi|^{2\theta}-a^{2}|\xi|^{2-2\rho}.

We remark that λ~j(|ξ|)\tilde{\lambda}_{j}(|\xi|) are the principle parts of eigenvalues λj(|ξ|)\lambda_{j}(|\xi|) for j=1,,4j=1,\dots,4 (one may recall the statement of Proposition 2.1).
According to the representation of solutions for ξ𝒵int(ε)\xi\in\mathcal{Z}_{\operatorname{int}}(\varepsilon) in Proposition 2.1, we may obtain

χint(ξ)|ξ|s(W(t,ξ)T1W~(t,ξ;ρ,θ))=χint(ξ)|ξ|s(J1(t,|ξ|)+J2(t,|ξ|)+J3(t,|ξ|))W0(ξ),\displaystyle\chi_{\operatorname{int}}(\xi)|\xi|^{s}\left(W(t,\xi)-T_{1}\widetilde{W}(t,\xi;\rho,\theta)\right)=\chi_{\operatorname{int}}(\xi)|\xi|^{s}\left(J_{1}(t,|\xi|)+J_{2}(t,|\xi|)+J_{3}(t,|\xi|)\right)W_{0}(\xi),

where

J1(t,|ξ|)\displaystyle J_{1}(t,|\xi|) =T1diag(eλj(|ξ|)teλ~j(|ξ|)t)j=14T11,\displaystyle=T_{1}\operatorname{diag}\left(e^{-\lambda_{j}(|\xi|)t}-e^{-\tilde{\lambda}_{j}(|\xi|)t}\right)_{j=1}^{4}T_{1}^{-1},
J2(t,|ξ|)\displaystyle J_{2}(t,|\xi|) =T1N2(|ξ|)diag(eλj(|ξ|)t)j=14(I4×4+N2(|ξ|))1T11,\displaystyle=T_{1}N_{2}(|\xi|)\operatorname{diag}\left(e^{-\lambda_{j}(|\xi|)t}\right)_{j=1}^{4}(I_{4\times 4}+N_{2}(|\xi|))^{-1}T_{1}^{-1},
J3(t,|ξ|)\displaystyle J_{3}(t,|\xi|) =T1(I4×4+N2(|ξ|))diag(eλj(|ξ|)t)j=14N2(|ξ|)(I4×4+N2(|ξ|))1T11.\displaystyle=-T_{1}(I_{4\times 4}+N_{2}(|\xi|))\operatorname{diag}\left(e^{-\lambda_{j}(|\xi|)t}\right)_{j=1}^{4}N_{2}(|\xi|)(I_{4\times 4}+N_{2}(|\xi|))^{-1}T_{1}^{-1}.

Here the matrix N2(|ξ|)=𝒪(|ξ|12ρ)N_{2}(|\xi|)=\mathcal{O}\left(|\xi|^{1-2\rho}\right) is defined in (2.17). In the above equation we used

(It×4+N2(|ξ|))1=I4×4N2(|ξ|)(I4×4+N2(|ξ|))1.\displaystyle(I_{t\times 4}+N_{2}(|\xi|))^{-1}=I_{4\times 4}-N_{2}(|\xi|)(I_{4\times 4}+N_{2}(|\xi|))^{-1}.

We now begin to estimate J1(t,|ξ|)J_{1}(t,|\xi|) and J2(t,|ξ|)+J3(t,|ξ|)J_{2}(t,|\xi|)+J_{3}(t,|\xi|), respectively. By means value theorem, we know that

χint(ξ)|eλj(|ξ|)teλ~j(|ξ|)t|(1+t)χint(ξ)|ξ|1+2θ2ρeλ~j(|ξ|)t.\displaystyle\chi_{\operatorname{int}}(\xi)\left|e^{-\lambda_{j}(|\xi|)t}-e^{-\tilde{\lambda}_{j}(|\xi|)t}\right|\lesssim(1+t)\chi_{\operatorname{int}}(\xi)|\xi|^{1+2\theta-2\rho}e^{-\tilde{\lambda}_{j}(|\xi|)t}.

Thus,

χint(ξ)|ξ|sJ1(t,ξ)W0(ξ)L2\displaystyle\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s}J_{1}(t,\xi)W_{0}(\xi)\right\|_{L^{2}} (1+t)χint(ξ)|ξ|s+1+2θ2ρdiag(eλ~j(|ξ|)t)j=14W0(ξ)L2\displaystyle\lesssim(1+t)\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s+1+2\theta-2\rho}\operatorname{diag}\left(e^{-\tilde{\lambda}_{j}(|\xi|)t}\right)_{j=1}^{4}W_{0}(\xi)\right\|_{L^{2}}
(1+t)s22ρ2mm(22ρ)2θ122ρU0Lm.\displaystyle\lesssim(1+t)^{-\frac{s}{2-2\rho}-\frac{2-m}{m(2-2\rho)}-\frac{2\theta-1}{2-2\rho}}\|U_{0}\|_{L^{m}}.

