Universität Konstanz
Global classical solution for one-dimensional nonlinear thermoelastiticity with second sound on the semi-axis
Yuxi Hu
Konstanzer Schriften in Mathematik Nr. 280, Juli 2011
ISSN 1430-3558
© Fachbereich Mathematik und Statistik Universität Konstanz
Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-140667
Global classical solution for one-dimensional
nonlinear thermoelasticity with second sound on the semi-axis
Yuxi HU
∗Department of Mathematics Shanghai Jiao Tong University
200240 Shanghai, P.R.China
Abstract
In this paper, we will give a global existence theorem in one-dimensional thermoelasticity with second sound inR+. For this purpose, we first give the decay rates of the linearized equations with the help of Fourier sine and cosine transformation and the local existence theorem using theorems for quasi-linear symmetric hyperbolic systems. Then we establish some estimates inL2, L1 andL∞ norms to get a uniform a priori estimate. Finally, we use the usual continuation argument to get global solution.
Keywords: second sound, linear decay rates, semi-axis, global solution
1 Introduction
The equations of thermoelasticity describe the elastic and the thermal behavior of elastic, heat con- ductive media, in particular the reciprocal actions between elastic stresses and temperature differences.
The common modeling of heat conduction using Fourier law, essentially saying that the heat flux is a certain function of the gradient of the temperature, leads to the well known paradox of infinite propaga- tion of signals, in particular of heat signals. Many efforts are made to remove this paradox (see [3] [4]
[23]). The common feature of these efforts is that all lead to hyperbolic equation rather than parabolic equation and thus has limit speed of propagation. In our case, we consider Cattaneo’s law instead of Fourier’s law and thus get a hyperbolic system.
The general equations for nonlinear thermoelasticity described by the displacement vectoru=u(t, x), the temperatureT =T(t, x) and the heat fluxq=q(t, x), are as follows
ρutt− ∇′S=ρb, (1.1)
εt−tr{SFt}+∇′q=r, (1.2)
where ρ is the material density in a domain Ω ofRn, n = 1,2,3, S is the Piola-Kirchhoff stress tensor and b is the specific external body force, ε is the internal energy, q is the heat flux, r is the external heat supply, while∇′ denotes the divergence operator, trB denotes the trace of a matrixBandF is the deformation gradient
F = 1 +∇u.
The two basic nonlinear differential equations arise from the balance of linear momentum and the balance of energy. For the derivation of the above equations, see Racke and Jiang [10] and Carlson [2].
∗Exchange student from Shanghai Jiaotong University to Konstanz University; Email:huyuxi@sjtu.edu.cn.
We denote byη the entropy and by
ψ:=ε−T η
the Helmholtz free energy. The constitutive assumptions in thermoelasticity with second sound are that S, ε, η, ψare functions of (∇u, θ,∇θ, q) whereθ=T−T0denotes the temperature difference,T0>0 the constant reference temperature. It is always assumed that these functions are smooth and that
detF ̸= 0.
With the help of the second law of thermodynamics, it turns out ψ=ψ(∇u, θ, q) is independent of∇θ and
η=η(∇u, θ, q) =−ψθ(∇u, θ, q), S=S(∇u, θ, q) =ψ∇u(∇u, θ, q).
So, we can rewrite (1.2) as
(θ+T0)(tr{−ψθ∇u∇ut} −ψθθθt) +∇′q= (θ+T0)ψθqqt−ψqqt. (1.3) If the thermal behavior is described by Fourier’s law, that is,
q=−κθx, (1.4)
then (1.1) and (1.2) constitute the system of classical thermoelasticity. While, the heat flux replacing Fourier’s law by Cattaneo’s law, is given by the following equation
τ(∇u, θ)qt+q+κ(∇u, θ)∇θ= 0, (1.5) where τ is the relaxation time describing the time lag in the response of the heat flux to a gradient in the temperature andκdenotes the heat conductivity tensor. The equations (1.1), (1.3), (1.5) constitute the system of nonlinear thermoelasticity with second sound.
