Computers and Mathematics with Applications 66 (2013) 4155

Contents lists available at SciVerse ScienceDirect

Computers and Mathematics with Applications

journal homepage: www.elsevier.com/locate/camwa

Decay of solutions for a mixture of thermoelastic

one dimensional solids
Jaime E. Muoz Rivera a,b , Maria Grazia Naso c, , Ramon Quintanilla d
a
National Laboratory for Scientific Computation, Rua Getulio Vargas 333, Quitadinha-Petrpolis 25651-070, Rio de Janeiro, RJ, Brazil
b
Instituto de Matemtica - UFRJ, Av. Horcio Macedo, Centro de Tecnologia, Cidade Universitria - Ilha do Fundo 21941-972,
Rio de Janeiro, RJ, Brazil
c
Dipartimento di Ingegneria Civile, Architettura, Territorio, Ambiente e di Matematica, Universit degli Studi di Brescia, Via Valotti 9,
25133 Brescia, Italy
d
Matemtica Aplicada 2, UPC C. Coln 11, 08222 Terrassa, Barcelona, Spain

article info abstract

Article history: We study a PDE system modeling thermomechanical deformations for a mixture of ther-
Received 4 October 2012 moelastic solids. In particular we investigate the asymptotic behavior of the solutions. First,
Received in revised form 19 February 2013 we identify conditions on the constitutive coefficients to guarantee that the imaginary axis
Accepted 27 March 2013
is contained in the resolvent. Subsequently, we find the necessary and sufficient conditions
to guarantee the exponential decay of solutions. When the decay is not of exponential type,
Keywords:
we prove that the solutions decay polynomially and we find the optimal polynomial decay
Thermoelastic mixtures
Exponential decay
rate.
Weakly coupled system 2013 Elsevier Ltd. All rights reserved.

1. Introduction

In recent years an increasing interest has been directed to understand the so-called nonclassical thermomechanical
theories of materials as micromorphic materials, porous materials, etc. One of them is the mixture of materials. The origin of
the modern formulation of continuum thermomechanical theories of mixtures goes back to the papers by Kelly [1], Truesdell
and Toupin [2], Eringen and Ingram [3,4], Mller [5], Green and Naghdi [6,7], Bowen [8], Bowen and Wiese [9], Atkin and
Craine [10] and Bowen and Drumheller [11]. Recent books on this subject are given by Samohyl [12] and Rajagopal and
Tao [13]. For several theories of mixture of materials [8,6,14], the spatial description is used and the independent variables
are the displacement gradients and the relative velocity. The first theory, where the Lagrangian description is proposed and
where the independent variables are the displacement gradients and the relative displacement, was presented in the works
by Bedford and Stern [15,16]. These models have been applied by Tiersten and Jahanmir [17] to derive a theory of composites
where the relative displacement of the individual constituents is infinitesimal. Theory of mixtures is amply accepted in the
scientific community and it has a wide field of applications.
Mathematical studies for this theory have been developed. Questions concerning the existence, uniqueness, continuous
dependence and asymptotic behavior have been studied in [1824]. We believe that the mathematical and physical studies
are necessary issues to clarify the scope of its applicability. We here want to deep the study of the time decay of solutions
for the one-dimensional mixture of two thermoelastic solids. Our aim is to clarify when we can expect the exponential or
polynomial decay for the solutions of our problem. From a mathematical point of view the question is of interest. We have
the coupling of two conservative equations with one dissipative equation. We want to see when the coupling between these
three equations is strong enough to guarantee the exponential or at least polynomial decay.

Corresponding author. Tel.: +39 030 3715735; fax: +39 030 3715745.
E-mail addresses: [email protected] (J.E. Muoz Rivera), [email protected] (M.G. Naso), [email protected] (R. Quintanilla).

0898-1221/$ see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.camwa.2013.03.022
42 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155

We consider a rod which is composed by a mixture of two interacting continua occupying the finite interval (0, ). The
displacement of each constituent is denoted by u and w , respectively, where u = u(x, t ) : (0, ) (0, T ) R and w =
w(y, t ) : (0, ) (0, T ) R, with T > 0. We assume that the particles under consideration are in the same position at
time t = 0, so that x = y. We suppose also that the temperature = (x, t ) : (0, ) (0, T ) R is the same at both
constituents in each point x and at time t. The mass density of each constituent at time t = 0 is denoted by i , i = 1, 2.
Partial stresses associated with each constituent are T and S respectively, P is the internal body force, is the entropy
density, Q is the heat flux vector and 0 is the absolute temperature in the reference configuration.
When the supply terms are absent the system of equations are proposed by:
the equations of motion
1 utt = Tx P , 2 wtt = Sx + P , (1.1)
the energy equation
(1 + 2 )0 t = Qx , (1.2)
the constitutive equations
T = a11 ux + a12 wx 1 , S = a12 ux + a22 wx 2 , (1.3)
P = (u w), = 1 ux + 2 wx + , Q = K x . (1.4)

If we substitute the constitutive equations (1.3)(1.4) into evolutive equations (1.1)(1.2), we obtain the system of the field
equations

1 utt a11 uxx a12 wxx + (u w) 1 x = 0 in (0, ) (0, T ),

2 wtt a12 uxx a22 wxx (u w) 2 x = 0 in (0, ) (0, T ), (1.5)
t xx 1 uxt 2 wxt = 0 in (0, ) (0, T ),

where := K 01 (1 + 2 )1 .
We note that is related with the thermal conductivity, is the thermal capacity and the parameters aij are related with
the conservative part of the system as well as . The coefficients i , i = 1, 2 are the coupling terms between the mechanical
(conservative) and the thermal (dissipative) parts and they will play a relevant role in our study.
We supplement the system (1.5) with the initial conditions

u(x, 0) = u0 (x), ut (x, 0) = u1 (x) in (0, ),

w(x, 0) = w0 (x), wt (x, 0) = w1 (x) in (0, ), (1.6)
(x, 0) = 0 (x) in (0, ),
for some given functions u0 , u1 , w0 , w1 , 0 : (0, ) R, and with homogeneous Dirichlet boundary conditions

u(0, t ) = 0, u(, t ) = 0 in (0, T ),

w(0, t ) = 0, w(, t ) = 0 in (0, T ), (1.7)
(0, t ) = 0, (, t ) = 0 in (0, T ).
The analysis proposed in this paper can be also applied under Neumann thermal homogeneous boundary conditions, that is
x (0, t ) = 0, x (, t ) = 0 in (0, T ).

