Stability Results of Coupled Wave Models With Locally Memory in A Past History Framework Via Nonsmooth Coefficients On The Interface

10991476, 2021, 8, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/mma.7235 by UFPR - Universidade Federal do Parana, Wiley Online Library on [09/02/2023].
See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Received: 28 August 2020 Revised: 11 January 2021 Accepted: 11 January 2021
DOI: 10.1002/mma.7235
RESEARCH ARTICLE
Stability results of coupled wave models with locally

memory in a past history framework via nonsmooth
coefficients on the interface
Mohammad Akil1 Haidar Badawi2 Serge Nicaise2 Ali Wehbe3
1
Laboratoire LAMA, University Savoie
Mont Blanc, Chambéry, France In this paper, we investigate the stabilization of a locally coupled wave equations
2
LAMAV, University Polytechnic with local viscoelastic damping of past history type acting only in one equation
Hauts-de-France, Valenciennes, France via non smooth coefficients. First, using a general criteria of Arendt–Batty,
3
Faculty of Sciences 1 Khawarizmi
we prove the strong stability of our system. Second, using a frequency domain
Laboratory of Mathematics and
Applications-KALMA, Lebanese approach combined with the multiplier method, we establish the exponential
University, Beirut, Lebanon stability of the solution if and only if the two waves have the same speed of
Correspondence
propagation. In case of different speed propagation, we prove that the energy
Mohammad Akil, Laboratoire LAMA, of our system decays polynomially with rate t−1 . Finally, we show the lack of
University Savoie Mont Blanc, Chambéry, exponential stability if the speeds of wave propagation are different.
France.
Email: mohammad.akil@univ-smb.fr
K E Y WO R D S
Communicated by: A. Miranville coupled wave equation, exponential stability, frequency domain approach, past history damping,
polynomial stability, strong stability
Funding information
There are no funders to report for this
M S C C L A S S I F I C AT I O N
submission.
35L05; 74D05; 93D20; 93D23; 93D15
1 I N T RO DU CT ION
In this paper, we investigate the indirect stability of coupled elastic wave equations with localized past history damping.
More precisely, we consider the following system:
⎧ u − (au − b(x) ∫ ∞ g(s)u (x, t − s)ds) + c(x)𝑦 = 0, (x, t) ∈ (0, L) × (0, ∞),
⎪ tt x 0 x x t
⎪ 𝑦 tt − 𝑦 xx − c(x)u t = 0, (x, t) ∈ (0, L) × (0, ∞),
⎨ u(0, t) = u(L, t) = 𝑦(0, t) = 𝑦(L, t) = 0, t > 0, (1.1)
⎪ (u(x, −s), ut (x, 0)) = (u0 (x, s), u1 (x)), (x, s) ∈ (0, L) × (0, ∞),
⎪ (𝑦(x, 0), 𝑦t (x, 0)) = (𝑦0 (x), 𝑦1 (x)), x ∈ (0, L),
⎩
where L and a are positive real numbers. We suppose that there exist a nonzero constant c0 and positive constants 𝛼, 𝛽, 𝛾,
and b0 such that 0 < 𝛼 < 𝛽 < 𝛾 < L, and define
{
b0 , x ∈ (0, 𝛽),
b(x) = 0, x ∈ (𝛽, L), b(·)
{
c0 , x ∈ (𝛼, 𝛾),
c(x) = 0, x ∈ (0, 𝛼) ∪ (𝛾, L). c(·)
6950 © 2021 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/mma Math Meth Appl Sci. 2021;44:6950–6981.
10991476, 2021, 8, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/mma.7235 by UFPR - Universidade Federal do Parana, Wiley Online Library on [09/02/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
AKIL ET AL. 6951
FIGURE 1 Geometric description of the functions b(x) and c(x)

[Colour figure can be viewed at wileyonlinelibrary.com]
The general integral term represents a history term with the relaxation function g that is supposed to satisfy the following
hypotheses:
{
g ∈ L1 ([0, ∞)) ∩ C1 ([0, ∞)) is a positive function such that
∞ ̃
g(0) ∶= g0 > 0, ∫0 g(s)ds ∶= g̃ , b(x) ∶= a − b(x)̃g > 0, and (H)
g′ (s) ≤ −mg(s), for somem > 0, ∀s ≥ 0.
Remark that, the last assumption in (H) implies that
g(s) ≤ g0 e−ms , ∀s ≥ 0. (1.2)
Moreover, from the definition of b(·) (see Figure 1), we have
{
̃
b(x) ∶= a−b(x)̃g = b̃0 ∶= a − b0 g̃ , x ∈ (0, 𝛽), ̃
(b(·))
a, x ∈ (𝛽, L).
The notion of indirect damping mechanisms has been introduced by Russell,1 and since this time, it retains the attention of
many authors. In particular, the fact that only one equation of the coupled system is damped refers to the so-called class of
“indirect” stabilization problems initiated and studied in previous studies2-4 and further studied by many authors; see for
instance5-7 and the rich references therein. In 1996, Liu and Zheng8 studied the one-dimensional linear thermoviscoelastic
system { ∞
utt − 𝛼uxx + ∫0 g(s)uxx (·, t − s)ds + 𝛾c𝜃x = 0, in (0, 𝜋) × (0, ∞),
(1.3)
𝜃t + 𝛾uxt − 𝜃xx = 0, in (0, 𝜋) × (0, ∞),
where 𝛼 > 0, 𝛾 ≥ 0, and c > 0 and proved that the system is exponential stable. In 2008, Rivera et al.9 studied the stability of
one-dimensional Timoshenko system with past history acting only in one equation; they showed that the system is expo-
nential stable if and only if the equations have the same wave speeds of propagation. In case that the wave speeds of the
equations are different, they proved that the solution of the system decays polynomially to zero. In 2011, Guesmia10 stud-
ied the asymptotic stability of the following abstract linear dissipative integrodifferential equation with infinite memory
∞
utt (t) + Au(t) − g(s)Bu(t − s)ds = 0, ∀ t > 0. (1.4)
∫0
where A : D(A) → H and B : D(B) → H are self-adjoint linear positive definite operators with domains D(A) ⊂ D(B) ⊂ H such
that the embeddings are dense and compact, H is a Hilbert space, and g ∶ R+ → R+ is the convolution kernel function. He
showed that the stability of the system holds for a relatively large class of convolution kernels, and he provided a relation
between the decay rate of the solution and the growth of the kernel at infinity. In 2012, Matos et al.11 studied the stability
of the abstract coupled wave equations with past history, by considering
⎧ u + A u − ∫ ∞ g(s)A u(t − s)ds + 𝛽v = 0,

⎪ tt 1 0 2
⎪ vtt + B v + 𝛽u = 0, in L 2
(R+ , ),
⎨ u(−t) = u0 (t), t ≥ 0, (1.5)
⎪ v(0) = v0 ,
⎪ ut (0) = u1 , vt (0) = v1 ,
⎩
where A1 , A2 , and B are self-adjoint positive-definite operators with the domain D(A1 ) ⊆ D(A2 ) ⊂  and D(B) ⊂ 
with compact embeddings in , g : [0, ∞) → [0, ∞) is a smooth and summable function and 𝛽 is a small positive constant.
They showed that the abstract setting is not strong enough to produce exponential stability, and they proved that the
6952 AKIL ET AL.
solution decays polynomially to zero. In 2014, Fatori et al.12 studied a fully hyperbolic thermoelastic Timoshenko system
with past history where the thermal effects are given by Cattaneo's law; they established
( the )exponential
( ) stability of the
𝜌1 b𝜌1 𝜏𝜌1 𝛿 2
solution if and only if the coefficients of their system satisfy the next relation 𝜒0 ∶= 𝜏 − 𝜌3 𝜅
𝜌2 − 𝜅
− 𝜌3 𝜅
= 0. In
the case 𝜒 0 ≠ 0, they established optimal polynomial stability rates. In the same year, Santos et al.13
studied the stability
of one-dimensional Bresse system with past history acting in the shear angle displacement; they showed the exponential
decay of the solution if and only if the wave speeds are the same. Otherwise, they showed that the Bresse system is
polynomial stable with optimal decay rate. In 2014, Jin et al.14 studied the stability of the abstract Cauchy problem for a
system of coupled equations with fading memory
⎧ utt (t) + Au(t) − ∫ t g(t − s)Au(s)ds + 𝛼u(t) + 𝛽Bv(t) = 𝑓 (u(t)), t > 0,

0
⎪ vtt (t) + Av(t) + 𝛽Bu(t) = 0, t > 0,
⎨ u(0) = u , u (0) = u , (1.6)
⎪ 0 t 1
⎩ v(0) = v0 , vt (0) = v1 ,
where 𝛼( ≥ 0,)𝛽 ≥ 0, A is a positive self-adjoint linear operator in a Hilbert space H, B is a symmetric linear operator in H,
√
𝑓 ∶ D A → H denotes external forces, and g is the memory kernel. If 𝛽 > 0, they established a polynomial decay
rate of order t−1 of the full energy, while if 𝛽 = 0, they proved the same decay rate but only on the energy of u. In 2015,
Guesmia15 studied the asymptotic behavior for coupled abstract evolution equations with one infinite memory
{ ∞
̃
utt (t) + Au(t) − ∫0 g(s)Bu(t − s)ds + Bv(t) = 0, ∀ t > 0,
̃ (1.7)
vtt (t) + Ãv(t) + Bu(t) = 0, ∀ t > 0,
where A : D(A) → H, Ã ∶ D(Ã) → H, and B : D(B) → H are self-adjoint linear positive definite operators with domains
D(A) ⊂ D(B) ⊂ H and D(Ã) ⊂ H such that the embeddings are dense and compact, B̃ ∶ H → H is a self-adjoint bounded
operator, H is a real Hilbert space, and g ∶ R+ → R+ is the convolution kernel. He proved under a boundedness condition
on the past history data that the stability of the system holds for convolution kernels having much weaker decay rates
than the exponential one. In 2017, Alabau-Boussouira et al.16 studied the energy decay of the coupled wave equations
{
utt − Δu + 𝜌(x, ut ) + 𝛼(x)vt = 0, in Ω × (0, ∞),
vtt − Δv − 𝛼(x)ut = 0, in Ω × (0, ∞), (1.8)
u = v = 0, on Γ × (0, ∞),
where Ω is a bounded subset of Rn , Γ is a smooth boundary of Ω, 𝜌(x, ut ) is a nonlinear damping, and 𝛼 ∈ C(Ω) is positive
on a subset of positive measure (but may vanish in some parts of Ω). They proved that the total energy of the whole
system (1.8) decays as fast as the damped single equation. Also, they gave a one-step general explicit decay formula for
arbitrary nonlinearity. In 2018, Abdallah et al.17 studied the stability of a one-dimensional Bresse system with infinite
memory-type control and/or with heat conduction given by Cattaneo's law acting in the shear angle displacement. In
the absence of thermal effect, under the same speed propagation, they established the exponential stability of the system.
However, in the case of different speed propagation, they established a polynomial energy decay rate. In 2018, Cavalcanti
et al.18 studied the asymptotic stability of the multidimensional damped wave equation, by considering:
∞
𝜌(x)utt − Δu + g(s)div[a(x)(u(·, t − s)]ds + b(x)ut = 0, in Ω × (0, ∞), (1.9)
∫0
where Ω is an open bounded and connected set of Rn , n ≥ 2, 𝜌(x) is constant, a(x) ≥ 0 is a smooth function, b(x) ≥ 0 is a
bounded function acting effectively in a region A of Ω where a = 0. Considering that the well-known geometric control
condition (𝜔, T0 ) holds and supposing that the relaxation function g is bounded by a function that decays exponentially
to zero, they proved that the solution to the corresponding partial viscoelastic model decays exponentially to zero, even
in the absence of the frictional dissipative effect. Moreover, they proved by removing the frictional damping term b(x)ut
and by assuming that 𝜌 is not constant that localized viscoelastic damping is strong enough to assure that the system
is exponentially stable. In 2019, Hao and Wang19 studied the stability of the abstract thermoelastic system with infinite
AKIL ET AL. 6953
memory
∞
⎧ utt + Au + But − ∫0 g(s)Au(t − s)ds − A𝛼 𝜃 = 0, t > 0,
⎪ 𝜃t + kA𝛽 𝜃 + A𝛼 ut = 0, t > 0,
⎨ u(−t) = u (t), t ≥ 0,
(1.10)
⎪ 0
⎩ ut (0) = u1 , 𝜃(0) = 𝜃0 ,
where 𝛼 ∈ [0, 1), 𝛽 ∈ (0, 1], A : D(A) → H and B : D(B) → H are self-adjoint linear positive definite operators with domains
D(A) ⊂ D(B) ⊂ H such that the embeddings are dense and compact, and H is a real Hilbert space. They obtained the stability
result and provided a direct relationship between the decay rate of the energy and the decay rate of kernel function g. In
2019, Hassan and Messaoudi20 studied the stability of an abstract class of weakly dissipative second-order system with
finite memory
{ t
utt + Au − ∫−∞ g(t − s)A𝛼 u(s)ds = 0, t > 0, (1.11)
u(−t) = u0 (t), t ≥ 0, ut (0) = u1 .
where A : D(A) ⊂ H → H is a positive definite self-adjoint operator on H, H is a real separable Hilbert space, g is the convo-
lution kernel, and 𝛼 ∈ [0, 1]. They established a new general decay rate for the solution of the system under approbritae
conditions on the memory kernel g. In 2019, Jin et al.21 studied the stability of an abstract Cauchy problem for a system
of coupled equations with one infinite memory, by considering
⎧ utt (t) + A1 u(t) − ∫0∞ g(s)Au(t − s)ds + Bv(t) = 0, t > 0,