Due to the fact that N2(|ξ|)=𝒪(|ξ|12ρ)N_{2}(|\xi|)=\mathcal{O}\left(|\xi|^{1-2\rho}\right), we have

χint(ξ)|ξ|s(J2(t,ξ)+J3(t,|ξ|))W0(ξ)L2\displaystyle\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s}(J_{2}(t,\xi)+J_{3}(t,|\xi|))W_{0}(\xi)\right\|_{L^{2}} χint(ξ)|ξ|s+12ρdiag(eλj(|ξ|)t)j=14W0(ξ)L2\displaystyle\lesssim\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s+1-2\rho}\operatorname{diag}\left(e^{-\lambda_{j}(|\xi|)t}\right)_{j=1}^{4}W_{0}(\xi)\right\|_{L^{2}}
(1+t)s22ρ2mm(22ρ)12ρ22ρU0Lm\displaystyle\lesssim(1+t)^{-\frac{s}{2-2\rho}-\frac{2-m}{m(2-2\rho)}-\frac{1-2\rho}{2-2\rho}}\|U_{0}\|_{L^{m}}

Summarizing the above estimates leads to

χint(ξ)|ξ|s(W(t,ξ)T1W~(t,ξ;ρ,θ))L2(1+t)s22ρ2mm(22ρ)2θ122ρU0Lm,\displaystyle\left\|\chi_{\operatorname{int}}(\xi)|\xi|^{s}\left(W(t,\xi)-T_{1}\widetilde{W}(t,\xi;\rho,\theta)\right)\right\|_{L^{2}}\lesssim(1+t)^{-\frac{s}{2-2\rho}-\frac{2-m}{m(2-2\rho)}-\frac{2\theta-1}{2-2\rho}}\|U_{0}\|_{L^{m}},

where we used our condition ρ+θ<1\rho+\theta<1.
Finally, applying the Parseval-Plancherel theorem, we complete the proof of the theorem. ∎

Theorem 4.2.

Let us consider the Cauchy problem (2.7) with 0ρ<1/2<θ10\leq\rho<1/2<\theta\leq 1 and U0L1,γU_{0}\in L^{1,\gamma} with γ(0,1]\gamma\in(0,1]. Then, the following refinement estimates hold:

χint(D)(U(t,)T1U~(t,;ρ,θ))H˙s(1+t)s+γ22ρ122ρq(ρ,θ)U0L1,γ,\displaystyle\left\|\chi_{\operatorname{int}}(D)\left(U(t,\cdot)-T_{1}\widetilde{U}(t,\cdot;\rho,\theta)\right)\right\|_{\dot{H}^{s}}\lesssim(1+t)^{-\frac{s+\gamma}{2-2\rho}-\frac{1}{2-2\rho}-q(\rho,\theta)}\|U_{0}\|_{L^{1,\gamma}},

where the function q=q(ρ,θ)q=q(\rho,\theta) is defined in (4.22) and the matrix T1T_{1} is defined in (2.12).

Proof.

We may immediately compete the proof of this result by following the procedure from the proofs of Theorems 3.2 and 4.1. ∎

Remark 4.1.

According to Theorems 3.1, 3.2, 4.1 and 4.2, the decay rate (1+t)q(ρ,θ)(1+t)^{-q(\rho,\theta)} can be gained by subtracting the solutions U~(t,x;ρ,θ)\widetilde{U}(t,x;\rho,\theta) for the reference systems (4.19), (4.20) and (4.21). From the value of q(ρ,θ)q(\rho,\theta), we also find that the threshold for diffusion structure is ρ+θ=1\rho+\theta=1.

5. Concluding remarks

Remark 5.1.

Let us discuss about smoothing effect of solutions. We first introduce the Gevrey space Γκ\Gamma^{\kappa} with κ[1,)\kappa\in[1,\infty) (see [32]), where

Γκ:={fL2:there exists a constant c such that exp(cξ1κ)(f)L2}.\displaystyle\Gamma^{\kappa}:=\left\{f\in L^{2}\,:\,\text{there exists a constant }c\text{ such that }\exp\left(c\langle\xi\rangle^{\frac{1}{\kappa}}\right)\mathcal{F}(f)\in L^{2}\right\}.