a11 a12
Constants 1 , 2 , , , and are positive, and the matrix a12 a22
is positive definite. Thus, we also assume that the co-
efficients of the matrix verify

a11 > 0 and a11 a22 a212 > 0. (1.8)

In this paper we restrict our attention to the case that 1 + 2 = 0.1 Without loss of generality we may suppose that 1 +
2 > 0.2
As we pointed out before our problem (1.5)(1.7) contains three partial differential equations. The two firsts are conser-
vative (there is no decay of energy) and the third one is dissipative. We will show that generically the coupling is so strong
that the thermal dissipation brings the whole system to the exponential/polynomial decay.
To describe the rate of decay of the solutions, it is usual to use expressions like slow decay or exponential decay. When
the solutions of the system are exponentially stable we will say that the system has exponential decay. Otherwise we will
speak about slow decay. The difference has relevant physical consequences. When the system has exponential decay, then

1 The case when + = 0 needs of a further analysis.

1 2
2 This is because u(x, t ), w(x, t ), (x, t ) is a solution of the problem with coupling coefficients and if and only if u (x, t ) = u( x, t ), w(
x, t ) =
1 2
w( x, t ), (
x, t ) = ( x, t ) is a solution of the problem with coupling coefficients 1 and 2 .
 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155 43

the thermomechanical displacements are damped in a way that they can be neglected significantly earlier compared to
the slow decay. Therefore, the nature of the solutions highly determines the temporal behavior of the system. Thus, this
mathematical aspect highly clarifies the physical nature of the system. Concerning the terminology, we talk indifferently of
stability of our problem or of its associated solution.
In a recent paper [18] Alves, et al. proved that, under suitable conditions on the coefficients of the problem, the solution
decays exponentially. We here want to continue this study and to improve the results under the following points. The first
one is that we consider null thermal Dirichlet boundary conditions (see (1.7) 3 ) which is different from the study proposed
in [18]. This is relevant, because the mathematical analysis for these boundary conditions proposes a greater difficulty than
the problem corresponding to the Neumann boundary conditions. Second aspect we want to emphasize is that we give
a complete characterization of the spectrum of the infinitesimal operator of the C0 -semigroup associated to our problem.
Consequently we find the necessary and sufficient conditions to guarantee that the imaginary axis is contained in the re-
solvent, and then we can characterize the decay of the solutions. The last point to highlight here is that, when there are not
eigenvalues in the imaginary axis and the decay is not exponential, we prove the polynomial decay of the solutions.
In classical linear thermoelasticity the asymptotic behavior of solutions, as t +, has been studied by many authors.
We refer, e.g., to the book of Liu and Zheng [25] for a survey on these topics. It is also suitable to point out that the slow decay
which can happens for a one dimensional thermoelastic mixture suggests a difference from the general behavior for the clas-
sical thermoelasticity. In fact, for the classical one dimensional thermoelasticity the decay is always exponential, however
in our case thought the decay is generically exponential, there are a family of cases where the decay is not exponential.
It is also worth recalling several results concerning time decay in thermoelastic mixtures when viscous effects are also
present [19,26,27]. Some regularity properties as analyticity have been proved in these cases, but this is because we have
more dissipation mechanisms. In our paper we only have one dissipative mechanism which is the temperature. The coupling
with the mechanical part is not strong enough to guarantee the analyticity.
Furthermore, for the problem (1.5)(1.7) we cannot expect that the solution always decays. For instance, in case that
1 + 2 = 03 and 2 (a11 + a12 ) = 1 (a12 + a22 ), the system admits isothermal solutions of the form u = w and = 0. They
are undamped and do not decay to zero. Also, when 1 = 2 = 0, the mechanical and thermal parts are uncoupled and the
displacements are described by a conservative equation which cannot decay. They are very particular cases, but we will see
that there are some other cases where there are undamped solutions or where the solutions decay in a way that cannot be
controlled by an exponential function. By applying some recent results (see, e.g., [28]), the polynomial decay will be proved
in these cases.
Our main aim is to show that the C0 -semigroup (and then the solution U (t )) associated to system (1.5)(1.7) is exponen-
tially stable if and only if suitable conditions (see Lemma 3.4, condition (1) or (iii)) on the coefficients are satisfied, whereas
in case that condition (ii) of Lemma 3.4 is verified we will obtain the polynomial decay.
This paper is organized as follows. In Section 2, we establish the well-posedness of system (1.5)(1.7). In Section 3 we
identify the conditions on the constitutive coefficients to guarantee that the imaginary axis is contained in the resolvent (see
Lemma 3.4). Then, under condition (i) or (iii) of Lemma 3.4 the exponential decay of solutions is shown (see Theorem 3.11).
In Section 4 we prove the lack of exponential decay when conditions (i) and (iii) of Lemma 3.4 are not satisfied. In Section 5
we show the polynomial decay in case that the decay is not of exponential type, analyzing separately Dirichlet thermal
boundary conditions and Neumann thermal boundary conditions. Moreover in this later case we find the optimal decay
rate.