⎪ vtt (t) + A2 v(t) + Bu(t) = 0, t > 0,
⎨ u(−t) = u (t), ∀t ≥ 0, u (0) = u , (1.12)
⎪ 0 t 1
⎩ v(0) = v0 , vt (0) = v1 ,
where A, A1 , and A2 are positive self-adjoint linear operators in a Hilbert space H, B is a positive self-adjoint bounded linear
operator in H, and g is the memory kernel. They established a polynomial energy decay rate of order t−1 . In 2011, Almeida
et al.22 studied the stability of coupled wave equations with past history effective only in one equation, by considering the
following system:
⎧ u − Δu + ∫ ∞ g(s)Δu(·, t − s)ds + 𝛼v = 0, in Ω × (0, ∞),
⎪ tt 0
⎪ vtt − Δv + 𝛼u = 0, in Ω × (0, ∞),
⎨ u = v = 0, on Γ × (0, ∞) (1.13)
⎪ u(x, 0), v(x, 0)) = (u0 (x), v0 (x)) in Ω,
⎪ ut (x, 0), vt (x, 0)) = (u1 (x), v1 (x)) in Ω,
⎩
where Ω is an open bounded set of Rn with smooth boundary Γ and 𝛼 > 0. They showed that the dissipation given by
the memory effect is not strong enough to produce exponential decay. They proved that the solution of the system (1.13)
1
decays polynomially with rate t− 2 . Also, in 2020, Cordeiro et al.23 established the optimality of the decay rate.
But to the best of our knowledge, it seems that no result in the literature exists concerning the case of coupled wave
equations with localized past history damping, especially in the absence of smoothness of the damping and coupling
coefficients. The goal of the present paper is to fill this gap by studying the stability of system (1.1).
This paper is organized as follows: In Section 2, we prove the well-posedness of our system by using semigroup approach.
In Section 3, following a general criteria of Arendt Batty, we show the strong stability of our system in the absence of
the compactness of the resolvent. Next, in Section 4, by using the frequency domain approach combining with a specific
multiplier method, we establish exponential stability of the solution if and only if the waves have same speed propagation
(i.e., a = 1). In the case a ≠ 1, we prove that the energy of our system decays polynomially with the rate t−1 . Finally, in
Section 5, we show the lack of exponential stability in case that the speeds of wave propagation are different, i.e., when
a ≠ 1.
2 W E L L- P OSE DN E SS O F T HE SYSTEM
In this section, we will establish the well-posedness of system (1.1) by using semigroup approach. To this aim, as in
Dafermos,24 we introduce the following auxiliary change of variable
𝜔(x, s, t) ∶= u(x, t) − u(x, t − s), (x, s, t) ∈ (0, 𝛽) × (0, ∞) × (0, ∞). (2.1)
6954 AKIL ET AL.
Then, system (1.1) becomes

( )
̃ (u, 𝜔)
utt − Sb(·) + c(·)𝑦t = 0, (x, t) ∈ (0, L) × (0, ∞), (2.2)
x
𝑦tt − 𝑦xx − c(·)ut = 0, (x, t) ∈ (0, L) × (0, ∞), (2.3)
𝜔t (·, s, t) + 𝜔s (·, s, t) − ut = 0, (x, s, t) ∈ (0, 𝛽) × (0, ∞) × (0, ∞), (2.4)

where
{
Sb̃ (u, 𝜔) ∶= b̃0 ux + b0 ∫0 g(s)𝜔x (x, s)ds, x ∈ (0, 𝛽),
∞
̃ (u, 𝜔) ∶=
Sb(·) 0 ̃ (u, 𝜔))
(Sb(·)
aux , x ∈ (𝛽, L).
With the following boundary conditions

{
u(0, t) = u(L, t) = 𝑦(0, t) = 𝑦(L, t) = 0, t > 0,
𝜔(·, 0, t) = 0, (x, t) ∈ (0, 𝛽) × (0, ∞), (2.5)
𝜔(0, s, t) = 0, (s, t) ∈ (0, ∞) × (0, ∞),
and the following initial conditions

{
u(·, −s) = u0 (·, s), ut (·, 0) = u1 (·), (x, s) ∈ (0, L) × (0, ∞),
𝑦(·, 0) = 𝑦0 (·), 𝑦t (·, 0) = 𝑦1 (·), x ∈ (0, L), (2.6)
𝜔(·, s, 0) = u0 (·, 0) − u0 (·, s), (x, s) ∈ (0, 𝛽) × (0, ∞).
The energy of system (2.2)–(2.6) is given by
E(t) = E1 (t) + E2 (t) + E3 (t), (2.7)
where
1
L(
̃
) 1
L( ) b0 𝛽 ∞
E1 (t) = |ut |2 + b(·)|ux |2
dx, E 2 (t) = |𝑦 t |2
+ |𝑦 x |2
dx and E 3 (t) = g(s)|𝜔x (·, s, t)|2 dsdx.
2 ∫0 2 ∫0 2 ∫0 ∫0
Lemma 2.1. Under the hypotheses (H). Let U = (u, ut , 𝑦, 𝑦t , 𝜔) be a regular solution of system (2.2)–(2.6). Then, the
energy E(t) satisfies the following estimation
𝛽 ∞
d b
E(t) = 0 g′ (s)|𝜔x (·, s, t)|2 dsdx. (2.8)
dt 2 ∫0 ∫0
Proof. First, multiplying (2.2) by ut , integrating over (0, L), using integration by parts with (2.5), using the definition
of (Sb(·) ̃
̃ (u, 𝜔)), (b(·)), and (c(·)), then taking the real part, we obtain
{ } { }
d 𝛽 ∞ 𝛾
E (t)
dt 1
= − ℜ b0 ∫0 ∫0 g(s)𝜔x (·, s, t)utx dsdx − ℜ c0 ∫𝛼 𝑦t ut dx . (2.9)
Now, multiplying (2.3) by 𝑦t , integrating over (0, L), using the definition of (c(·)), then taking the real part, we get
{ 𝛾 }
d
E2 (t) = ℜ c0 ut 𝑦t dx . (2.10)
dt ∫𝛼
Deriving (2.4) with respect to x, we obtain
𝜔xt (·, s, t) + 𝜔xs (·, s, t) − utx = 0 in (0, 𝛽) × (0, ∞) × (0, ∞). (2.11)
AKIL ET AL. 6955
Multiplying (2.11) by b0 g(s)𝜔x (·, s, t), integrating over (0, 𝛽) × (0, ∞), then taking the real part, we get
𝛽 ∞ { 𝛽 ∞ }
d b d
E3 (t) = − 0 g(s) |𝜔x (·, s, t)|2 dsdx + ℜ b0 g(s)𝜔x (·, s, t)utx dsdx .
dt 2 ∫0 ∫0 ds ∫0 ∫0
Using integration by parts with respect to s in the above equation with the help of (2.5) and the hypotheses (H), we
obtain { }
𝛽 ∞ 𝛽 ∞
d b
E3 (t) = 0 g′ (s)|𝜔x (·, s, t)|2 dsdx + ℜ b0 g(s)𝜔x (·, s, t)utx dsdx . (2.12)
dt 2 ∫0 ∫0 ∫0 ∫0
Finally, adding (2.9), (2.10), and (2.12), we obtain (2.8). The proof is thus complete.
Under the hypotheses (H) and from Lemma 2.1, system (2.2)–(2.6) is dissipative in the sense that its energy is
non-increasing with respect to time (i.e. E′ (t) ≤ 0). Now, we define the following Hilbert space  by
( )2
 ∶= H01 (0, L) × L2 (0, L) × g ,
where { }
g ∶= L2g ((0, ∞); HL1 (0, 𝛽)) and HL1 (0, 𝛽) ∶= 𝜔̃ ∈ H 1 (0, 𝛽) | 𝜔(0)
̃ =0 .
The space g is a Hilbert space of HL1 (0, 𝛽)-valued functions on (0, ∞), equipped with the following inner product
𝛽 ∞
(𝜔1 , 𝜔2 )g ∶= g(s)𝜔1x 𝜔2x dsdx, ∀ 𝜔1 , 𝜔2 ∈ g .
∫0 ∫0
The Hilbert space  is equipped with the following inner product
( ) L ( ) 𝛽 ∞
U, U 1 = ̃
b(·)ux ux + vv + 𝑦x 𝑦x + zz
1 1 1 1 dx + b
0 g(s)𝜔x (·, s)𝜔1x (·, s)dsdx, (2.13)
 ∫0 ∫0 ∫0
where U = (u, v, 𝑦, z, 𝜔(·, s))⊤ ∈  and U 1 = (u1 , v1 , 𝑦1 , z1 , 𝜔1 (·, s))⊤ ∈ . Now, we define the linear unbounded operator
 ∶ D() ⊂  →  by
{ ⊤
}
( v, 𝑦, z, 𝜔(·,
) s)) 2 ∈  | 𝑦 ∈ H (0, L) ∩ H0 (0, L), v, z ∈ H0 (0, L)
2 1 1
U = (u,
D() = , (2.14)
̃ (u, 𝜔)
Sb(·) ∈ L (0, L), 𝜔s (·, s) ∈ g , 𝜔(·, 0) = 0 in (0, 𝛽)
x
and
⎛ u ⎞ ⎛( v) ⎞
⎜ v ⎟ ⎜ Sb(·) ̃ (u, 𝜔) x − c(·)z ⎟
⎜ 𝑦 ⎟ = ⎜ z ⎟, (2.15)
⎜ z ⎟ ⎜ 𝑦 + c(·)v ⎟
⎜ 𝜔(·, s) ⎟ ⎜ xx ⎟
⎝ ⎠ ⎝ −𝜔s (·, s) + v ⎠
for all U = (u, v, 𝑦, z, 𝜔(·, s))⊤ ∈ D().
Now, if U = (u, ut , 𝑦, 𝑦t , 𝜔(·, s))⊤ , then system (2.2)–(2.6) can be written as the following first order evolution equation
Ut = U, U(0) = U0 , (2.16)
where U0 = (u0 (·, 0), u1 , 𝑦0 , 𝑦1 , 𝜔0 (·, s))⊤ ∈ .