By using Proposition 2.2 with the same approach of [30], we immediately obtain the following results.

Theorem 5.1.

Let us consider the Cauchy problem (2.7) with 0ρ<1/2<θ<10\leq\rho<1/2<\theta<1 and U0L2U_{0}\in L^{2}. Then, the solutions satisfy |D|sU(t,)Γ122θ|D|^{s}U(t,\cdot)\in\Gamma^{\frac{1}{2-2\theta}} with s0s\geq 0. However, when 0ρ<1/2<θ=10\leq\rho<1/2<\theta=1, the solutions do not belong to any Gevrey space.

It is well-known that smoothing effect is mainly determined by asymptotic behavior of eigenvalues localized in large frequency zone (see Proposition 2.2). For this reason, we may observe smoothing effect is only influenced by the damping term (Δ)θut(-\Delta)^{\theta}u_{t} with θ(1/2,1]\theta\in(1/2,1] in the Cauchy problem (1.1).

Remark 5.2.

From [30, 5], we know the solution to elastic waves with friction utu_{t} does not have smoothing effect. However, in doubly dissipative elastic waves (1.1), the structural damping (Δ)θut(-\Delta)^{\theta}u_{t} with θ(1/2,1)\theta\in(1/2,1) brings Gevrey smoothing for the solutions even when ρ=0\rho=0.

Remark 5.3.

In the present paper we focus on energy estimates with initial data taking from HsLmH^{s}\cap L^{m} for s0s\geq 0, m[1,2]m\in[1,2] or from HsL1,γH^{s}\cap L^{1,\gamma} for γ(0,1]\gamma\in(0,1]. Here we restrict ourselves on estimating solutions in the L2L^{2} norm. For estimating the solutions in the LqL^{q} norm with 2q2\leq q\leq\infty, by applying Lemma 4.2 in [4], one may obtain LpLqL^{p}-L^{q} estimates with 1p2q1\leq p\leq 2\leq q\leq\infty and diffusion phenomena in a LpLqL^{p}-L^{q} framework.

Remark 5.4.

Our aim in this paper is to investigate dissipative structure and diffusion phenomena for doubly dissipative elastic waves (1.1) in two spaces dimensions, especially, we obtain a new threshold for diffusion structure. We think it is also possible to study three dimensional doubly dissipative elastic waves without any new difficulties. The crucial point is to derive asymptotic behavior of eigenvalues and representation of solutions by using suitable diagonalization procedure.

Acknowledgments

The PhD study of the author is supported by Sächsiches Landesgraduiertenstipendium.