2. Existence and uniqueness of solutions

In this section we prove the existence and uniqueness of the solutions for the problem (1.5)(1.7). It is worth noting
that a similar result was obtained in [22]. But it is suitable to state it here again, because we will consider the case that the
functions take values in the complex field. Also it will be very useful to recall the energy and the dissipation of the system
as well as the matrix operator associated to it. As the arguments are very similar we only sketch them.
With the usual notation, we introduce the spaces L2 (0, ), H01 (0, ) and H 1 (0, ) acting on a bounded interval (0, ) R.
Let , and denote the L2 -inner product and L2 -norm, respectively. We recall also the Poincar inequality
0 2 x 2 H01 ,
where 0 = 2 /2 > 0. Thus H01 can be endowed with the inner product , H 1 = x , x . In case of Dirichlet thermal
0
boundary conditions, let us consider the vectorial space
H = H01 (0, ) H01 (0, ) L2 (0, ) L2 (0, ) L2 (0, ),
with the inner product
1
(u, w, v, , ), (
u, w
,
v,
,
)H = a11 ux ,
ux + a12 ux , w
x + wx ,
ux + a22 wx , w
x
2

+ (u w), (
uw
) + 1 v,
v + 2 ,
+ ,
.

3 We recall that this case is not analyzed in this paper.

44 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155

This is a Hilbert space and its corresponding norm is given by

1
(u, w, v, , )2H = a11 ux , ux + a12 (ux , wx + wx , ux ) + a22 wx , wx
2

+ (u w), (u w) + 1 v, v + 2 , + , .

In particular there exists a positive constant c such that inequality

(u, w, v, , )2H c ux 2 + wx 2 + v2 + 2 + 2 ,

is satisfied. For the sake of simplicity, here and in what follows we shall employ the same symbol c for different constants,
even in the same formula.
Analogously, for Neumann thermal boundary conditions, we introduce

HN = H01 (0, ) H01 (0, ) L2 (0, ) L2 (0, ) L2 (0, ),

where

L2 (0, ) = f L2 (0, ) : f (x) dx = 0 .
0

Putting v = ut and = wt , let us introduce U = (u, w, v, , ) and the matrix operator

0 0 I 0 0

0 0 0 I 0
1

a11 a12
()xx ()xx + ()x

I I 0 0
A= 1 1 1 1 1 ,

(2.1)
a12 a22 2
()xx + ()xx I ()x

I 0 0
2 2 2 2 2
1 2
0 0 ()x ()x ()xx

with the domain

D(A) = {U H : AU H } .

The domain D(A) is dense in the Hilbert space H . Our initialboundary value problem (1.5)(1.7) can be rewritten as the
following initial value problem
d
U (t ) = AU (t ), U (0) = U0 ,
dt
where U0 = (u0 , w0 , u1 , w1 , 0 ),

Lemma 2.1. The operator A is the infinitesimal generator of a C0 -semigroup of contractions, denoted by S (t ) = eAt , t 0.

Proof. We will prove that A is a dissipative operator and 0 (A) (resolvent of the operator). Then our conclusion will be
a consequence of the LumerPhillips theorem (see, e.g., [29]).
If U D(A) then

ReAU , U H = x 2 0, (2.2)
2
and therefore the operator A is dissipative.
Given F = (f , g , h, p, q) H , we will show that there exists a unique vector U = (u, w, v, , ) in D(A) such that AU
= F , that is

v=f in H01 (0, ),

=g in H01 (0, ),
a11 uxx + a12 wxx (u w) + 1 x = 1 h in L2 (0, ), (2.3)
a12 uxx + a22 wxx + (u w) + 2 x = 2 p in L2 (0, ),
1 vx + 2 x + xx = q in L2 (0, ).

Using (2.3) 1 and (2.3) 2 in (2.3) 5 we have the equation

xx = q 1 fx 2 gx L2 (0, ). (2.4)
 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155 45

Thus, there exists a unique H 2 satisfying (2.4). It follows from (2.3) 3 and (2.3) 4 that we have to get u and w satisfying:

a11 uxx + a12 wxx (u w) = 1 h 1 x in L2 (0, ), (2.5)

f :=

a12 uxx + a22 wxx + (u w) = 2 p 2 x in L2 (0, ). (2.6)

g :=

As the sesquilinear form B : H01 (0, ) H01 (0, ) C given by

B((u, w), (
u, w
)) = a11 ux ,
ux + a12 (ux , w ux ) + a22 wx , w
x + wx , x + u w, (
uw
)
is continuous and coercive, it follows using the LaxMilgram theorem (see, e.g., [30]) that there exists a unique vector (u, w)
satisfying system (2.5)(2.6). Therefore, we obtain the vector U satisfying the equation AU = F . It is easy to show that
U H c F H , for a positive constant c, and we conclude that 0 (A). 
We finish this section establishing the following result

Proposition 2.2. For any U0 = (u0 , w0 , u1 , w1 , 0 ) H there exists a unique solution U = (u, w, ut , wt , ) of system (1.5)
(1.7) satisfying

u, w C (0, T ; H01 (0, )) C 1 (0, T ; L2 (0, )),

C (0, T ; L2 (0, )) L2 (0, T ; H 1 (0, )).

3. Exponential stability

The main aim of this section is to give the necessary and sufficient conditions to guarantee the imaginary axis is contained
in the resolvent and the sufficient conditions to have the exponential decay of solutions.
It is useful to recall the following known result (see [31]):

Theorem 3.1. Let S (t ) = eAt be a C0 -semigroup of contractions on Hilbert space. Then S (t ) is exponentially stable if and only if
iR (A) and

lim (iI A)1 L(H ) < . (3.1)

||

First we need to know when the imaginary axis is contained in the resolvent (A). Putting F = (f , h, g , p, q) H , we
consider the following system

(i I A)U = F ,
or, in components,

iu v = f , (3.2)
iw = g , (3.3)
i1 v (a11 uxx u) (a12 wxx + w) 1 x = h, (3.4)
i2 (a12 uxx + u) (a22 wxx w) 2 x = p, (3.5)
i 1 vx 2 x xx = q, (3.6)

with R. Since operator A is dissipative, we have

Re(iI A)U , U H = x 2 . (3.7)
2
Let us introduce some notation. We define the matrices

a11 a12 a12 a22
M1 := and M2 := ,
1 2 1 2
and we denote by

D1 := det(M1 ), D2 := det(M2 ),
(3.8)
0 := 2 1 D2 + 1 2 D1 , 1 := 2 1 1 2 .
46 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155

Lemma 3.2. If 0 = 0, then operator A has only one imaginary eigenvalue if and only if

D 1 1 (1 + 2 )1
1+ > 0 and N.
0 0
If 0 = 0, then operator A has
m imaginary eigenvalues if and only if
1 = 0 and D1 < 0,
where m is the maximal natural number such that 2 D1 m2 + 2 (1 + 2 ) > 0;
infinitely many imaginary eigenvalues if and only if
1 = 0 and D1 > 0.