For classical thermoelasticity, many results have been obtained. See [19], [20] [5], [6], [21]. For thermoelasticity with second sound, we already have some results. In one dimension case, Tarabek [24]
had proved a global well posedness for cauchy problem and decay to an equilibrium. Wang and Racke [17]
gave the details for local existence and decay rates. The exponential stability was obtained by Racke [18]
for bounded domain. Extensions were given by Messaoudi and Said-Houari [16]. In three dimension case, the global well-posedness and the exponential stability were given by Irmscher [8] for radially symmetric domain. However, the fully nonlinear Cauchy problem and exterior domains have not yet been treated.
My work here is to be able to prove a global existence theorem for an initial boundary value problem in R+ for nonlinear thermoelasticity with second sound.
In one space dimension, (1.1), (1.3), (1.5) turn into
ρutt−auxx+bθx=−Sqqx+ρb, (θ+T0)(˜aθt+buxt) +qx=dqt+r, τ qt+q+κθx= 0,
(1.6)
with
a(∇u, θ, q) =ψuxux, b=−ψuxθ,˜a=−ψθθ, d= (θ+T0)ψθq−ψq.
In our case, we assume there’s no external energy term, that is,b =r= 0. Without loss of generality, we assumeρ= 1. And for simplicity, we also assumeSq =d= 0 and that Cattaneo’s law is linear, that is,τ, κare positive constants. So the above equations become
utt−a(∇u, θ, q)uxx+b(∇u, θ, q)θx= 0,
(θ+T0)(˜a(∇u, θ, q)θt+b(∇u, θ, q)uxt) +qx= 0, τ qt+q+κθx= 0,
(1.7)
with initial and boundary conditions {
u(0,·) =u0, ut(0,·) =u1, θ(0,·) =θ0, q(0,·) =q0, ∀x∈Ω = [0,¯ +∞) ux|∂Ω=θ|∂Ω= 0.
Letω=ux, v=ut, then system (1.7) becomes
ωt−vx= 0, (1.8)
vt−a(ω, θ, q)ωx+b(ω, θ, q)θx= 0, (1.9)
˜
a(ω, θ, q)θt+b(ω, θ, q)vx+c(θ)qx= 0, (1.10)
τ qt+q+κθx= 0, (1.11)
wherec(θ) = θ+T1
0(|θ| ≤K < T0will be a posterior estimate justified by the global small solution). With initial and boundary conditions
ω(0,·) = (u0)x=:ω0, v(0,·) =u1=:v0, θ(0,·) =θ0, q(0,·) =q0, (1.12)
ω|∂Ω=θ|∂Ω= 0. (1.13)
Some smooth assumptions on the coefficients of equations (1.8)-(1.11) are needed.
Assumption 1.1. a , b ,˜a , careC3−functions of their arguments and there exist positive constantsγ0, γ1 andK < T0 such that if |ω|,|θ|,|q| ≤K, then
γ0≤a(ω, θ, q),|b(ω, θ, q)|,˜a(ω, θ, q), c(θ)≤γ1. (1.14) We introduce some notations which will be frequently used throughout the paper. For a non-negative integer N, let
DNu= ∑
l+m=N
∂tl∂mxu.
We denote by Wm,p(Ω),0 ≤ m ≤ ∞,1 ≤p ≤ ∞, the usual Sobolev space with the norm ∥ · ∥Wm,p. For convenience, Hm(Ω) and Lp(Ω) stand for Wm,2(Ω) and W0,p(Ω) respectively. Let X be a Banach space. We denote byLp([α, β], X) (1≤p≤ ∞) and∥ · ∥Lp([α,β,X])the space of all measurablep-th power functions from [α, β] to X and its norm respectively. ForT >0, we use the notation
∥u∥p,T ,k= sup
0≤t≤T
(1 +t)k∥u(t)∥Lp,1≤p≤ ∞, k≥0.