References

  • [1] R.C. Charão, R. Ikehata, Decay of solutions for a semilinear system of elastic waves in an exterior domain with damping near infinity, Nonlinear Anal. 67 (2) (2007) 398–429.
  • [2] R.C. Charão, R. Ikehata, Energy decay rates of elastic waves in unbounded domain with potential type of damping, J. Math. Anal. Appl. 380 (1) (2011) 46–56.
  • [3] W. Chen, Decay properties and asymptotic profiles for elastic waves with Kelvin-Voigt damping in 2D, Preprint (2018).
  • [4] W. Chen, Cauchy problems for thermoelastic plate equations with different damping mechanisms, Preprint (2019).
  • [5] W. Chen, M. Reissig, Weakly coupled systems of semilinear elastic waves with different damping mechanisms in 3D, Math. Methods Appl. Sci. 42 (2) (2019) 667–709.
  • [6] M. D’Abbicco, L1L1L^{1}-L^{1} estimates for a doubly dissipative semilinear wave equation, NoDEA Nonlinear Differential Equations Appl. 24 (1) (2017) 23 pp.
  • [7] M. D’Abbicco, M.R. Ebert, Diffusion phenomena for the wave equation with structural damping in the LpLqL^{p}-L^{q} framework, J. Differential Equations 256 (7) (2014) 2307–2336.
  • [8] M. D’Abbicco, M. Reissig, Semilinear structural damped waves, Math. Methods Appl. Sci. 37 (11) (2014) 1570–1592.
  • [9] T. Hosono, T. Ogawa, Large time behavior and LpL^{p}-LqL^{q} estimate of solutions of 2-dimensional nonlinear damped wave equations, J. Differential Equations 203 (1) (2004) 82–118.
  • [10] R. Ikehata, New decay estimates for linear damped wave equations and its application to nonlinear problem, Math. Methods Appl. Sci. 27 (8) (2004) 865–889.
  • [11] R. Ikehata, Asymptotic profiles for wave equations with strong damping, J. Differential Equations, 257 (6) (2014) 2159–2177.
  • [12] R. Ikehata, R.C. Charão, C.R. da Luz, Optimal decay rates for the system of elastic waves in 𝐑n\mathbf{R}^{n} with structural damping, J. Evol. Equ. 14 (1) (2014) 197–210.
  • [13] R. Ikehata, H. Michihisa, Moment conditions and lower bounds in expanding solutions of wave equations with double damping terms, Preprint (2018).
  • [14] R. Ikehata, M. Onodera, Remarks on large time behavior of the L2L^{2}-norm of solutions to strongly damped wave equations, Differential Integral Equations 30 (7-8) (2017).
  • [15] R. Ikehata, A. Sawada, Asymptotic profile of solutions for wave equations with frictional and viscoelastic damping terms, Asymptot. Anal. 98 (1-2) (2016) 59–77.
  • [16] R. Ikehata, H. Takeda, Critical exponent for nonlinear wave equations with frictional and viscoelastic damping terms, Nonlinear Anal. 148 (2017) 228–253.
  • [17] R. Ikehata, H. Takeda, Asymptotic profiles of solutions for structural damped wave equations, J. Dynam. Differential Equations 31 (1) (2019) 537–571.
  • [18] R. Ikehata, H. Takeda, Large time behavior of global solutions to nonlinear wave equations with frictional and viscoelastic damping terms, Osaka Math. J. (in press).
  • [19] K. Jachmann, A unified treatment of models of thermoelasticity, PhD Thesis, TU Bergakademie Freiberg, 2008.
  • [20] K. Jachmann, Cauchy problem for linear thermoelastic systems. II. Applications, Int. J. Dyn. Syst. Differ. Equ. 2 (3-4) (2009) 202–222.
  • [21] K. Jachmann, M. Reissig, Cauchy problem for linear thermoelastic systems. I. A unified approach, Int. J. Dyn. Syst. Differ. Equ. 2 (3-4) (2009) 170–201.
  • [22] G. Karch, Selfsimilar profiles in large time asymptotics of solutions to damped wave equations, Studia Math. 143 (2) (2000) 175–197.
  • [23] P. Marcati, K. Nishihara, The LpL^{p}-LqL^{q} estimates of solutions to one-dimensional damped wave equations and their application to compressible flow through porous media, J. Differential Equations 191 (2) (2003) 445-469.
  • [24] A. Matsumura, On the asymptotic behavior of solutions of semi-linear wave equations, Publ. Res. Inst. Math. Sci. 12 (1) (1976/77) 169–189.
  • [25] H. Michihisa, L2L^{2} asymptotic profiles of solutions to linear damped wave equations, Preprint (2017).
  • [26] H. Michihisa, Expanding methods for evolution operators of strongly damped wave equations, Preprint (2018).
  • [27] T. Narazaki, LpL^{p}-LqL^{q} estimates for damped wave equations and their applications to semi-linear problem, J. Math. Soc. Japan 56 (2) (2004) 585–626.
  • [28] K. Nishihara, LpL^{p}-LqL^{q} estimates of solutions to the damped wave equation in 3-dimensional space and their application, Math. Z. 244 (3) (2003) 631–649.
  • [29] G. Ponce, Global existence of small solutions to a class of nonlinear evolution equations, Nonlinear Anal. 9 (5) (1985) 399–418.
  • [30] M. Reissig, Structurally damped elastic waves in 2D, Math. Methods Appl. Sci. 39 (15) (2016) 4618–4628.
  • [31] M. Reissig, Y.G. Wang, Cauchy problems for linear thermoelastic systems of type III in one space variable, Math. Methods Appl. Sci. 28 (11) (2005) 1359–1381.
  • [32] L. Rodino, Linear partial differential operators in Gevrey spaces, World Scientific Publishing Co., Inc., River Edge, NJ, 1993.
  • [33] Y. Shibata, On the rate of decay of solutions to linear viscoelastic equation, Math. Methods Appl. Sci. 23 (3) (2000) 203–226.
  • [34] H. Takeda, Higher-order expansion of solutions for a damped wave equation, Asymptot. Anal. 94 (1-2) (2015) 1–31.
  • [35] M. Wu, S. Chai, S. Li, Energy decay rates for the elasticity system with structural damping in the Fourier space, J. Math. Anal. Appl. 452 (1) (2017) 361–377.
  • [36] K. Yagdjian, The Cauchy problem for hyperbolic operators. Multiple characteristics. Micro-local approach, Mathematical Topics, Akademie Verlag, Berlin, 1997.