Proof. Let us suppose that there exists at least one imaginary eigenvalue i, with R. This implies that there exists U
= 0 such that (iI A)U = 0. Because of dissipative property (3.7), we obtain = 0. Hence, we get

2 1 u a11 uxx a12 wxx + (u w) = 0, (3.9)

2 w a12 uxx a22 wxx (u w) = 0,
2
(3.10)
1 vx 2 x = 0. (3.11)

Relation (3.11) implies that 1 ux 2 wx = 0. Thus, we have w = 1 u. Using this identity the above system turns into
2

1
(a11 2 a12 1 )uxx = 2 1 u
(1 + 2 )u, (3.12)
2 2
1 1
(a12 2 a22 1 )uxx = 2 2 u + (1 + 2 )u. (3.13)
2 2 2
That is

(a11 2 a12 1 )uxx = 2 2 1 u (1 + 2 )u, (3.14)

(a12 2 a22 1 )uxx = 1 2 u + (1 + 2 )u.
2
(3.15)

This implies that

(1 + 2 ) = 2 2 1 D1 2 , (3.16)
(1 + 2 ) = 1 2 + D2 ,
2 2
(3.17)

where = / , with N. Then, recalling notation (3.8), it follows that there exists only one solution if and only if
2 2 2 2

0 = 0.
In this case we will have
2 1 1 2 1
2 = (1 + 2 ) = (1 + 2 ) ,
2 1 D2 + 1 2 D1 0
(1 + 2 ) D 1 1

D1 + D2
2 = (1 + 2 ) = 1+ .
2 1 D2 + 1 2 D1 2 1 0
As / must be a natural number and 2 must be positive we see

1 D1 1

(1 + 2 ) N and 1 + > 0.
0 0
When 0 = 0 there exists 0 = 0, such that

(D1 , D2 ) = 0 (2 1 , 1 2 ).
So we have from (3.16) and (3.17) that

(1 + 2 ) = 2 2 1 + 0 2 1 2 ,
(1 + 2 ) = 2 1 2 + 0 1 2 2 ,
 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155 47

and there exist solutions if and only if 1 = 0, that is 1 2 = 2 1 . In this case we will have
(1 + 2 ) = 2 1 2 D1 2 .
Therefore
2 1 2 = D1 2 + (1 + 2 ),
and we have infinitely many eigenvalues when D1 > 0. Hence the solution is given by
(1 + 2 ) D1
2 = + 2 .
1 2 1 2
If D1 < 0 then there exist at most m eigenvalues where m is the maximal number satisfying
2 D1 m2 + 2 (1 + 2 ) > 0.
When 0 = 0 and 1 = 0 then there is no imaginary eigenvalue. 

Remark 3.3. Note that the case D1 = D2 = 0 is not possible, because in this case there exist 0 , 1 R such that
(a11 , a12 ) = 0 (1 , 2 ) and (a12 , a22 ) = 1 (1 , 2 ),

a11 a12
which implies that matrix a12 a22
is singular.

Lemma 3.4. Inclusion iR (A) holds if and only if one of the following conditions is verified:

D ( + )
(i) 1 + 1 1 < 0 or 1
0
2 1
N, provided 0 = 0.
0
(ii) 0 = 0 and 1 = 0.
(iii) 0 = 0, 1 = 0 and 2 D1 + 2 (1 + 2 ) < 0.
Proof. The proof is an immediate consequence of Lemma 3.2. 

Remark 3.5. If condition iR (A) holds, the C0 -semigroup S (t ) is strongly stable, namely, limt S (t )U0 H = 0 for all
U0 H (see [32]).
We now introduce the following coefficients

A1 := 1 2 a11 + 2 1 a12 ,
(3.18)
A2 := 1 2 a12 + 2 1 a22 .
Note that 0 = 2 A1 1 A2 .

Lemma 3.6. Under the assumptions of Lemma 3.4, if U = (u, w, v, , ) is a solution of the resolvent system (3.2)(3.6), for
any > 0 there exists c > 0 such that
1 v + 2 2 c U H F H + c F 2H + A1 ux + A2 wx 2 + 1 ux + 2 wx 2 + u w2 .

Proof. Let us define the function such that

xx = 1 v + 2 , (0) = () = 0.
Multiplying Eq. (3.6) by x we get
i , x xx , x 1 vx + 2 x , x = q, x .
It follows that

1 v + 2 2 = x x 0 x , 1 v + 2 ix , q, x , (3.19)

where x x 0 = x () x () x (0) x (0) .

Note that
1/2 1/2
|x (0)|2 + |x ()|2 c H 3/2 c H 1 H 2 ,
1/2
|x (0)|2 + |x ()|2 c 1 v + 2 H 1/2 c 1 v + 2 H 1 1 v + 2 1/2 .
Therefore we get
1/2
x x 0 c x 1/2 xx 1/2 1 v + 2 H 1 1 v + 2 1/2 .
48 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155

Using Eq. (3.6) we find

xx c || + c || 1 ux + 2 wx + c F H . (3.20)
Multiplying Eqs. (3.4) and (3.5) by 2 1 and 1 2 respectively, summing up the product result, and recalling (3.18), we get

i1 2 (1 v + 2 ) = A1 uxx + A2 wxx (2 1 1 2 )(u w) + (2 12 + 1 22 )x + 2 1 h + 1 2 p. (3.21)

The above inequality implies
|| 1 v + 2 H 1 c + c A1 ux + A2 wx + c u wH 1 + c F H , (3.22)
and then we get
xx 1 v + 2 H 1 c A1 ux + A2 wx 2 + c 1 ux + 2 wx 2 + c u w2 + c 2 + c F 2H .
Therefore

x x 0 c U H F H + A1 ux + A2 wx 2 + 1 ux + 2 wx 2 + u w2 + 1 v + 2 2 + c F 2H .
Similarly
ix , || x 1 v + 2 H 1

c x 2 + A1 ux + A2 wx 2 + u w2 + c F 2H
c U H F H + A1 ux + A2 wx 2 + u w2 + c F 2H .
Finally, after the use of the arithmeticgeometric mean inequality we obtain the desired estimate. 