Throughout this paperCorC′ will denote a general positive constant which is not necessarily the same in any two places. ForU = (ω, v, θ, q) a function oft andx, we denote
K(U) =
0 −∂x 0 0
−a∂x 0 b∂x 0 0 ˜ab∂x 0 ac˜∂x
0 0 κτ∂x 1 τ
U, E0=
( 1 0 0 0 0 0 1 0
) .
Define (∂tkU(0, x) fort≥0 recursively by
(∂tkU)(0, x) =∂tk−1(−K(U)U)(0, x), k≥1, U(0, x) =U0(x) = (ω0, v0, θ0, q0). (1.15) Furthermore, we need the smooth assumptions on initial data to ensure global solution.
Assumption 1.2. (1) Uk0∈H3−k(Ω), (regularity on initial data) (2)E0·Uk0|∂Ω= 0, (compatibility condition)
whereUk0=∂ktU(0, x)andk= 0,1,2,3.
This paper is mainly motivated by Jiang’s paper [9]. In that paper, he was able to prove a global solution for equations of classical one-dimensional thermoelasticity inR+for small smooth data. It seems that many results in classical thermoelasticity can be extended to thermoelasticity with second sound, see [17], [18] and [20]. However, it is not true, for example, for Timoshenko-type thermoelastic systems, where a system can be or remain exponentially stable under Fourier’s law, while it loses this property under Cattaneo’s law, see [7]. Our question is that whether a weak damping effect given by equation (1.5) is still predominating to ensure decay rates and global solution compared with a strong impact of dissipation induced by equation (1.4).
We arrange our paper as follows. In section 2, we give a brief introduction to our problem and then study the behavior of the linearized system. Some comparison with Jiang’s paper[9] will be given there.
In section 3, we study local existence theorem based on a local existence theorem for hyperbolic systems.
In section 4, we use energy methods to studyL2 estimates for the local solution. In section 5, we obtain L1 andL∞estimates based on decay behavior of the linearized system and finally get global solution by usual continuation arguments.
2 Linear Decay Rates
The methods used for obtaining global existence theorems for small data consist of proving suitable a priori estimate, where one often exploits the decay of solutions to the linearized equations. This requires a precise analysis of the asymptotic behavior of such solutions as time tend to infinity, which will finally allow us to describe the asymptotic behavior of solutions to the nonlinear systems as well. In this section, we will give the decay rates of the corresponding linear equations to our system (1.8)-(1.11). Denote
a0=a(0,0,0), b0=b(0,0,0), c0=c(0),˜a0= ˜a(0,0,0),
where a, b, c,a˜ are defined in section 1. So, by Assumption 1.1, we knowa0 >0, b0 ̸= 0, c0 >0,a˜0 >0.
Then we rewrite equations (1.8)-(1.11) as follows
7.7
ωt−vx= 0,
vt−a0ωx+b0θx= (a−a0)ωx+ (b0−b)θx,
˜
a0θt+b0vx+c0qx= (˜a0−˜a)θt+ (b0−b)vx+ (c0−c)qx, τ qt+q+κθx= 0.
(2.16)
Take transformation byU∗= (ω∗, v∗, θ∗, q∗) = (√a0ω, v,√
˜
a0θ,√τ c0
κ q), system (??) becomes
ω∗t − √a0v∗x= 0, v∗t − √a0ω∗x+√b0
˜
a0θx∗=f(U∗), θ∗t+√b0
˜
a0vx∗+√
κc0
τ0˜a0q∗x=g(U∗), qt∗+1τq∗+√
κc0
τ0˜a0θ∗x= 0,
where
f(U∗) =(a√−aa00)ωx∗+(b√0−b)
˜ a0 θx∗, g(U∗) =
[(˜a0−˜a)
√˜a0 θt∗+ (b0−b)v∗x+ (c0−c)√
κ τ c0qx∗
]√
˜ a0−1. Let α = √
a0, β = √b0
˜ a0, γ =
√κc0
τ˜a0. We still denote (ω∗, v∗, θ∗, q∗) by (ω, v, θ, q). So, the linearized equations wheref(U∗) =g(U∗) = 0 are as follows:
wt−αvx= 0, (2.1)
vt−αωx+βθx= 0, (2.2)
θt+βvx+γqx= 0, (2.3)
qt+1
τq+γθx= 0, (2.4)
with boundary and initial conditions
ω|∂Ω=θ|∂Ω= 0, t≥0 (2.5)
(ω(0, x), v(0, x), θ(0, x), q(0, x))′ = (ω0, v0, θ0, q0)′≡U0, x∈Ω (2.6) where Ω = (0,∞) andα, γ, τ >0, β̸= 0 are constants.