Proposition 2.1. Under the hypotheses (H), the unbounded linear operator  is m-dissipative in the energy space .
Proof. For all U = (u, v, 𝑦, z, 𝜔(·, s))⊤ ∈ D(), from (2.13) and (2.15), we have
{ } { } { } { }
L L ( ) L L
ℜ(U, U) = ℜ ̃ x ux dx
b(·)v +ℜ ̃ (u, 𝜔) vdx
Sb(·) +ℜ zx 𝑦x dx +ℜ 𝑦xx zdx
∫0 ∫0 x ∫0 ∫0
{ 𝛽 ∞ } { 𝛽 ∞ }
+ ℜ b0 g(s)vx 𝜔x (·, s)dsdx −ℜ b0 g(s)𝜔xs (·, s)𝜔x (·, s)dsdx .
∫0 ∫0 ∫0 ∫0
6956 AKIL ET AL.
Using integration by parts to the second and fourth terms in the above equation, then using the fact that U ∈ D() ,
we obtain
{ 𝛽 }
∞
b0 𝛽 ∞ d
ℜ(U, U) = −ℜ b0 g(s)𝜔xs (·, s)𝜔x (·, s)dsdx =− g(s) |𝜔x (·, s)|2 dsdx.
∫0 ∫0 2 ∫0 ∫0 ds
Using integration by parts with respect to s in the above equation and the fact that 𝜔(·, 0) = 0 in(0, 𝛽) with the help of
hypotheses (H), we get
𝛽 ∞
b
ℜ(U, U) = 0 g′ (s)|𝜔x (·, s)|2 dsdx ≤ 0, (2.17)
2 ∫0 ∫0
which implies that  is dissipative. Now, let us prove that  is maximal. For this aim, let F =
(𝑓 1 , 𝑓 2 , 𝑓 3 , 𝑓 4 , 𝑓 5 (·, s))⊤ ∈ , we want to find U = (u, v, 𝑦, z, 𝜔(·, s))⊤ ∈ D() unique solution of
−U = F. (2.18)
Equivalently, we have the following system

−v = 𝑓 1 , (2.19)
( )
̃ (u, 𝜔) x + c(·)z = 𝑓 ,
2
− Sb(·) (2.20)
−z = 𝑓 3 , (2.21)
−𝑦xx − c(·)v = 𝑓 4 , (2.22)
𝜔s (·, s) − v = 𝑓 5 (·, s), (2.23)

with the following boundary conditions
u(0) = u(L) = 𝑦(0) = 𝑦(L) = 0, 𝜔(·, 0) = 0 in (0, 𝛽) and 𝜔(0, s) = 0 in (0, ∞). (2.24)
From (2.19), (2.23), and (2.24), we get
s
𝜔(x, s) = 𝑓 5 (x, 𝜉)d𝜉 − s𝑓 1 , (x, s) ∈ (0, 𝛽) × (0, ∞). (2.25)
∫0
Since v = −𝑓 1 ∈ H01 (0, L) and 𝑓 5 (·, s) ∈ g , then from (2.23) and (2.25), we get 𝜔s (·, s) ∈ g and 𝜔(·, s) ∈ HL1 (0, 𝛽) a.e.
∞
in (0, ∞). Now, to obtain that 𝜔(·, s) ∈ g , it is sufficient to prove that ∫0 g(s)||𝜔x (·, s)||2L2 ds < ∞ where || · ||L20,𝛽 ∶=
0,𝛽
|| · ||L2 (0,𝛽) . For this aim, let 𝜖 1 , 𝜖 2 > 0 , under the hypotheses (H), we have
𝜖2 2 𝜖
1
g(s)||𝜔x (·, s)||2L2 ds ≤ − g′ (s)||𝜔x (·, s)||2L2 ds. (2.26)
∫𝜖1 0,𝛽 m ∫𝜖1 0,𝛽
Using integration by parts in (2.26), we obtain
𝜖2 [ 𝜖2 ( ) ]
1 d
g(s)||𝜔x (·, s)||2L2 ds ≤ g(s) ||𝜔x (·, s)||L2 ds + g(𝜖1 )||𝜔x (·, 𝜖1 )||L2 − g (𝜖2 ) ‖𝜔x (·, 𝜖2 )‖L2 .
2 2 2
∫𝜖1 0,𝛽 m ∫𝜖1 ds 0,𝛽 0,𝛽 0,𝛽
Moreover, from Young's inequality, we have
2𝜖 ( ) 2 𝜖 { 𝛽 }
1 d 2
g(s) ||𝜔x (·, s)||2L2 ds = g(s)ℜ 𝜔x (·, s)𝜔sx (·, s)dx ds
m ∫𝜖1 ds 0,𝛽 m ∫𝜖1 ∫0
𝜖 𝜖 (2.27)
2 2
1 2
≤ g(s)||𝜔x (·, s)||2L2 ds + 2 g(s)||𝜔sx (·, s)||2L2 ds.
2 ∫𝜖1 0,𝛽 m ∫𝜖1 0,𝛽
AKIL ET AL. 6957
Inserting (2.27) in the above inequality, we get
𝜖2 2 𝜖
4 2 2
g(s)||𝜔x (·, s)||2L2 ds ≤ g(s)||𝜔sx (·, s)||2L2 ds + g(𝜖1 )||𝜔x (·, 𝜖1 )||2L2 − g (𝜖2 ) ‖𝜔x (·, 𝜖2 )‖2L2 .
∫𝜖1 0,𝛽
2 ∫
m 𝜖1 0,𝛽 m 0,𝛽 m 0,𝛽
Using the fact that 𝜔s (·, s) ∈ g , 𝜔(·, 0) = 0 in(0, 𝛽) and the hypotheses (H) in the above inequality, (in particular 1.2),
we obtain, as 𝜖 1 → 0+ and 𝜖 2 → ∞, that
∞
g(s)||𝜔x (·, s)||2L2 ds < ∞,
∫0 0,𝛽
and consequently, 𝜔(·, s) ∈ g . Now, see the definition of (Sb(·)

̃ (u, 𝜔)), substituting (2.19), (2.21), and (2.25) in (2.20)
and (2.22), we get the following system
[ ( s )]
Sb(·)
̃ u, 𝑓 5 (·, 𝜉)d𝜉 − s𝑓 1 + c(·)𝑓 3 = − 𝑓 2 , (2.28)
∫0 x
𝑦xx − c(·)𝑓 1 = − 𝑓 4 , (2.29)
u(0) = u(L) = 𝑦(0) = 𝑦(L) = 0, (2.30)

where ( ) { ( s )
b̃0 ux + b0 ∫0 g(s) ∫0 𝑓x5 (x, 𝜉) − s𝑓x1 , on (0, 𝛽),
s ∞
Sb(·)
̃ u, 𝑓 5 (·, 𝜉)d𝜉 − s𝑓 1 = (2.31)
∫0 aux , on (𝛽, L).
Let (𝜙, 𝜓) ∈ H01 (0, L) × H01 (0, L). Multiplying (2.28) and (2.29) by 𝜙 and 𝜓, respectively, integrating over (0, L), using
formal integrations by parts, then using the definition of (Sb(·) ̃
̃ (u, 𝜔)), (b(·)), and (c(·)), we obtain
L L 𝛾 𝛽 ∞ ( s )
̃
b(·)ux 𝜙x dx = 𝑓 2 𝜙dx + c0 𝑓 3 𝜙dx − b0 g(s) 𝑓x5 (·, 𝜉)d𝜉 − s𝑓x1 𝜙x dsdx, (2.32)
∫0 ∫0 ∫𝛼 ∫0 ∫0 ∫0
and
L L 𝛾
𝑦x 𝜓x dx = 𝑓 4 𝜓dx − c0 𝑓 1 𝜓dx. (2.33)
∫0 ∫0 ∫𝛼
Adding (2.32) and (2.33), we obtain
((u, 𝑦), (𝜙, 𝜓)) = (𝜙, 𝜓), ∀(𝜙, 𝜓) ∈ H01 (0, L) × H01 (0, L), (2.34)
where
L L
((u, 𝑦), (𝜙, 𝜓)) = ̃
b(·)ux 𝜙x dx + 𝑦x 𝜓x dx
∫0 ∫0
and
L ( ) 𝛾 ( ) 𝛽 ∞ ( s )
(𝜙, 𝜓) = 𝑓 2 𝜙 + 𝑓 4 𝜓 dx + c0 𝑓 3 𝜙 − 𝑓 1 𝜓 dx − b0 g(s) 𝑓x5 (·, 𝜉)d𝜉 − s𝑓x1 𝜙x dsdx.
∫0 ∫𝛼 ∫0 ∫0 ∫0
( )2
It is easy to see that,  is a sesquilinear, continuous, and coercive form on H01 (0, L) × H01 (0, L) and  is a antilinear
and continuous form on H01 (0, L) × H01 (0, L). Then, it follows by Lax–Milgram theorem that (2.34) admits a unique
solution (u, 𝑦) ∈ H01 (0, L) × H01 (0, L). By taking test functions (𝜙, 𝜓) ∈ ((0, L))2 , we see that (2.28)–(2.30) hold in the
distributional sense, from which we deduce that 𝑦 ∈ H 2 (0, L) ∩ H01 (0, L), while (Sb(·) ̃ (u, 𝜔))x ∈ L (0, L). Consequently,
2
U ∈ D() is a unique solution of 2.18. Then,  is an isomorphism, and since 𝜌 () is open set of C (see Theorem 6.7
[Chapter III] in Kato25 ), we easily get R(𝜆I − ) =  for a sufficiently small 𝜆 > 0. This, together with the dissipative-
ness of , imply that D () is dense in  and that  is m-dissipative in  (see Theorems 4.5 and 4.6 in Pazy26 ). The
proof is thus complete.
According to Lumer–Philips theorem (see Pazy26 ), Proposition 2.1 implies that the operator  generates a C0 semigroup
of contractions et in  which gives the well-posedness of (2.16). Then, we have the following result:
6958 AKIL ET AL.
Theorem 2.1. Under the hypotheses (H), for all U0 ∈ , system (2.16) admits a unique weak solution
U(x, s, t) = et U0 (x, s) ∈ C0 (R+ , ).
Moreover, if U0 ∈ D(), then the system (2.16) admits a unique strong solution
U(x, s, t) = et U0 (x, s) ∈ C0 (R+ , D()) ∩ C1 (R+ , ).
3 STRO NG STABILITY
( )
This section is devoted to the proof of the strong stability of the C0 semigroup et t≥0 . To obtain the strong stability of
( )
the C0 semigroup et t≥0 , we use the theorem of Arendt and Batty27 (see Theorem A2 in Appendix B1).
( )
Theorem 3.1. Assume that the hypotheses (H) hold. Then, the C0 semigroup of contraction et t≥0 is strongly stable in
; i.e., for all U0 ∈ , the solution of (2.16) satisfies
lim ||et U0 || = 0.