Remark 3.7. In case of 0 = 0, from (3.20) and (3.22) this inequality

xx 1 v + 2 H 1 c 2 + c 1 ux + 2 wx 2 + c u w2H 1 + c U H F H + c F 2H
holds. It implies that
1 v + 2 2 c U H F H + C F 2H + 1 ux + 2 wx 2 + u w2H 1 .

Lemma 3.8. Under assumptions (i), (ii) or (iii) of the Lemma 3.4, we have that for any there exists a positive constant c such
that
1 ux + 2 wx 2 c F 2H + U 2H + c U H F H .

Proof. From (3.4)(3.5) we have

a11 uxx + a12 wxx = J ,
a12 uxx + a22 wxx = K ,
where
J := i1 v + (u w) 1 x h, K := i2 (u w) 2 x p.
So we have that
Ja22 Ka12 Ka11 Ja12
uxx = , wxx = , (3.23)
D D
where D := a11 a22 a212 . Thus, we see
1 Ja22 1 Ka12 + 2 Ka11 2 Ja12 JD2 + KD1
1 uxx + 2 wxx = = .
D D
Multiplying the above expression by 1 u + 2 w we find that
1 ux + 2 wx 2 c 1 v + 2 1 D2 v 2 D1 + c 1 u + 2 w u w
+ c 1 ux + 2 wx + c F H 1 u + 2 w,
and using Poincar inequality and Lemma 3.6 our conclusion follows. 

Lemma 3.9. Let us assume that || is large enough. If 0 = 0, then the following estimate
A1 ux + A2 wx 2 c F 2H + U 2H ,
holds.
 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155 49

Proof. Multiplying Eq. (3.21) by A1 u + A2 w we get

A1 ux + A2 wx 2 = 1 2 1 v + 2 , A1 v + A2 (1 2 2 1 ) u w, A1 u + A2 w

c
|| U 2H

+ 1 2 1 v + 2 , A1 f + A2 g + (12 2 + 22 1 )x , A1 u + A2 w
+ 2 1 h + 1 2 p, A1 u + A2 w.
If is large enough, it follows that

A1 ux + A2 wx 2 c 1 v + 2 2 + c 2 + c U H F H + U 2H .
Using (3.7) and Lemma 3.6 we get

A1 ux + A2 wx 2 c F 2H + c U H F H + U 2H ,
and our conclusion follows. 

Remark 3.10. If 0 = 0, there exists a constant 0 such that

A1 ux + A2 wx = 0 (1 ux + 2 wx ) , (3.24)

where

2 1 a11 + (2 1 + 1 2 ) a12 + 1 2 a22

0 = , (3.25)
1 + 2
and the desired estimate on A1 ux + A2 wx 2 is a consequence of Lemma 3.8.

Theorem 3.11. If condition (i) or (iii) of Lemma 3.4 holds, then the C0 -semigroup S (t ) = eAt is exponentially stable.

Proof. Let us assume that condition (i) of Lemma 3.4 holds. From Lemmas 3.8 and 3.9 we get
A1 ux + A2 wx 2 c F 2H + U 2H ,
1 ux + 2 wx 2 c F 2H + U 2H .
By condition (i) we find
ux 2 + wx 2 c A1 ux + A2 wx 2 + 1 ux + 2 wx 2 c F 2H + c U 2H .

Multiplying (3.4) and (3.5) by v and respectively we conclude that

v2 + 2 c F 2H + c U 2H ,
and then
U 2H c F 2H + c U 2H .
Therefore, taking small enough, the exponential stability holds.
In case of 1 = 0 and 0 = 0, from Eqs. (3.4) and (3.5) we find
i1 2 (v ) D1 (u w)xx + (2 + 1 )(u w) = 2 h 1 p.
Then we have
1 2 v 2 D1 ux wx 2 + (2 + 1 )u w2 = R,
where
R = 2 h 1 p, u w 1 2 v , f g .
Since D1 < 0, there exists a positive constant c such that
v 2 + ux wx 2 c U H F H .
From Remark 3.7 and Lemma 3.8 we conclude that there exists a positive constant c such that
U 2H c F 2H .
So the exponential stability holds. 
50 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155

4. Lack of exponential stability

In this section we will show that when conditions (i) and (iii) fail there exists a lack of exponential stability. However, we
only propose this analysis in the case that thermal boundary conditions are of the Neumann type. That is, we assume that
the heat flux vector vanishes at x = 0, . Moreover, to make the calculations easier we will assume in this section that the
length of the interval is . That is, = .
Let us consider f = g = q = 0. So the resolvent system can be written as

2 1 u a11 uxx a12 wxx + (u w) 1 x = h, (4.1)

2 w a12 uxx a22 wxx (u w) 2 x = p,
2
(4.2)
i xx 1 iux 2 iwx = 0. (4.3)
Multiplying (4.1) by 2 1 and (4.2) by 1 2 summing up the product result and using the fact that 0 = 0, there exists a
constant 0 = 0 such that

2 1 2 (1 u + 2 w) = 0 (1 u + 2 w)xx (1 2 2 1 )(u w) + (12 2 + 22 1 )x + 2 1 h + 1 2 p, (4.4)

where identity (3.24) has been used. Multiplying (4.1) by 2 and (4.2) by 1 and taking the difference we see that
i1 2 (v ) (2 a11 1 a12 )uxx (2 a12 1 a22 )wxx + (2 + 1 )(u w)
= (1 2 2 1 )x + 2 h 1 p. (4.5)
Taking 0 and 1 such that
(2 a11 1 a12 , 2 a12 1 a22 ) = 1 (1 , 2 ) + 0 (1, 1)
Solving the system we get