We denote byFs(f(t, x)),Fc(f(t, x)) the Fourier sine and cosine transform of f(t, x) with respect to x, i.e.
Fs(f(t, x))(ξ) =
√2 π
∫ ∞
0
f(t, x) sinξxdx, Fc(f(t, x))(ξ) =
√2 π
∫ ∞
0
f(t, x) cosξxdx.
Taking Fourier sine transform on both sides of (2.1) and (2.3) and Fourier cosine transform on both sides of (2.2) and (2.4), then we get
Uˆt+A(ξ) ˆU = 0, t >0, (2.7) Uˆ(0) = (Fs(ω0),Fc(v0),Fs(θ0),Fc(q0))′, (2.8) where ˆU = ˆU(t, ξ) = (Fs(ω(x, t))(ξ),Fc(v(x, t))(ξ),Fs(θ(x, t))(ξ),Fc(q(x, t))(ξ))′ and
A(ξ) =
0 αξ 0 0
−αξ 0 βξ 0 0 −βξ 0 −γξ
0 0 γξ τ1
.
Let det(λI+A(ξ)) be zero, then we get the following characteristic equation λ4+1
τλ3+ (α2+β2+γ2)ξ2λ2+α2+β2
τ ξ2λ+α2γ2ξ4= 0. (2.9) Using formulae solving quartic equations, we have
λ1=
√a1+2y+
√
−3a1−2y−√a2a2
1 +2y
2 −4τ1,
λ2=
√a1+2y−√
−3a1−2y−√2a2
a1 +2y
2 −4τ1,
λ3=
−√ a1+2y+
√−3a1−2y+√a2a2
1 +2y
2 −4τ1,
λ4=
−√a1+2y−√
−3a1−2y+√a2a2
1 +2y
2 −4τ1,
(2.10)
where
a1= (α2+β2+γ2)ξ2−8τ32, a2= α2+β2τ2−γ2ξ2+8τ13,
a3=α2γ2ξ4+γ2−3α16τ2−23β2ξ2−256τ3 4, and
y=−56a1+ (−Q2 +√
∆)13 + (−Q2 −√
∆)13,
P = [−121(α2+β2+γ2)2−α2γ2]ξ4+α24τ+β22ξ2=:P1ξ4+P2ξ2,
Q= [−1081 (α2+β2+γ2)3+13(α2+β2+γ2)α2γ2]ξ6+−2α4−2β4−4α24τ2β22+β2γ2−2α2γ2ξ4=:Q1ξ6+Q2ξ2
∆ = Q42 +P273.
In order to getL1 decay estimates, we first investigate the low and high frequency expansions of eigen- values in (2.10) in terms ofξ, which are expressed in the following two Lemmas.