t→+∞
According to Theorem A2, to prove Theorem 3.1, we need to prove that the operator  has no pure imaginary
eigenvalues and 𝜎() ∩ iR is countable. The proof of Theorem 3.1 has been divided into the following two Lemmas.
Lemma 3.1. Under the hypothesis (H), we have
ker(i𝜆I − ) = {0}, ∀𝜆 ∈ R.
Proof. From Proposition 2.1, we have 0 ∈ 𝜌(). We still need to show the result for 𝜆 ∈ R⋆ . For this aim, suppose
that there exists a real number 𝜆 ≠ 0 and U = (u, v, 𝑦, z, 𝜔(·, s))⊤ ∈ D() such that
U = i𝜆U. (3.1)

v = i𝜆u, (3.2)
( )
̃ (u, 𝜔) x − c(·)z = i𝜆v,
Sb(·) (3.3)
z = i𝜆𝑦, (3.4)
𝑦xx + c(·)v = i𝜆z, (3.5)
−𝜔s (·, s) + v = i𝜆𝜔(·, s). (3.6)

From (2.17) and (3.1), we obtain
b0 𝛽 ∞ ′
0 = ℜ(i𝜆U, U) = ℜ(U, U) = g (s)|𝜔x (·, s)|2 dsdx. (3.7)
2 ∫0 ∫0
Thus, we have
𝜔x (·, s) = 0 in (0, 𝛽) × (0, ∞). (3.8)
From (3.8), we have
𝜔(·, s) = k(s) in (0, 𝛽) × (0, ∞), (3.9)
where k(s) is a constant depending on s. Then, from (3.9) and the fact that 𝜔(·, s) ∈ g (i.e. 𝜔(0, s) = 0), we get
𝜔(·, s) = 0 in (0, 𝛽) × (0, ∞). (3.10)

AKIL ET AL. 6959
From (3.2), (3.6), and (3.10), we obtain

u = v = 0 in (0, 𝛽). (3.11)
Inserting (3.2) and (3.4) in (3.3) and (3.5), then using (3.8) together with the definition of (Sb(·) ̃
̃ (u, 𝜔)) and (b(·)), we
obtain the following system:
̃
𝜆2 u + (b(·)ux )x − c(·)i𝜆𝑦 = 0, in (0, L), (3.12)
𝜆2 𝑦 + 𝑦xx + c(·)i𝜆u = 0, in (0, L), (3.13)
u(0) = u(L) = 𝑦(0) = 𝑦(L) = 0. (3.14)

From (3.11), (3.12), the definition of (c(·)) and (3.4), we obtain
𝑦 = z = 0 in (𝛼, 𝛽). (3.15)
Now, from (3.15) and the fact that y ∈ C1 ([0, L]), we get
𝑦(𝛼) = 𝑦x (𝛼) = 0. (3.16)
Next, from (3.13), (3.16), and the definition of (c(·)), we obtain the following system
𝜆2 𝑦 + 𝑦xx = 0, in (0, 𝛼), (3.17)
𝑦(0) = 𝑦(𝛼) = 𝑦x (𝛼) = 0. (3.18)

Thus, from the above system and by using Holmgren uniqueness theorem, we obtain
𝑦 = 0 in (0, 𝛼). (3.19)
Therefore, from (3.4) and (3.19), we obtain
𝑦 = z = 0 in (0, 𝛼). (3.20)
̃ (u, 𝜔)) and the fact that b̃0 = a − b0 g

According to the definition of (Sb(·) ̃ , we obtain
{ ∞
aux − b0 g̃ ux + b0 ∫0 g(s)𝜔x (x, s)ds, on(0, 𝛽),
̃ (u, 𝜔) =
Sb(·) (3.21)
aux , on(𝛽, L).
From (3.8), (3.11), and (3.21), we get

( )
̃ (u, 𝜔) = aux in (0, L) and consequently Sb(·)
Sb(·) ̃ (u, 𝜔) x = auxx in (0, L). (3.22)
Thus, from (3.22) and the fact that U ∈ D(), we obtain
uxx ∈ L2 (0, L) and consequently u ∈ C1 ([0, L]). (3.23)
Now, from (3.11), (3.15), (3.23), and the fact that y ∈ C1 ([0, L]), we obtain
u(𝛽) = ux (𝛽) = 𝑦(𝛽) = 𝑦x (𝛽) = 0. (3.24)
̃
Next, from the definition of (b(·)) and (c(·)), the System (3.12)–(3.13) can be written in (𝛽, 𝛾) as the following system
𝜆2 u + auxx − c0 i𝜆𝑦 = 0, in (𝛽, 𝛾), (3.25)
𝜆2 𝑦 + 𝑦xx + c0 i𝜆u = 0, in (𝛽, 𝛾), (3.26)

6960 AKIL ET AL.
u(𝛽) = ux (𝛽) = 𝑦(𝛽) = 𝑦x (𝛽) = 0. (3.27)

Let V = (u, ux , 𝑦, 𝑦x )⊤ , then system (3.25)–(3.27) can be written as the following
Vx = BV, V(𝛽) = 0. (3.28)
where
⎛ 0 1 0 0⎞
⎜ −a−1 𝜆2 0 a−1 i𝜆c0 0⎟
B=⎜ 0 0 0 1⎟.
⎜ ⎟
⎝ −i𝜆c0 0 −𝜆2 0⎠
The solution of the differential Equation (3.28) is given by
V(x) = eB(x−𝛽) V(𝛽). (3.29)
Thus, from (3.29) and the fact that V(𝛽) = 0, we get
V = 0 in (𝛽, 𝛾) and consequently u = ux = 𝑦 = 𝑦x = 0 in (𝛽, 𝛾). (3.30)
So, from (3.2), (3.4), and (3.30), we get
u = v = 0 in (𝛽, 𝛾) and 𝑦 = z = 0 in (𝛽, 𝛾). (3.31)
Now, from (3.30) and the fact that u, y ∈ C1 ([0, L]), we obtain
u(𝛾) = ux (𝛾) = 𝑦(𝛾) = 𝑦x (𝛾) = 0. (3.32)
̃
Next, from the definition of (b(·)) and (c(·)), the system (3.12)–(3.13) can be written in (𝛾, L) as the following system
𝜆2 u + auxx = 0, in (𝛾, L), (3.33)
𝜆2 𝑦 + 𝑦xx = 0, in (𝛾, L), (3.34)
u(L) = u(𝛾) = ux (𝛾) = 0, (3.35)
𝑦(L) = 𝑦(𝛾) = 𝑦x (𝛾) = 0. (3.36)

From the above system and by using Holmgren uniqueness theorem, we deduce that
u = 𝑦 = 0 in (𝛾, L). (3.37)
Thus, from (3.2), (3.4) and (3.37), we obtain
u = v = 0 in (𝛾, L) and 𝑦 = z = 0 in (𝛾, L). (3.38)
Finally, from (3.10) (3.11), (3.15), (3.20), (3.31), and (3.38), we obtain
U = 0. (3.39)
The proof is thus complete.

Lemma 3.2. Under the hypotheses (H), for all 𝜆 ∈ R, we have
R(i𝜆I − ) = .
AKIL ET AL. 6961
Proof. From Proposition 2.1, we have 0 ∈ 𝜌(). We still need to show the result for 𝜆 ∈ R⋆ . For this aim, let F =
(𝑓 1 , 𝑓 2 , 𝑓 3 , 𝑓 4 , 𝑓 5 (·, s))⊤ ∈ , we want to find U = (u, v, 𝑦, z, 𝜔(·, s))⊤ ∈ D() solution of
(i𝜆I − )U = F. (3.40)

i𝜆u − v = 𝑓 1 , (3.41)
( )
̃ (u, 𝜔)
i𝜆v − Sb(·) + c(·)z = 𝑓 2 , (3.42)
x
i𝜆𝑦 − z = 𝑓 3 , (3.43)
i𝜆z − 𝑦xx − c(·)v = 𝑓 4 , (3.44)
i𝜆𝜔(·, s) + 𝜔s (·, s) − v = 𝑓 5 (·, s), (3.45)

u(0) = u(L) = 𝑦(0) = 𝑦(L) = 0, 𝜔(·, 0) = 0 in (0, 𝛽) and 𝜔(0, s) = 0 in (0, ∞). (3.46)
From (3.41), (3.45), and (3.46), we have