2 (a11 + a12 ) 1 (a22 + a12 ) 2 2 a11 (2 1 + 1 2 )a12 + 1 1 a22

1 = , 0 = . (4.6)
1 + 2 1 + 2
Therefore we can rewrite the Eq. (4.5) as
i1 2 (v ) 0 (uxx wxx ) + 12 (u w) + 1 x = 1 (1 u + 2 w)xx + 2 h 1 p, (4.7)
with 12 := (1 + 2 ). Denoting by V = 1 u + 2 w, W = u w , Eqs. (4.4), (4.7) and (4.3) can be written as
2 1 2 V 0 Vxx 1 W (12 2 + 22 1 )x = 2 1 h + 1 2 p,
2 1 2 W 0 Wxx + 12 W + 1 x 1 Vxx = 2 h 1 p,
i xx iVx = 0.
Taking 2 1 h + 1 2 p = 0 and 2 h 1 p = sin(x), we can look for solutions of the form
V = A sin(x), W = B sin(x), = C cos(x).
Under the above condition we get that system (4.1)(4.3) is equivalent to
p1 A 1 B 12 C = 0,
1 2 A + p2 B 1 C = 1,
iA + p3 C = 0,
where 12 = 12 2 + 22 1 and

p1 () = 2 1 2 0 2 , p2 () = 1 2 2 0 2 + 12 , p3 = i + 2 .
The matrix of the system is given by

12

p1 1
1 2 p2 1 .
i 0 p3

It follows that
p1 p3 i12 2
B= .
p1 p2 p3 1 1 2 p3 i2 (12 + 12 p2 )
Taking p2 () = c0 , it follows that
1 2 2 = 0 2 + 12 c0 ,
 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155 51

where c0 is such that

c0 p1 1 1 2 = 0.
This implies that B c0 for c0 R and large. Therefore we have



21 U 2H 2 |B sin(x)|2 dx 2 B2 c 4 at least. (4.8)
0 2
Therefore there is no exponential stability.

Remark 4.1. The analysis developed in this section is strongly based on the boundary conditions. With these boundary
conditions it is possible to obtain the precise solution to the different proposed equations. One would like to obtain a similar
result for the case of the Dirichlet thermal boundary conditions. However, we cannot develop the analysis in a similar way
for the alternative conditions, because we cannot obtain the precise solutions.

5. Polynomial decay

In this section we obtain the polynomial stability of solutions in the case (ii) proposed below. Our argument is strongly
based on the following theorem (see [28, Theorem 2.4]).

Theorem 5.1. Let (S (t ))t 0 be a bounded C0 -semigroup on a Hilbert space H with generator A such that iR (A). Then, for a
fixed > 0,
(iI A)1 L(H ) = O (|| ) , S (t )A1 L(H ) = O t 1/ , t .

The main result of this section is

Theorem 5.2. Let us suppose that condition (ii) holds. If satisfies Neumann boundary condition at least in one end, then the
corresponding solution U (t ) = eAt U0 decays polynomially as
c
U (t )HN U0 D(A) .
t 1/2
Otherwise, for satisfying Dirichlet boundary condition, we have the solution decays as
c
U (t )H U0 D(A) .
t 1/4
Moreover the rate of decay to Neumann boundary condition is optimal.
Proof. We note that in that follows we will use the same symbol R for different functions such that |R| c U H F H .
Using (3.21) and the fact that 0 = 0, there exists a constant 0 = 0 such that

i1 2 (1 v + 2 ) = 0 (1 u + 2 w)xx (1 2 2 1 )(u w) + (12 2 + 22 1 )x + 2 1 h + 1 2 p, (5.1)

where we have used
(A1 , A2 ) = 0 (1 , 2 ),
and
A1 A2 2 1 a11 + 1 2 a12 2 1 a12 + 1 2 a22
0 = = = = .
1 2 1 2
Multiplying Eq. (5.1) by 1 u + 2 w we get
c c c
1 ux + 2 wx 2 c 1 v + 2 2 + u w2 + U H F H + F 2H .
||2 ||2 ||2
From the above inequality and using Remark 3.7 we see

1 v + 2 2 c U H F H + C F 2H + u w2H 1 . (5.2)
On the other hand from (3.23) and recalling the definition of J and K , we have that
i [1 (a22 + a12 )v 2 (a11 + a12 )] D(u w)xx + 3 (u w) + 3 x = F2 , (5.3)
where
3 = (a11 + 2a12 + a22 ), 3 = (D1 + D2 ), F2 = (a22 + a12 )h (a11 + a12 )p.
52 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155

Taking 0 and 1 such that

(1 (a22 + a12 ), 2 (a11 + a12 )) = 1 (1 , 2 ) + 0 (1, 1),

and, solving the system, we get

1 (a22 + a12 ) 2 (a12 + a11 ) 2 1 a11 + (2 1 + 1 2 )a12 + 1 2 a22

1 = = 0 , 0 = ,
1 + 2 1 + 2
recalling (4.6), since 1 = 1 , we can rewrite the Eq. (5.3) as

i0 (v ) D(u w)xx + 3 (u w) + 3 x = i1 (1 v + 2 ) + F2 , (5.4)

when 1 = 0. Multiplying Eqs. (4.7) and (5.4) by 0 and 1 2 respectively and summing up the corresponding result, it
follows

i1 2 (0 0 )(v ) (1 2 D 0 0 )(u w)xx + [3 1 2 (1 + 2 )0 ] (u w)

[(D1 + D2 )1 2 1 0 ] x = F3 + 1 [i1 2 (1 v + 2 ) 0 (1 u + 2 w)xx ],
with F3 = 1 2 F2 0 F1 . Multiplying (3.4) by 1 2 and (3.5) by 2 1 , we get

i1 2 (0 0 )(v ) (1 2 D 0 0 )(u w)xx + [3 1 2 (1 + 2 )0 1 1 ] (u w)

+ (D1 + D2 )1 2 1 0 1 12 2 1 1 22 x = F4 ,

where F4 = F3 + 1 (1 2 h + 2 1 p). That is, there exist constants bi , i = 1, 2, 3, 4, such that

ib1 (v ) b2 (u w)xx + b3 (u w) + b4 x = F4 .