Lemma 2.1. (i)λ1(0) =−1τ, λj(0) = 0 (j= 2,3,4), and asξ→0, we have
λ1(ξ) =−1τ+O(ξ), λ2(ξ) =−ατ α2+β2γ22ξ2+O(ξ3), λ3(ξ) =−2(ατ β22+βγ22)ξ2+i√
α2+β2ξ+O(ξ3), λ4(ξ) =−2(ατ β22+βγ22)ξ2−i√
α2+β2ξ+O(ξ3),
(2.11)
and
dλ2
dξ +2τ αα2+β2γ22ξ=O(ξ2),
dλ3 dξ −i√
α2+β2=O(ξ),
dλ4 dξ +i√
α2+β2=O(ξ),
(2.12)
and asξ→ ∞, we have
λ1=−2τ1 −a−bξ−2+O(ξ−4) +i(eξ+f ξ−1+O(ξ−3)), λ2=−2τ1 −a−bξ−2+O(ξ−4)−i(eξ+f ξ−1+O(ξ−3)), λ3=a+bξ−2+O(ξ−4) +i(cξ+dξ−1+O(ξ−3)), λ4=a+bξ−2+O(ξ−4)−i(cξ+dξ−1+O(ξ−3)),
(2.13)
and
dλ1
dξ =ie+O(ξ−2),
dλ2
dξ =−ie+O(ξ−2),
dλ3
dξ =ic+O(ξ−2),
dλ4
dξ =−ic+O(ξ−2),
(2.14)
where
a= 1 4τ
[
2γ2 α2+β2+γ2+√
(α2+β2+γ2)−4α2γ2 −1 ]
,
c=
√
α2+β2+γ2−√
(α2+β2+γ2)−4α2γ2
2 ,
e=
√
α2+β2+γ2+√
(α2+β2+γ2)−4α2γ2
2 .
(ii)Except at most finite values ofξ >0, λj̸=λi(i̸=j, i, j= 1,2,3,4).
Proof. Ifξ→0,
∆ = Q2 4 +P3
27 =1
4(Q1ξ6+Q2ξ4)2+ 1
27(P1ξ4+P2ξ2)3
= 1
4(Q21ξ12+ 2Q1Q2ξ10+Q22ξ8) + 1
27(P13ξ12+ 3P12P2ξ10+ 3P1P22ξ8+P23ξ6)
= (Q21 4 +P13
27)ξ12+ (Q1Q2
2 +P12P2
9 )ξ10+ (Q22
4 +P1P22
9 )ξ8+ 1 27P23ξ6. Using the formulae (1 +x)α= 1 +αx+O(x2) asx→0, we get√
∆ =
√P23
27ξ3+O(ξ5). Then we have
−Q2 +√
∆ =
√P23
27ξ3−Q22ξ4+O(ξ5) and (−Q2 +√
∆)13 =
√P2
3ξ− 2PQ22ξ2+O(ξ3). According to the definition ofy, we gety=−56a1−2QP22ξ2 and therefore get
−3a1−2y−2a2(a1+ 2y)−12 =−4(α2+β2)ξ2+O(ξ3),
and
−3a1−2y+ 2a2(a1+ 2y)−12 = 1
τ2+ ( 2β2γ2
α2+β2 −4γ2)ξ2+O(ξ3).
So, we conclude that asξ→0,
λ1(ξ) =−1τ +O(ξ), λ2(ξ) =−ατ α2+β2γ22ξ2+O(ξ3), λ3(ξ) =−2(ατ β22+βγ22)ξ2+i√
α2+β2ξ+O(ξ3), λ4(ξ) =−2(ατ β22+βγ22)ξ2−i√
α2+β2ξ+O(ξ3).
Asξ→ ∞, we get as before
∆ = Q2 4 +P3
27 = (Q21 4 +P13
27)ξ12+ (Q1Q2
2 +P12P2
9 )ξ10+ (Q22
4 +P1P22
9 )ξ8+ 1 27P23ξ6. Since Q421 +P2713 =−α2γ2[(α2+β2+γ2)2−4α2γ2]2
432 <0, we get asξ→ ∞ (−Q
2 +
√Q2 4 +P3
27)13 = (−Q 2 +i
√
−(Q2 4 +P3
27))13 = [Q2
4 −(Q2 4 +P3
27) ]16
eiθ31 =
√
−P 3eiθ31, whereθ1 satisfies tanθ1=
√
−(Q42+P273)
−Q2 .Then after some calculations, we get cosθ1= 3√
3Q1
2P1
√−P1
(1 +1 2(2Q2
Q1 −3P2
P1
)ξ−2+O(ξ−4), sinθ1=
√
4P13+ 27Q21
4P13 +27Q21
8P13 (4P13+ 27Q2
4P13 )−12(2Q2
Q1 −3P2
P1
)ξ−2+O(ξ−4).