s
1
𝜔(x, s) = (i𝜆u − 𝑓 1 )(1 − e−i𝜆s ) + 𝑓 5 (x, 𝜉)ei𝜆(𝜉−s) d𝜉, (x, s) ∈ (0, 𝛽) × (0, ∞). (3.47)
i𝜆 ∫0
̃ (u, 𝜔)), inserting (3.41), (3.43), and (3.47) in (3.42) and (3.44), we obtain the following
Due the definition of (Sb(·)
system
⎧ 2 ( )
̃ (u, 𝜔) x + i𝜆c(·)𝑦 = F1 ∶= 𝑓 + c(·)𝑓 + i𝜆𝑓 ,
2 3 1
⎪ −𝜆 u − Sb(·)
⎨ −𝜆 𝑦 − 𝑦xx − i𝜆c(·)u = F2 ∶= 𝑓 − c(·)𝑓 + i𝜆𝑓 3 ,
2 4 1 (3.48)
⎪ u(0) = u(L) = 𝑦(0) = 𝑦(L) = 0,
⎩
̃ (u, 𝜔) takes the form
where here Sb(·)
{
b̂0 ux +
b0 ∞ ∞ s
∫0 g(s)(1 − e−i𝜆s )𝑓x1 ds + b0 ∫0 g(s) ∫0 𝑓x5 (x, 𝜉)ei𝜆(𝜉−s) d𝜉ds, on (0, 𝛽),
̃ (u, 𝜔) =
Sb(·) i𝜆
aux , on (𝛽, L),
and ∞
b̂0 ∶= a − b0 g(s)e−i𝜆s ds.
∫0
Let (𝜙, 𝜓) ∈ H01 (0, L) × H01 (0, L). Multiplying the first equation of (3.48) and the second equation of (3.48) by 𝜙 and 𝜓
respectively, integrating over (0, L), then using formal integrations by parts, we obtain
L L 𝛽 ∞
̂ b
− 𝜆2 u𝜙dx + b(·)ux 𝜙x dx + 0 g(s)(1 − e−i𝜆s )𝑓x1 𝜙x dsdx
∫0 ∫0 i𝜆 ∫0 ∫0
𝛽 ∞ s 𝛾 L
(3.49)
+ b0 g(s) ei𝜆(𝜉−s)
𝑓x5 (·, 𝜉)𝜙x d𝜉dsdx + i𝜆c0 𝑦𝜙dx = F1 𝜙dx
∫0 ∫0 ∫0 ∫𝛼 ∫0
and
L L 𝛾 L
−𝜆2 𝑦𝜓dx + 𝑦x 𝜓x dx − i𝜆c0 u𝜓dx = F2 𝜓dx, (3.50)
∫0 ∫0 ∫𝛼 ∫0
where {
̂
b(x) = b̂0 , x ∈ (0, 𝛽),
a, x ∈ (𝛽, L).
Adding (3.49) and (3.50), we get
((u, 𝑦), (𝜙, 𝜓)) = (𝜙, 𝜓), ∀(𝜙, 𝜓) ∈ V ∶= H01 (0, L) × H01 (0, L), (3.51)
6962 AKIL ET AL.
where
((u, 𝑦), (𝜙, 𝜓)) = 1 ((u, 𝑦), (𝜙, 𝜓)) + 2 ((u, 𝑦), (𝜙, 𝜓))
with {
1 ((u, 𝑦), (𝜙, 𝜓)) = ∫0 ̂
L L
b(·)ux 𝜙x dx + ∫0 𝑦x 𝜓x dx,
2∫L L (3.52)
2 ((u, 𝑦), (𝜙, 𝜓)) = −𝜆 0 (u𝜙 + 𝑦𝜓)dx − i𝜆c0 ∫0 (u𝜓 − 𝑦𝜙)dx,
and
( s )
L
b0 𝛽 ∞ 𝛽 ∞
(𝜙, 𝜓) = (F1 𝜙 + F2 𝜓)dx − g(s)(1 − e−i𝜆s )𝑓x1 𝜙x dsdx − b0 g(s) ei𝜆(𝜉−s) 𝑓x5 (·, 𝜉)d𝜉 𝜙x dsdx.
∫0 i𝜆 ∫0 ∫0 ∫0 ∫0 ∫0
Let V′ be the dual space of V. Let us define the following operators
B ∶ V → V′ Bi ∶ V → V′
(u, 𝑦) → B(u, 𝑦) and (u, 𝑦) → Bi (u, 𝑦) , i ∈ {1, 2}, (3.53)
such that {
(B(u, 𝑦))(𝜙, 𝜓) = ((u, 𝑦), (𝜙, 𝜓)), ∀(𝜙, 𝜓) ∈ V,
(Bi (u, 𝑦))(𝜙, 𝜓) = i ((u, 𝑦), (𝜙, 𝜓)), ∀(𝜙, 𝜓) ∈ V, i ∈ {1, 2}. (3.54)
We need to prove that the operator B is an isomorphism. For this aim, we divide the proof into three steps:
Step 1. In this step, we want to prove that the operator B1 is an isomorphism. For this aim, it is easy to see that 1
is sesquilinear, continuous and coercive form on V. Then, from (3.54) and Lax-Milgram theorem, the operator B1 is
an isomorphism.
Step 2. In this step, we want to prove that the operator B2 is compact. For this aim, from (3.52) and (3.54), we have
|2 ((u, 𝑦), (𝜙, 𝜓))| ≲ ||(u, 𝑦)||(L2 (0,L))2 ||(𝜙, 𝜓)||(L2 (0,L))2 , (3.55)
( )2
and consequently, using the compact embedding from V into L2 (0, L) , we deduce that B2 is a compact operator.
Therefore, from the above steps, we obtain that the operator B = B1 + B2 is a Fredholm operator of index zero. Now,
following Fredholm alternative, we still need to prove that the operator B is injective to obtain that the operator B is
an isomorphism.
Step 3. In this step, we want to prove that the operator B is injective (i.e., ker(B) = {0}). For this aim, let (u, ̃ ∈ ker(B)
̃ 𝑦)
which gives
̃ (𝜙, 𝜓)) = 0, ∀(𝜙, 𝜓) ∈ V.
̃ 𝑦),
((u,
Equivalently, we have
L L L L
̂
b(·)ũ x 𝜙x dx + 𝑦̃x 𝜓x dx − 𝜆2 ̃ + 𝑦𝜓)dx
(u𝜙 ̃ − i𝜆 ̃ − 𝑦𝜙)dx
c(·)(u𝜓 ̃ = 0, ∀(𝜙, 𝜓) ∈ V.
∫0 ∫0 ∫0 ∫0
Thus, we find that

⎧ ̂
⎪ −𝜆 ũ − (b(·)ũ x )x + i𝜆c(·)𝑦̃ = 0,
2
⎨ −𝜆 𝑦̃ − 𝑦̃xx − i𝜆c(·)ũ = 0,
2
⎪ u(0)
̃ ̃
= u(L) = 𝑦(0)
̃ = 𝑦(L)̃ = 0.
⎩
Therefore, the vector Ũ defined by
Ũ = (u,
̃ i𝜆u,
̃ 𝑦,
̃ i𝜆𝑦, ̃ ⊤
̃ (1 − e−i𝜆s )u)
belongs to D() and satisfies
i𝜆Ũ − Ũ = 0,
and consequently Ũ ∈ ker(i𝜆I − ). Then, according to Lemma 3.1, we obtain Ũ = 0 and consequently ũ = 𝑦̃ = 0 and
ker(B) = {0}. Finally, from Step 3 and Fredholm alternative, we deduce that the operator B is isomorphism. It is easy
to see that the operator  is a antilinear and continuous form on V. Consequently, (3.51) admits a unique solution
(u, 𝑦) ∈ V. By using the classical elliptic regularity, we deduce that U ∈ D() is a unique solution of (3.40). The proof
is thus complete.
AKIL ET AL. 6963
Proof of Theorem 3.1. From Lemma 3.1, we obtain the the operator  has no pure imaginary eigenvalues (i.e. 𝜎p () ∩
iR = ∅). Moreover, from Lemma 3.2 and with the help of the closed graph theorem of Banach, we deduce that
𝜎() ∩ iR = ∅. Therefore, according to Theorem A2, we get that the C0 semigroup (et )t≥0 is strongly stable. The
Remark 3.1. We mention Akil et al.28 for a direct approach of the strong stability of Timoshenko system in the absence
of compactness of the resolvent.
4 EXPONENTIAL AND POLY NOMIAL STA BILITY
In this section, under the hypotheses (H), we show the influence of the ratio of the wave propagation speed on the stability
of system (2.2)–(2.6). Our main result in this part is the following theorems.
Theorem 4.1. Under the hypotheses (H), if a = 1, then the C0 semigroup et is exponentially stable; i.e., there exists
constants M ≥ 1 and 𝜖 > 0 independent of U0 such that
||et U0 || ≤ Me−𝜖t ||U0 || .
Theorem 4.2. Under the hypotheses (H), if a ≠ 1, then there exists C > 0 such that for every U0 ∈ D(), we have
C
E(t) ≤ ||U0 ||2D() , t > 0.
t
Since iR ⊂ 𝜌() (see Section 3), according to Theorems A3 and A4, to proof Theorems 4.1 and 4.2, we still need to prove
the following condition
‖ ‖ ( )
sup ‖(i𝜆I − )−1 ‖ = O |𝜆|𝓁 , with 𝓁 = 0 for Theorem 4.1 and 𝓁 = 2 for Theorem 2.4. (H1 )
𝜆∈R
‖ ‖ ()
We will prove condition (H1 ) by a contradiction