In the case 1 = 0, Eq. (5.4) is already of this type. Multiplying the above equation by (v ) and taking its real part we get

Rex , (v ) = ReF5 , (v ), (5.5)

F4 + 3 (f g ) + 2 (f g )xx . Multiplying Eq. (4.4) by u w we get
1 ib ib
where F5 = b4

1 u w2 = 1 2 (1 v + 2 ), (v ) 0 (1 ux + 2 wx ), (ux wx )

=I0

+ (12 2 + 22 1 )x , (u w) + R, (5.6)

where R is such that

|R| c u w F H .
Multiplying Eq. (4.7) by (1 u + 2 w)/0
1 2 12
(1 ux + 2 wx ), (ux wx ) = (1 v + 2 ), (v ) (1 u + 2 w), (u w)
0 0
1 1
(1 u + 2 w), x (1 v + 2 )x 2 + R.
0 0 2
Substitution of I0 in (3.23) given by the above identity we get

0

1
u w2 = 1 2 1 Re(1 v + 2 ), (v ) + c2 Re(1 u + 2 w), (u w)
1 0
+ c3 Rex , (u w) + c4 Re(1 u + 2 w), x + c5 (1 v + 2 )x 2 + Re(R). (5.7)

Let us suppose that satisfies Neumann boundary condition, that is we assume that x (0) = 0. Integrating Eq. (3.6) over
[0, x] and multiplying the resulting equation by (v ), we get
x

(1 v + 2 ), (v ) = ix , (v ) x , (v ) q ds , (v ) ,
0

where

xx = , x (0) = 0, () = 0.
 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155 53

Using (5.5) and (4.7) we get

12
Re(1 v + 2 ), (v ) = Rex , [0 (u w) + 1 (1 u + 2 w)] + Rex , (u w) + Re(R)
1 2 1 2
c |x |, [|u w| + |1 v + 2 |] + c | |, |u w| + Re(R), (5.8)

where

|R| c U HN F HN .
Inserting this inequality into (5.7) we get

u w2 c 1 v + 2 2 + c x 2 + R.
From (5.2) it follows that

v 2 c ||2 U HN F HN + c ||2 F 2HN .

Using the above equation and that i(u w) = v + (f g ), from (3.19) we conclude that

1 v + 2 2 c ||2 U HN F HN + c ||2 F 2HN ,

from large enough. Denoting by

v = F1 , 1 v + 2 = G1 ,
we have that
2 F1 + G1 1 F1 + G1
v= , = .
1 + 2 1 + 2
So we obtain

v2 + 2 c ||2 U HN F HN + c ||2 F 2HN . (5.9)

Therefore multiplying Eq. (3.4) by u and (3.5) by w and summing up the product results we get

a11 ux , ux + a12 wx , ux + a12 ux , wx + a22 wx , wx

= 1 v2 + 2 2 u w2 + x , (1 u + 2 w) + R,
where R is such that

|R| c U HN F HN .
Since the matrix (aij ) is positive definite, there exists positive constant such that

ux 2 + wx 2 c v2 + 2 + c U HN F HN .

From (5.9) we get

ux 2 + wx 2 c ||2 F 2HN + c ||2 U HN F HN

for || large enough. Then we have that

U 2HN c ||2 F 2HN + c ||2 U HN F HN .

Therefore we get

U 2HN c ||4 F 2HN ,

which implies that the solution decay polynomially as t 1/2 . Using inequality (4.8) we conclude that the rate of decay is
optimal for the Neumann conditions in both ends. Because if there exists a better decay such as t 1/(2) , then we will
have that ||2 U H must be bounded. But this is contradictory to inequality (4.8). Finally, if satisfies Dirichlet boundary
condition, then multiplying by 2 Eq. (5.6) we arrive to

v 2 c ||4 1 v + 2 2 + 1 ux + 2 wx 2 + x 2 + ux wx 2 + ||2 R.

Using the same above procedure we conclude that

U 2H c ||8 F 2H ,
and then our conclusion follows. 
54 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155

Remark 5.3. As we pointed out at the end of the previous section, in the case of the Dirichlet thermal boundary conditions
we do not know how to find a similar result to the one obtained for the Neumann thermal boundary conditions. Thus, we
cannot conclude that the decay estimate, obtained for the Dirichlet thermal boundary conditions, is optimal. However, we
want to recall that results where the decay is different, depending on the boundary conditions, have been proved in other
thermomechanical situations as the Bresse system (see [33]).

Remark 5.4. It is worth noting that it is possible to obtain rate of decay for the L2 -norm of the spatial derivative for the
temperature. We do it for the situation where the decay is slowest, but it can be adapted to the other faster situation. As
c
U H U0 D(A) ,
t 1/4
we see that
c
2 U0 2D(A) .
t 1/2
Using the semigroup property and the linearity we get that
c
AU H U0 D(A2 ) .
t 1/4
In view of [34, Proposition 3.1] we conclude that
c
AU H U0 D(A) .
t 1/8
Therefore we obtain that
c
xx 2 1/4 U0 2D(A) .
t
After the use of the interpolation inequality we see
c
x 2 C xx U0 2D(A) .
t 3/8
This inequality proposes decay rates for the several spatial derivatives of the temperature.

Acknowledgments

We express our gratitude to the referees for their helpful suggestions.

The work of J.E. Muoz Rivera is supported by CNPq (Brazil). The work of R. Quintanilla is supported by the project
Ecuaciones en Derivadas Parciales en Termomecnica. Teora y Aplicaciones (MTM2009-08150) of the Spanish Ministry
of Education.