Let
θ11= 3√ 3Q1
2P1√
−P1, θ12== 3√ 3Q1
4P1√
−P1(2Q2
Q1 −3P2
P1 ), θ13=
√
4P13+ 27Q21
4P13 , θ14=27Q21
8P13 (4P13+ 27Q2
4P13 )−12(2Q2 Q1 −3P2
P1
).
We calculate cosθ31 in the following formula:
cosθ1 3 = 1
2[(cosθ1+isinθ1)13+ (cosθ1−isinθ1)13] =m+nξ−2+O(ξ−4),
where {
m=12 [
(θ11+iθ13)13 + (θ11−iθ13)13 ]
,
n= 16[(θ11+iθ13)−23(θ12+iθ14) + (θ11−iθ13)−23(θ12−θ14)].
Therefore, we geta1+ 2y=c1ξ2+c2+O(ξ−2) where
c1=
[
−23(α2+β2+γ2) + 4
√−P31m ]
<0, c2=4τ12 + 2
√−P31PP21m+ 4n
√−P31.
So, we get√
a1+ 2y=i[g1ξ+g2ξ−1+O(ξ−3)] and (a1+ 2y)−12 =−i[f1ξ−1+f2ξ−3+O(ξ−5)] where
g1=√
−c1= [
2
3(α2+β2+γ2)−4
√−P31m ]12
, g2=−2√c−2c
1 =−12 [
2
3(α2+β2+γ2)−4
√−P31m ]−12
(4τ12 + 2
√−P31PP21m+ 4n
√−P31),
f1= 23(α2+β2+γ2)−4
√−P31m,
f2= 12 [
2
3(α2+β2+γ2)−4
√−P31m ]−32
(4τ12 + 2
√−P31PP21m+ 4n
√−P31).
Based on above calculations, we have
−3a1−2y− 2a2
√a1+ 2y =−2 [
(α2+β2+γ2)ξ2− 3 8τ2
]
−(c1ξ2+c2+O(ξ−2)) + 2i( 1
8τ3+α2+β2−γ2
2τ ξ2)(f1ξ−1+f2ξ−3+O(ξ−5)
=r1ξ2+s1+O(ξ−2) +i(t1ξ+u1ξ+O(ξ−3)), where
r1=−2(α2+β2+γ2)−c1, r2= 3
4τ2−c2, t1= α2+β2−γ2
τ f1, u1= α2+β2−γ2
τ + 1
8τ3f1. So, we have
(−3a1−2y− 2a2
√a1+ 2y)12 =[
(r1ξ2+s1+O(ξ−2))2+ (t1ξ+u1ξ−1+O(ξ−3))2]14 eiθ2/2,
= (√
−r1ξ+2r1s1+t1 2r132
ξ−1+O(ξ−3)) cosθ2 2 +i(√
−r1ξ+2r1s1+t1 2r132
ξ−1+O(ξ−3)) sinθ2 2 , whereθ2 satisfies tanθ2=t1rξ+u1ξ−1+O(ξ−3)
1ξ2+s1+O(ξ−2) . Sincer1<0 andt1>0, we get π2 < θ2< π and cosθ2
2 = t1
2r1
ξ−1−1 2(7t41
8r41 +u1r1−t1s1
r21 )ξ−3,sinθ2
2 = 1−1 8
t21
r12ξ−2+O(ξ−4).