{ } argument. For this purpose, suppose that (H1 ) is false, then there exists
(𝜆n , U n ∶= (un , vn , 𝑦n , zn , 𝜔n (·, s))⊤ ) n≥1 ⊂ R∗ × D() with
|𝜆n | → ∞ and ||U n || = ||(un , vn , 𝑦n , zn , 𝜔n (·, s))⊤ || = 1, (4.1)
such that
(𝜆n )𝓁 (i𝜆n I − )U n = F n ∶= (𝑓 1,n , 𝑓 2,n , 𝑓 3,n , 𝑓 4,n , 𝑓 5,n (·, s))⊤ → 0 in . (4.2)
For simplicity, we drop the index n. Equivalently, from (4.2), we have
i𝜆u − v = 𝜆−𝓁 𝑓 1 → 0 in H01 (0, L), (4.3)
( )
̃ (u, 𝜔)
i𝜆v − Sb(·) + c(·)z = 𝜆−𝓁 𝑓 2 → 0 in L2 (0, L), (4.4)
x
i𝜆𝑦 − z = 𝜆−𝓁 𝑓 3 → 0 in H01 (0, L), (4.5)
i𝜆z − 𝑦xx − c(·)v = 𝜆−𝓁 𝑓 4 → 0 in L2 (0, L), (4.6)
i𝜆𝜔(·, s) + 𝜔s (·, s) − v = 𝜆−𝓁 𝑓 5 (·, s) → 0 in g . (4.7)
Here we will check the condition (H1 ) by finding a contradiction with (4.1) by showing ‖U‖ = o(1). For clarity, we
divide the proof into several lemmas.
6964 AKIL ET AL.
Lemma 4.1. Under the hypotheses (H), the solution U = (u, v, 𝑦, z, 𝜔(·, s))⊤ ∈ D() of system (4.3)–(4.7) satisfies the
following estimations:
𝛽 ∞ ( ) 𝛽 ∞ ( )
− g′ (s)|𝜔x (·, s)|2 dsdx = o |𝜆|−𝓁 and g(s)|𝜔x (·, s)|2 dsdx = o |𝜆|−𝓁 , (4.8)
∫0 ∫0 ∫0 ∫0
𝛽 𝛽
| |2
|ux |2 dx = o(|𝜆|−𝓁 ) and |Sb̃0 (u, 𝜔)| dx = o(|𝜆|−𝓁 ). (4.9)
∫0 ∫0 | |
Proof. First, taking the inner product of (4.2) with U in  and using (2.17), we get
b0 𝛽 ∞ ′
− g (s)|𝜔x (·, s)|2 dsdx = −ℜ(U, U) = 𝜆−𝓁 ℜ(F, U) ≤ |𝜆|−𝓁 ||F|| ||U|| . (4.10)
2 ∫0 ∫0
Thus, from (4.10), (H) and the fact that ||F|| = o(1) and ||U|| = 1, we obtain the first estimation in (4.8). From
hypotheses (H), we obtain
𝛽 ∞ 𝛽 ∞
1
g(s)|𝜔x (·, s)|2 dsdx ≤ − g′ (s)|𝜔x (·, s)|2 dsdx. (4.11)
∫0 ∫0 m ∫0 ∫0
Then, from the first estimation in (4.8) and (4.11), we obtain the second estimation in (4.8). Next, inserting (4.3)
in (4.7), then deriving the resulting equation with respect to x, we get
i𝜆𝜔x (·, s) + 𝜔sx (·, s) − i𝜆ux = 𝜆−𝓁 𝑓x5 (·, s) − 𝜆−𝓁 𝑓x1 in (0, 𝛽) × (0, ∞). (4.12)
Multiplying (4.12) by 𝜆−1 g(s)ux , integrating over (0, 𝛽) × (0, ∞), then taking the imaginary part, we obtain
𝛽 ∞ { 𝛽 ∞ } { 𝛽 ∞ }
g(s)|ux |2 dsdx = ℑ i g(s)𝜔x (·, s)ux dsdx +ℑ 𝜆−1 g(s)𝜔xs (·, s)ux dsdx
∫0 ∫0 ∫0 ∫0 ∫0 ∫0
{ 𝛽 ∞ } { 𝛽 ∞ }
− ℑ 𝜆−(𝓁+1) g(s)𝑓x5 (·, s)ux dsdx + ℑ 𝜆−(𝓁+1) g(s)𝑓x1 ux dsdx .
∫0 ∫0 ∫0 ∫0
Using integration by parts with respect to s in the above equation, then using hypotheses (H) and the fact that 𝜔(·, 0) =
0 in(0, 𝛽), we get
𝛽 { 𝛽 ∞ } { 𝛽 ∞ }
g̃ |ux |2 dx = ℑ i g(s)𝜔x (·, s)ux dsdx +ℑ 𝜆−1 −g′ (s)𝜔x (·, s)ux dsdx
∫0 ∫0 ∫0 ∫0 ∫0
{ 𝛽 ∞ } { 𝛽 } (4.13)
− ℑ 𝜆−(𝓁+1) g(s)𝑓x5 (·, s)ux dsdx + ℑ g̃ 𝜆−(𝓁+1) 𝑓x1 ux dx .
∫0 ∫0 ∫0
Using Young's inequality and Cauchy–Schwarz inequality in (4.13) with the help of hypotheses (H), we obtain
𝛽 𝛽 𝛽
g̃ 1
∞
g̃ |ux |2 dx ≤ |ux |2 dx + g(s)|𝜔x (·, s)|2 dsdx
∫0 2 ∫0 2 ∫0 ∫0
( 𝛽 ∞ ) 12 ( 𝛽 ) 12
√
+ |𝜆| −1
g0 ′ 2
−g (s)|𝜔x (·, s)| dsdx |ux | dx
2
∫0 ∫0 ∫0
( 𝛽 ∞ ) 12 ( 𝛽 ) 12
√
+ |𝜆| −(𝓁+1)
g̃ 5 2
g(s)|𝑓x (·, s)| dsdx |ux | dx
2
∫0 ∫0 ∫0
( 𝛽 ) 12 ( 𝛽 ) 12
+ g̃ |𝜆|−(𝓁+1)
|𝑓x | dx
1 2
|ux | dx .
2
∫0 ∫0
AKIL ET AL. 6965
From the above inequality, (4.8) and the fact that ux is uniformly bounded in L2 (0, L) and 𝑓x1 → 0 in L2 (0, L), 𝑓 5 (·, s) →
0 in g , we obtain the first estimation in (4.9). Now, by using Cauchy–Schwarz inequality, we obtain
𝛽 𝛽 | ∞ |2 ( )2 𝛽 𝛽( ∞ )2
| |2 |b̃0 ux + b0 | ̃
|Sb̃0 (u, 𝜔)| dx = | g(s)𝜔x (·, s)ds| dx ≤ 2 b0 |ux | + 2b0
2 2
g(s)|𝜔x (·, s)|ds dx
∫0 | | ∫0 | ∫0 | ∫0 ∫0 ∫0
( )2 𝛽 𝛽 ∞
≤ 2 b̃0 |ux |2 + 2b20 g̃ g(s)|𝜔x (·, s)|2 dsdx.
∫0 ∫0 ∫0
Finally, from the above inequality, (4.8) and the first estimation in (4.9), we obtain the second estimation in (4.9). The
( )
Lemma 4.2. Let 0 < 𝜀 < min 𝛼, 𝛽−𝛼 5
. Under the hypotheses (H), the solution U = (u, v, 𝑦, z, 𝜔(·, s))⊤ ∈ D()
of (4.3)–(4.7) satisfies the following estimation:
𝛽−𝜀 ( 𝓁
)
|v|2 dx = o |𝜆|− 2 . (4.14)
∫𝜀
Proof. First, we fix a cut-off function h1 ∈ C1 ([0, 𝛽]) (see Figure 2) such that 0 ≤ h1 (x) ≤ 1, for all x ∈ [0, 𝛽] and
{
1 if x ∈ [𝜀, 𝛽 − 𝜀],
h1 (x) = 0 if x ∈ {0, 𝛽}, h1
and set
max |h′1 (x)| = Mh′1 .
x∈[0,𝛽]
From (4.4), we deduce that

( )
i𝜆v − Sb̃ (u, 𝜔) + c(·)z = 𝜆−𝓁 𝑓 2 in (0, 𝛽).
0 x
∞
Multiplying the above equation by −h1 ∫0
g(s)𝜔(·, s)ds and integrate over (0, 𝛽), using integration by parts with the
help of the properties of (h1 ) (i.e., h1 (0) = h1 (𝛽) = 0), then using the definition of (c(·)), we obtain
𝛽 ∞ 𝛽 ( ∞ )
−i𝜆 h1 v g(s)𝜔(·, s)dsdx = Sb̃ (u, 𝜔) h1 g(s)𝜔(·, s) dsdx
∫0 ∫0 ∫0 0 ∫0 x
𝛽 ∞ 𝛽 ∞
(4.15)
+ c0 h1 z g(s)𝜔(·, s)dsdx − 𝜆 −𝓁
h1 𝑓 2
g(s)𝜔(·, s)dsdx.
∫𝛼 ∫0 ∫0 ∫0
−i𝜆𝜔(·, s) = −𝜔s (·, s) + v + 𝜆−𝓁 𝑓 5 (·, s) in (0, 𝛽) × (0, ∞).
Inserting the above equation in the left-hand side of (4.15), then using the definition of c(·) and (h1 ), we get
𝛽 𝛽 ∞ 𝛽 ∞
g̃ h1 |v|2 dx = h1 v g(s)𝜔s (·, s)dsdx − 𝜆−𝓁 h1 v g(s)𝑓 5 (·, s)dsdx
∫0 ∫0 ∫0 ∫0 ∫0
𝛽 ∞ 𝛽 ∞
+ Sb̃ (u, 𝜔)h′1 g(s)𝜔(·, s)dsdx + Sb̃ (u, 𝜔)h1 g(s)𝜔x (·, s)dsdx (4.16)
∫0 0 ∫0 ∫0 0 ∫0
𝛽 ∞ 𝛽 ∞
+ c0 h1 z g(s)𝜔(·, s)dsdx − 𝜆−𝓁 h1 𝑓 2 g(s)𝜔(·, s)dsdx.
∫𝛼 ∫0 ∫0 ∫0
6966 AKIL ET AL.
Using integration by parts with respect to s with the help of 𝜔(·, 0) = 0 in(0, 𝛽) and hypotheses (H), Cauchy–Schwarz
inequality, Poincaré inequality, v is uniformly bounded in L2 (0, L), and (4.9), we get
| 𝛽 ∞ | | 𝛽 ∞ |
| | | |
| h1 v g(s)𝜔s (·, s)dsdx| = | h1 v −g′ (s)𝜔(·, s)dsdx|
|∫0 ∫0 | |∫0 ∫0 |
| | | |
( 𝛽 ) 12 ( 𝛽 ∞ ) 12
√
≤ g0 |v| dx
2 ′ 2
−g (s)|𝜔(·, s)| dsdx (4.17)
∫0 ∫0 ∫0
( 𝛽 ) 12 ( 𝛽 ∞ ) 12 ( )
√ 𝓁
≲ g0 |v| dx
2 ′ 2
−g (s)|𝜔x (·, s)| dsdx = o |𝜆|− 2 .
∫0 ∫0 ∫0
Using the definition of (h1 ), Cauchy–Schwarz inequality, Poincaré inequality, (4.8) and the fact that v and z are
uniformly bounded in L2 (0, L) and ||𝑓 2 ||L2 (0,L) = o(1), 𝑓 5 → 0 ing , we get
⎧| √ ( 𝛽 )1 ( )1
⎪ |𝜆−𝓁 ∫0 h1 v ∫0 g(s)𝑓 5 (·, s)dsdx|| ≲ 1 𝓁 g̃ ∫0 |v|2 dx
𝛽 ∞ 𝛽 ∞ o(1)
∫ ∫ ,
2 5 2 2
g(s)|𝑓 (·, s)| dsdx = |𝜆|
⎪ | | |𝜆| 0 0 x 𝓁
√ ( )( )1
⎪ ||c ∫ 𝛽 h z ∫ ∞ g(s)𝜔(·, s)dsdx|| ≤ |c | g̃ ∫ 𝛽 |z|2 dx ∫ 𝛽 ∫ ∞ g(s)|𝜔(·, s)|2 dsdx 2
⎪| 0 𝛼 1 0 | 0 𝛼 𝛼 0
⎨ √ ( 𝛽 2 )( 𝛽 ∞ )1 (4.18)
≲ |c0 | g̃ ∫𝛼 |z| dx ∫0 ∫0 g(s)|𝜔x (·, s)|2 dsdx = o(1)𝓁 ,
2
⎪
⎪ |𝜆| 2
⎪ | −𝓁 𝛽 ∞ | 1
√ ( 𝛽 2 2 ) 12 ( 𝛽 ∞ )1
⎪ ||𝜆 ∫0 h1 𝑓2 ∫0 g(s)𝜔(·, s)dsdx|| ≲ |𝜆|𝓁 g̃ ∫0 |𝑓 | dx ∫0 ∫0 g(s)|𝜔x (·, s)| dsdx = o(1)3𝓁 .
2 2
⎩ |𝜆| 2
On the other hand, we have

{
|Sb̃ (u, 𝜔)||h′1 |g(s)|𝜔(·, s)| ≤ 12 |h′1 ||Sb̃ (u, 𝜔)|2 g(s) + 12 |h′1 ||𝜔(·, s)|2 g(s),
0 0
|Sb̃ (u, 𝜔)||h1 |g(s)|𝜔x (·, s)| ≤ 12 |h1 ||Sb̃ (u, 𝜔)|2 g(s) + 12 |h1 ||𝜔x (·, s)|2 g(s).
0 0
̃ (u, 𝜔)) and (h1 ), Poincaré inequality and estimations (4.8)