References

[1] P. Kelly, A reacting continuum, Internat. J. Engrg. Sci. 2 (2) (1964) 129153.
[2] C. Truesdell, R. Toupin, The classical field theories, in: Handbuch der Physik, vol. Bd. III/1, Springer, Berlin, 1960, With an appendix on tensor fields by
J.L. Ericksen, pp. 226793; appendix, pp. 794858.
[3] A.C. Eringen, J.D. Ingram, A continuum theory of chemically reacting media I, Internat. J. Engrg. Sci. 3 (1965) 197212.
[4] J.D. Ingram, A.C. Eringen, A continuum theory of chemically reacting media II, Internat. J. Engrg. Sci. 5 (1967) 289322.
[5] I. Mller, A thermodynamic theory of mixtures of fluids, Arch. Ration. Mech. Anal. 29 (1968) 344369.
[6] A.E. Green, P.M. Naghdi, A dynamical theory of interacting continua, Internat. J. Engrg. Sci. 3 (1965) 231241.
[7] A.E. Green, P.M. Naghdi, A note on mixtures, Internat. J. Engrg. Sci. 6 (11) (1968) 631635.
[8] R.M. Bowen, Theory of mixtures, in: A.C. Eringen (Ed.), Continuum Physics III, Academic Press, New York, 1976, pp. 689722.
[9] R.M. Bowen, J.C. Wiese, Diffusion in mixtures of elastic materials, Internat. J. Engrg. Sci. 7 (7) (1969) 689722.
[10] R.J. Atkin, R.E. Craine, Continuum theories of mixtures: basic theory and historical development, Quart. J. Mech. Appl. Math. 29 (2) (1976) 209244.
[11] A. Bedford, D.S. Drumheller, Theories of immiscible and structured mixtures, Internat. J. Engrg. Sci. 21 (8) (1983) 863960.
[12] I. Samohyl, Thermodynamics of Irreversible Processes in Fluid Mixtures, Teubner Verlag, Leipzig, NJ, 1987.
[13] K.R. Rajagopal, L. Tao, Mechanics of Mixtures, in: Series on Advances in Mathematics for Applied Sciences, vol. 35, World Scientific Publishing Co. Inc.,
River Edge, NJ, 1995.
[14] T.R. Steel, Applications of a theory of interacting continua, Quart. J. Mech. Appl. Math. 20 (1967) 5772.
[15] A. Bedford, M. Stern, A multi-continuum theory of composite elastic materials, Acta Mech. 14 (1972) 85102.
[16] A. Bedford, M. Stern, Towards a diffusing continuum theory of composite elastic materials, J. Appl. Mech. 38 (1972) 814.
[17] H.F. Tiersten, M. Jahanmir, A theory of composites modeled as inerpenetreting solid continua, Arch. Ration. Mech. Anal. 65 (1977) 153192.
[18] M.S. Alves, J.E. Muoz Rivera, R. Quintanilla, Exponential decay in a thermoelastic mixture of solids, Internat. J. Solids Struct. 46 (78) (2009)
16591666.
[19] M.S. Alves, J.E. Muoz Rivera, M. Seplveda, O.V. Villagrn, Exponential stability in thermoviscoelastic mixtures of solids, Internat. J. Solids Struct. 46
(24) (2009) 41514162.
[20] M.S. Alves, B.M.R. Calsavara, J.E. Muoz Rivera, M. Seplveda, O.V. Villagrn, Analyticity and smoothing effect for the coupled system of equations of
Kortewegde Vries type with a single point singularity, Acta Appl. Math. 113 (1) (2011) 75100.
[21] D. Iean, R. Quintanilla, Existence and continuous dependence results in the theory of interacting continua, J. Elasticity 36 (1) (1994) 8598.
 J.E. Muoz Rivera et al. / Computers and Mathematics with Applications 66 (2013) 4155 55

[22] F. Martnez, R. Quintanilla, Some qualitative results for the linear theory of binary mixtures of thermoelastic solids, Collect. Math. 46 (3) (1995)
263277.
[23] R. Quintanilla, Existence and exponential decay in the linear theory of viscoelastic mixtures, Eur. J. Mech. A Solids 24 (2) (2005) 311324.
[24] R. Quintanilla, Exponential decay in mixtures with localized dissipative term, Appl. Math. Lett. 18 (12) (2005) 13811388.
[25] Z. Liu, S. Zheng, Semigroups Associated with Dissipative Systems, in: Chapman & Hall/CRC Research Notes in Mathematics, vol. 398, Chapman &
Hall/CRC, Boca Raton, FL, 1999.
[26] M.S. Alves, J.E. Muoz Rivera, M. Seplveda, O.V. Villagrn, Analyticity of semigroups associates with thermoeviscoelastic mixtures of solid, J. Therm.
Stresses 32 (2009) 9861004.
[27] M.S. Alves, J.E. Muoz Rivera, M. Seplveda, O.V. Villagrn, Stabilization in three-dimensional porous thermoviscoelastic mixtures, Q. J. Math. 64 (1)
(2013) 3757.
[28] A. Borichev, Y. Tomilov, Optimal polynomial decay of functions and operator semigroups, Math. Ann. 347 (2) (2010) 455478.
[29] A. Pazy, Semigroups of Linear Operators and Applications to Partial Differential Equations, in: Applied Mathematical Sciences, vol. 44, Springer-Verlag,
New York, 1983.
[30] H. Brezis, Analyse Fonctionnelle, Thorie et Applications. [Theory and Applications], in: Collection Mathmatiques Appliques pour la Matrise
(Collection of Applied Mathematics for the Masters Degree), Masson, Paris, 1983.
[31] J. Prss, On the spectrum of C0 -semigroups, Trans. Amer. Math. Soc. 284 (2) (1984) 847857. MR 743749 (85f:47044).
[32] F.L. Huang, Strong asymptotic stability of linear dynamical systems in Banach spaces, J. Differential Equations 104 (2) (1993) 307324.
[33] Z. Liu, B. Rao, Energy decay rate of the thermoelastic Bresse system, Z. Angew. Math. Phys. 60 (1) (2009) 5469.
[34] J. Prss, A. Batkai, K. Engel, R. Schnaubelt, Polynomial stability of operator semigroups, Math. Nachr. 279 (1) (2006) 14251440.