As a result, we have
(−3a1−2y− 2a2
a1+ 2y)12 =v1+w1ξ−2+O(ξ−4) +i(v2ξ+w2ξ−1+O(ξ−3)), where
v1= t1
2r
3 2
1
, w1= −7t41+ 4(2r1s1+t1)t1
16(−r1)72 −r1u1−t1s1
2(−r1)32 , v2=√
−r1, w2=− t21 8r
3 2
1
+2r1s1+t1
2r
3 2
1
. Finally we derive the formulae forλi asξ→ ∞
λ1= (v21 −4τ1) +w21ξ−2+O(ξ−4) +i(v2+g2 1ξ+w2+g2 2ξ−1+O(ξ−3)), λ2= ¯λ1,
λ3= (−v21 −4τ1)−w21ξ−2+O(ξ−4) +i(g1−2v2ξ+g2−2w2ξ−1+O(ξ−3)), λ4= ¯λ3.
For simplicity, we rewrite the above form as follows:
λ1=−2τ1 −a−bξ−2+O(ξ−4) +i(eξ+f ξ−1+O(ξ−3)), λ2=−2τ1 −a−bξ−2+O(ξ−4)−i(eξ+f ξ−1+O(ξ−3)), λ3=a+bξ−2+O(ξ−4) +i(cξ+dξ−1+O(ξ−3)), λ4=a+bξ−2+O(ξ−4)−i(cξ+dξ−1+O(ξ−3)).
asξ→ ∞ (2.15)
The formulae for a, b, c, c, e, f are very complicated in our case, we use the formulae giving roots to calculate some of them, though these formulae are not crucial in the proof of Theorem 2.7 below. We know that the roots satisfy the following equation:
λ1+λ2+λ3+λ4=−τ1,
λ1λ2+λ1λ3+λ1λ4+λ2λ3+λ2λ4+λ3λ4= (α2+β2+γ2)ξ2, λ1λ2λ3+λ1λ2λ4+λ1λ3λ4+λ2λ3λ4=−α2+βτ 2ξ2,
λ1λ2λ3λ4=α2γ2ξ4.
So, the coefficientsa, c, esatisfy the following equation e2c2=α2γ2, e2(2a) +c2(−1
τ −2a) =−α2+β2
τ , e2+c2= (α2+β2+γ2).
Therefore, we have
a= 1 4τ
[
2γ2 α2+β2+γ2+√
(α2+β2+γ2)−4α2γ2 −1 ]
,
c=
√
α2+β2+γ2−√
(α2+β2+γ2)−4α2γ2
2 ,
e=
√
α2+β2+γ2+√
(α2+β2+γ2)−4α2γ2
2 .
Remark 2.1. We note that the characteristic matrixA(ξ)is a4×4order matrix, which causes technical complications in the discussion of the eigenvalues. However, it is important to note that when τ goes to zero, the three eigenvalues λj(j= 2,3,4)are exactly the same with these in Jiang’s paper [9]. This fact is reasonable since when the parameter τ goes to zero, our system are essentially reduced to the classical system of thermoelasticity.
Lemma 2.2. (i) For any values ofξ withξ >0,Reλj(ξ)<0. (j= 1,2,3,4)
(ii) There exists constantsr1, r2 andCn(n= 1,2,3,4) depending onr1, r2 such that for j= 2,3,4
−C2|ξ|2≤Reλj(ξ)≤ −C1|ξ|2, 0≤ξ≤r1, Reλj(ξ)≤ −C3, r1≤ξ≤r2, Reλj(ξ)≤ −C4, ξ≥r2.
(2.16)
Proof. (i)Since v′(−A(ξ))v = −1τv42 ≤ 0 for any v = (v1, v2, v3, v4) ∈ R4, we know that −A(ξ) is dissipative for anyξ >0. Therefore, we derive that
Reλj(ξ)≤0, ∀ξ >0, j= 1,2,3,4.
Second we claim that there is no purely imaginary root for the characteristic equation (2.9). Otherwise, there exist someξ0>0 andj such thatReλj(ξ0) = 0. Then take λj into (2.9), we get
a4−1
τa3i−(α2+β2+γ2)ξ02a2+α2+β2
τ ξ02ai+α2γ2ξ04= 0.