Then, from the above inequalities, the definition of (Sb(·)
and (4.9), we obtain
M ( )
⎧ ||∫ 𝛽 S ̃ (u, 𝜔)h′ ∫ ∞ g(s)𝜔(·, s)dsdx|| ≤ h′1 g̃ ∫ 𝛽 |S ̃ (u, 𝜔)|2 dx + Cp ∫ 𝛽 ∫ ∞ g(s)|𝜔x (·, s)|2 dsdx
⎪| 0 b 0 1 0 | 2 0 b 0 0 0
⎨ = o(|𝜆|−𝓁 ), (4.19)
⎪ ||∫ 𝛽 S (u, 𝜔)h ∫ ∞ g(s)𝜔 (·, s)dsdx|| ≤ g̃ ∫ 𝛽 |S (u, 𝜔)|2 dx + ∫ 𝛽 ∫ ∞ g(s)|𝜔 (·, s)|2 dsdx = o(|𝜆|−𝓁 ),
⎩ | 0 b̃0 1 0 x
| 2 0 b̃0 0 0 x
where Cp > 0 is a Poincaré constant. Inserting inequalities (4.17)–(4.19) in (4.16), we obtain
𝛽 ( 𝓁
)
h1 |v|2 dx = o |𝜆|− 2 .
∫0
Finally, from the above estimation and the definition of (h1 ), we obtain the desired result (4.14). The proof is thus
complete.
( )
. Under the hypotheses (H), the solution the solution U = (u, v, 𝑦, z, 𝜔(·, s))⊤ ∈
D() of (4.3)–(4.7) satisfies the following estimation
𝛽−2𝜀
|a − 1||𝜆| 𝛽−𝜀
|𝑦x |2 dx ≤ |ux ||𝑦x |dx + o(1). (4.20)
∫𝛼+𝜀 |c0 | ∫𝛼
AKIL ET AL. 6967
FIGURE 2 Geometric description of

the functions h1 , h2 , and (h3 ) [Colour
figure can be viewed at
wileyonlinelibrary.com]
Proof. First, we fix a cut-off function h2 ∈ C1 ([0, L]) (see Figure 2) such that 0 ≤ h2 (x) ≤ 1, for all x ∈ [0, L] and
{
0 if x ∈ [0, 𝛼] ∪ [𝛽 − 𝜀, L],
h2 (x) = 1 if x ∈ [𝛼 + 𝜀, 𝛽 − 2𝜀], (h2 )
From (4.6), i𝜆−1 h2 𝑦xx is uniformly bounded in L2 (0, L). Multiplying (4.4) by i𝜆−1 h2 𝑦xx , using integration by parts over
(0, L) and over (𝛼, 𝛽 − 𝜀), the definitions of c(·) and (h2 ), and using the fact that ||𝑓 2 ||L2 (0,L) = o(1), we get
L L
i
L ( ) ic0 𝛽−𝜀 ( ′ ) o(1)
h′2 v𝑦x dx + h2 vx 𝑦x dx − ̃ (u, 𝜔) 𝑦xx dx −
h2 Sb(·) h2 z𝑦x + h2 zx 𝑦x dx = . (4.21)
∫0 ∫0 𝜆 ∫0 x 𝜆 ∫𝛼 |𝜆|𝓁
From (4.3) and (4.5), we obtain
i
vx = i𝜆ux − 𝜆−𝓁 𝑓x1 and − zx = 𝑦x + i𝜆−(𝓁+1) 𝑓x3 .
𝜆
Inserting the above equations in (4.21) and taking the real part, we get
{ L( } { } { L }
L L
i )
c0 h |𝑦 | dx + ℜ i𝜆
2
h u 𝑦 dx − ℜ ̃ (u, 𝜔) x h2 𝑦xx dx
Sb(·) = −ℜ ′
h v𝑦 dx
∫0 2 x ∫0 2 x x 𝜆 ∫0 ∫0 2 x
{ } { } { } (4.22)
1
L
c0 𝛽−𝜀 ′ ic0 𝛽−𝜀
o(1)
+ℜ h 2 𝑓 1
𝑦 x dx + ℜ i h z𝑦 x dx − ℜ h2 𝑓x3 𝑦x dx + .
𝜆 ∫0 𝜆 ∫𝛼 ∫𝛼
x
𝓁 2
𝜆 𝓁+1 |𝜆|𝓁
Using the fact that yx is uniformly bounded in L2 (0, L), ||𝑓x1 ||L2 (0,L) = o(1) and ||𝑓x3 ||L2 (0,L) = o(1), we get
{ L } { 𝛽−𝜀 }
ℜ 𝜆−𝓁 h2 𝑓x1 𝑦x dx = o(|𝜆|−𝓁 ) and − ℜ ic0 𝜆−(𝓁+1) h2 𝑓x3 𝑦x dx = o(|𝜆|−(𝓁+1) ). (4.23)
∫0 ∫𝛼
Using Cauchy–Schwarz inequality, the definition of (h2 ), yx and z are uniformly bounded in L2 (0, L), and estima-
tion (4.14), we get
{ } ( ) { }
L
− 𝓁4 c0 𝛽−𝜀 ′ ( )
−ℜ h′2 v𝑦x dx = o |𝜆| and ℜ i h2 z𝑦x dx = O |𝜆|−1 = o(1). (4.24)
∫0 𝜆 ∫𝛼
Inserting (4.23) and (4.24) in (4.22), then using the definition of (h2 ), we get
𝛽−𝜀 { 𝛽−𝜀 } { 𝛽−𝜀 ( ) }

i
c0 h2 |𝑦x | dx + ℜ
2
i𝜆 h2 ux 𝑦x dx −ℜ Sb̃ (u, 𝜔) h2 𝑦xx dx = o(1). (4.25)
∫𝛼 ∫𝛼 𝜆 ∫𝛼 0 x
( ) ( )
̃ (u, 𝜔)
From (4.4), i𝜆−1 h2 Sb(·) is uniformly bounded in L 2 (0, L). Multiplying (4.6) by i𝜆−1 S (u, 𝜔) , using inte-
̃
b(·)
x x
gration by parts over (0, L) and over (𝛼, 𝛽 − 𝜀), the definitions of c(·), (h2 ), and (Sb(·)
̃ (u, 𝜔)), and the fact that ||𝑓 ||L2 (0,L) =
4
6968 AKIL ET AL.
o(1), we get
L L L ( )
i
h′2 zSb(·) (u, 𝜔)dx + ̃ (u, 𝜔)dx −
h2 zx Sb(·) h2 𝑦xx Sb(·)
̃ (u, 𝜔) dx
∫0 ∫0 𝜆 ∫0 x
𝛽−𝜀
(4.26)
ic ( ′ )
+ 0 h2 v + h2 vx S̄b̃ (u, 𝜔)dx = o(|𝜆|−𝓁 ).
𝜆 ∫𝛼 0
From (4.5), we have

zx = i𝜆𝑦x − 𝜆−𝓁 𝑓x3 .
̃ (u, 𝜔)) and (h2 ), we get
Using the above equation and the definition of (Sb(·)
L 𝛽−𝜀 ( )
̃ (u, 𝜔)dx =
h2 zx Sb(·) h2 i𝜆𝑦x − 𝜆−𝓁 𝑓x3 S̄b̃ (u, 𝜔)dx
∫0 ∫𝛼 0
𝛽−𝜀 𝛽−𝜀 ∞
= i𝜆b̃0 h2 𝑦x ux dx + i𝜆b0 h 2 𝑦x g(s)𝜔x (·, s)dsdx (4.27)
∫𝛼 ∫𝛼 ∫0
− b̃0 𝜆−𝓁 h2 𝑓x3 ux dx − 𝜆−𝓁 b0 h2 𝑓x3 g(s)𝜔x (·, s)dsdx.
∫𝛼 ∫𝛼 ∫0
From (4.7), we have
i𝜆𝜔x (·, s) = 𝜔xs (·, s) + i𝜆ux + 𝜆−𝓁 𝑓x1 − 𝜆−𝓁 𝑓x5 (·, s) in (0, 𝛽) × (0, ∞).
From the above equation and by using integration by parts with respect to s, we get
𝛽−𝜀 ∞ 𝛽−𝜀 ∞
i𝜆b0 h 2 𝑦x g(s)𝜔x (·, s)dsdx = b0 h 2 𝑦x −g′ (s)𝜔x (·, s)dsdx
∫𝛼 ∫0 ∫𝛼 ∫0
𝛽−𝜀 𝛽−𝜀 𝛽−𝜀 ∞
(4.28)
+ i𝜆b0 g̃ h2 𝑦x ux dx + b0 g̃ 𝜆 −𝓁
h2 𝑦x 𝑓x1 dx − b0 𝜆 −𝓁
h 2 𝑦x g(s)𝑓x5 (·, s)dsdx.
∫𝛼 ∫𝛼 ∫𝛼 ∫0
Inserting (4.28) in (4.27), then using the fact that b̃0 = a − b0 g̃ , we get
L 𝛽−𝜀 𝛽−𝜀 ∞
h2 zx S ̃̄(·)b(u, 𝜔)dx = i𝜆a h2 𝑦x ux dx + b0 h 2 𝑦x −g′ (s)𝜔x (·, s)dsdx
∫0 ∫𝛼 ∫𝛼 ∫0
+ b0 g̃ 𝜆−𝓁 h2 𝑦x 𝑓x1 dx − b0 𝜆−𝓁 h 2 𝑦x g(s)𝑓x5 (·, s)dsdx (4.29)
∫𝛼 ∫𝛼 ∫0
− b̃0 𝜆−𝓁 h2 𝑓x3 ux dx − 𝜆−𝓁 b0 h2 𝑓x3 g(s)𝜔x (·, s)dsdx.
∫𝛼 ∫𝛼 ∫0
Using Cauchy-Schwarz inequality, the facts that yx , ux is uniformly bounded in L2 (0, L), and estimation (4.8),
||F|| = o(1), we get
⎧ ( ) ( )
𝛽−𝜀 ∞ −𝓁 𝛽−𝜀
⎪ b0 ∫𝛼 h2 𝑦x ∫0 −g′ (s)𝜔x (·, s)dsdx = o |𝜆| 2 , b0 g̃ 𝜆−𝓁 ∫𝛼 h2 𝑦x 𝑓x1 = o |𝜆|−𝓁 ,
⎪ 𝛽−𝜀 ( ) 𝛽−𝜀
⎨ − b0 𝜆−𝓁 ∫𝛼 h2 𝑦x ∫0 g(s)𝑓x5 (·, s)dsdx = o (|𝜆|−𝓁 ,)−b̃0 𝜆−𝓁 ∫𝛼 h2 𝑓x3 ux dx = o(|𝜆|−𝓁 ) and
∞
⎪ 𝛽−𝜀 3 ∞ − 3𝓁
⎪ − 𝜆 b0 ∫𝛼 h2 𝑓x ∫0 g(s)𝜔x (·, s)dsdx = o |𝜆| 2 .
−𝓁
⎩
Inserting the above estimations in (4.29), we get
L 𝛽−𝜀 ( 𝓁
)
̃ (u, 𝜔)dx = i𝜆a
h2 zx Sb(·) h2 𝑦x ux dx + o |𝜆|− 2 . (4.30)
∫0 ∫𝛼
AKIL ET AL. 6969
From (4.3), we have

i𝜆−1 vx = −ux − i𝜆−(𝓁+1) 𝑓x1 .
̃ (u, 𝜔)) and (h2 ), we get
Then from the above equation and the definition of (Sb(·)
c0 𝛽−𝜀 𝛽−𝜀 𝛽−𝜀

i h2 vx S̄b̃ (u, 𝜔)dx = − ux S̄b̃ (u, 𝜔)dx − i𝜆−(𝓁+1) 𝑓x1 S̄b̃ (u, 𝜔)dx. (4.31)
𝜆 ∫𝛼 0 ∫𝛼 0 ∫𝛼 0
Using Cauchy–Schwarz inequality, the definition of (h2 ), the fact that ux is uniformly bounded in L2 (0, L) and ||𝑓x1 || =
o(1), and estimation (4.9), we get
𝛽−𝜀 ( 𝓁
) 𝛽−𝜀 ( )
ux S̄b̃ (u, 𝜔)dx = o |𝜆|− 2 and − i𝜆−(𝓁+1) 𝑓x1 S̄b̃ (u, 𝜔)dx = o |𝜆|− 2 −1
3𝓁
−
∫𝛼 0 ∫𝛼 0
c0 𝛽−𝜀 ( 𝓁
)
i h2 vx S̄b̃ (u, 𝜔)dx = o |𝜆|− 2 . (4.32)
𝜆 ∫𝛼 0
̃ (u, 𝜔)), (4.9), and the fact that v and z are uniformly bounded in L (0, L),
2
Now, using the definition of (h2 ) and (Sb(·)
we get
L 𝛽−𝜀 ( 𝓁
) c 𝛽−𝜀 ( )
h′2 zS ̃̄(·)b(u, 𝜔)dx = h′2 zS̄b̃ (u, 𝜔)dx = o |𝜆|− 2 andi 0 h′2 vS̄b̃ (u, 𝜔)dx = o |𝜆|− 2 .
3𝓁
(4.33)
∫0 ∫𝛼 0 𝜆 ∫𝛼 0
Inserting (4.30), (4.32), and (4.33) in (4.26), using the definition of (h2 ), then taking the real part, we get
{ 𝛽−𝜀 } { 𝛽−𝜀 ( ) } ( )
i ̄
𝓁
ℜ i𝜆a h2 𝑦x ux dx −ℜ h2 𝑦xx Sb̃ (u, 𝜔) dx = o |𝜆|− 2 . (4.34)
∫𝛼 𝜆 ∫𝛼 0 x
Now, adding (4.25) and (4.34) and using the fact that 𝓁 ≥ 0, we get
𝛽−𝜀 { }
i𝜆(a − 1) 𝛽−𝜀
h2 |𝑦x |2 dx = ℜ h2 ux 𝑦x dx + o(1).
∫𝛼 c0 ∫𝛼
Using the definition of (h2 ) in the above equation, we get the desired estimation (4.20). The proof is thus complete.
( )
of (4.3)–(4.7) satisfies the following estimation
𝛽−3𝜀
3|a − 1||𝜆| 𝛽−𝜀
|z|2 dx ≤ |ux ||𝑦x |dx + o(1). (4.35)
∫𝛼+2𝜀 |c0 | ∫𝛼
Proof. First, we fix a cut-off function (h3 ) ∈ C1 ([0, L]) (see Figure 2) such that 0 ≤ h3 (x) ≤ 1, for all x ∈ [0, L] and
{
0 if x ∈ [0, 𝛼 + 𝜀] ∪ [𝛽 − 2𝜀, L],
h3 (x) = 1 if x ∈ [𝛼 + 2𝜀, 𝛽 − 3𝜀], (h3 )
Multiplying (4.6) by −i𝜆−1 h3 z, using integration by parts over (0, L), the fact that z is uniformly bounded in L2 (0, L)
and ||f4 ||=o(1), and the definition of c(·), we get
L L L 𝛽−2𝜀 ( )
i i c
h3 |z|2 dx − h′3 z𝑦x dx − h3 zx 𝑦x dx + i 0 h3 vzdx = o |𝜆|−(𝓁+1) . (4.36)
∫0 𝜆 ∫0 𝜆 ∫0 𝜆 ∫𝛼+𝜀
6970 AKIL ET AL.
From (4.5), we have

i
− zx = −𝑦x + i𝜆−(𝓁+1) 𝑓x3 .
𝜆
Inserting the above equation in (4.36), we get
L L L
h3 |z|2 dx = h3 |𝑦x |2 dx − i𝜆−(𝓁+1) h3 𝑓x3 𝑦x dx
∫0 ∫0 ∫0
𝛽−2𝜀
(4.37)
i
L
c ( )
+ h′3 z𝑦x dx − i 0 h3 vzdx + o |𝜆|−(𝓁+1) .
𝜆 ∫0 𝜆 ∫𝛼+𝜀
Using the fact that ||𝑓x3 ||L2 (0,L) = o(1), yx and z are uniformly bounded in L2 (0, L), and the definition of (h3 ), we get
L ( ) i L 𝛽−2𝜀
c
−i𝜆−(𝓁+1) h3 𝑓x3 𝑦x dx = o |𝜆|−(𝓁+1) , h′3 z𝑦x dx = o(1) and − i 0 h3 vzdx = o(1). (4.38)
∫0 𝜆 ∫0 𝜆 ∫𝛼+𝜀
Using (4.20) and the definition of (h3 ), we get
L 𝛽−2𝜀 𝛽−𝜀
3|a − 1|
h3 |𝑦x |2 dx ≤ 3 |𝑦x |2 dx ≤ |𝜆| |ux ||𝑦x |dx + o(1). (4.39)
∫0 ∫𝛼+𝜀 |c0 | ∫𝛼
Inserting (4.38) and (4.39) in (4.37) and using the definition of (h3 ), we get the desired estimation (4.35). The proof
has been completed.
Now, we fix a function 𝜒 ∈ C1 ([𝛽 − 3𝜀, 𝛾]) by
𝜒(𝛽 −3𝜀) = −𝜒(𝛾) = 1, and set max |𝜒(x)| = M𝜒 and max |𝜒 ′ (x)| = M𝜒 ′ . (𝜒)
x∈[𝛽−3𝜀,𝛾] x∈[𝛽−3𝜀,𝛾]
Remark 4.1. It is easy to see the existence of 𝜒(x). For example, we can take
1 ( )
𝜒(x) = −2x2 + 4(𝛽 − 3𝜀)x + 𝛾 2 − (𝛽 − 3𝜀)2 − 2𝛾(𝛽 − 3𝜀) ,
(𝛾 − 𝛽 + 3𝜀) 2
4
to get 𝜒(𝛽 − 3𝜀) = −𝜒(𝛾) = 1, 𝜒 ∈ C1 ([𝛽 − 3𝜀, 𝛾]), M𝜒 = 1 and M𝜒 ′ = 𝛾−𝛽+3𝜀
.
( )
of (4.3)–(4.7) satisfies the following estimations
|v(𝛾)|2 + |v(𝛽 − 3𝜀)|2 = O(|𝜆|) and |z(𝛾)|2 + |z(𝛽 − 3𝜀)|2 + |𝑦x (𝛾)|2 + |𝑦x (𝛽 − 3𝜀)|2 = O(1). (4.40)
Proof. First, deriving Equation (4.3) with respect to x, we obtain
i𝜆ux − vx = 𝜆−𝓁 𝑓x1 in (𝛽 − 3𝜀, 𝛾).
Multiplying the above equation by 2𝜒v, integrating over (𝛽 − 3𝜀, 𝛾), then taking the real part, we obtain
{ 𝛾 } 𝛾 { 𝛾 }
ℜ 2i𝜆 𝜒ux vdx − 𝜒(|v| )x dx = ℜ
2
2𝜆 −𝓁
𝜒𝑓x1 vdx . (4.41)
∫𝛽−3𝜀 ∫𝛽−3𝜀 ∫𝛽−3𝜀
Using integration by parts in (4.41), we obtain
𝛾 { 𝛾 } { 𝛾 }
[ ]𝛾
−𝜒|v|2 𝛽−3𝜀 = − 𝜒 |v| dx − ℜ
′ 2
2i𝜆 𝜒ux vdx +ℜ 2𝜆 −𝓁
𝜒𝑓x1 vdx . (4.42)
∫𝛽−3𝜀 ∫𝛽−3𝜀 ∫𝛽−3𝜀
AKIL ET AL. 6971
Using the definition of (𝜒) and Cauchy–Schwarz inequality in (4.42), we obtain
𝛾 ( 𝛾 )1 ( 𝛾 )1
2 2
|v(𝛾)|2 + |v(𝛽 − 3𝜀)|2 ≤ M𝜒 ′ |v|2 dx + 2|𝜆|M𝜒 |ux |2 dx |v|2 dx
∫𝛽−3𝜀 ∫𝛽−3𝜀 ∫𝛽−3𝜀
( )1 ( 𝛾 )1 (4.43)
𝛾 2 2
−𝓁
+ 2|𝜆| M𝜒 |𝑓x1 |2 dx |v|2 dx .
∫𝛽−3𝜀 ∫𝛽−3𝜀
Thus, from (4.43) and the fact that ux , v are uniformly bounded in L2 (0, L) and ||𝑓x1 ||L2 (0,L) = o(1), we obtain the first
estimation in (4.40). From (4.5) and (4.6) and the definition of c(·), we have
i𝜆𝑦x − zx = 𝜆−𝓁 𝑓x3 in (𝛽 − 3𝜀, 𝛾),
and
i𝜆z − 𝑦xx − c0 v = 𝜆−𝓁 𝑓 4 in (𝛽 − 3𝜀, 𝛾).
Multiplying the above equations by 2𝜒z and 2𝜒𝑦x respectively, integrating over (𝛽 − 3𝜀, 𝛾), taking the real part, then
using the fact that yx , z are uniformly bounded in L2 (0, L) and ||𝑓 2 ||L2 (0,L) = o(1) and ||𝑓x3 ||L2 (0,L) = o(1), we obtain
{ 𝛾 } 𝛾
ℜ 2i𝜆 𝜒𝑦x zdx − 𝜒(|z|2 )x dx = o(|𝜆|−𝓁 ) (4.44)
∫𝛽−3𝜀 ∫𝛽−3𝜀
and
{ 𝛾 } 𝛾 { 𝛾 }
ℜ 2i𝜆 𝜒z𝑦x dx − 𝜒(|𝑦x | )x dx − ℜ
2
2c0 𝜒v𝑦x dx = o(|𝜆|−𝓁 ). (4.45)
∫𝛽−3𝜀 ∫𝛽−3𝜀 ∫𝛽−3𝜀
Adding (4.44) and (4.45), then using integration by parts, we obtain
𝛾 { 𝛾 }
[ ]𝛾
−𝜒(|z|2 + |𝑦x |2 ) 𝛽−3𝜀 = − 𝜒 (|z| + |𝑦x | )dx + ℜ
′ 2 2
2c0 𝜒v𝑦x dx + o(|𝜆|−𝓁 ).
∫𝛽−3𝜀 ∫𝛽−3𝜀
Using the definition of (𝜒) and Cauchy–Schwarz inequality in the above equation, we obtain
|z(𝛾)|2 + |z(𝛽 − 3𝜀)|2 + |𝑦x (𝛾)|2 + |𝑦x (𝛽 − 3𝜀)|2

𝛾 ( 𝛾 )1 ( 𝛾 )1 (4.46)
2 2
≤ M𝜒 ′ (|z| + |𝑦x | )dx + 2c0 M𝜒
2 2
|v| dx
2
|𝑦x | dx
2
+ o(|𝜆|−𝓁 ).
∫𝛽−3𝜀 ∫𝛽−3𝜀 ∫𝛽−3𝜀
Finally, from (4.46) and the fact that v, yx , z are uniformly bounded in L2 (0, L), we obtain the second estimation
in (4.40). The proof is thus complete.
Lemma 4.6. Let 𝜃 ∈ C1 ([0, L]) be a function with 𝜃(0) = 𝜃(L) = 0. Under the hypotheses (H), the solution U =
(u, v, 𝑦, z, 𝜔(·, s))⊤ ∈ D() of (4.3)–(4.7) satisfies the following estimation
L( ) { L }
−1 | |2
𝜃 |v| + a |Sb(·)
′ 2
̃ (u, 𝜔)| + |z| + |𝑦x |
2 2
dx + ℜ 2a −1
̃ (u, 𝜔)dx
c(·)𝜃zSb(·)
∫0 | | ∫0
{ L } ( ) (4.47)
𝓁
−ℜ 2 c(·)𝜃v𝑦x dx = o |𝜆|− 2 .
∫0
Proof. First, from (4.3), we deduce that

i𝜆ux = −vx − 𝜆−𝓁 𝑓x1 . (4.48)
6972 AKIL ET AL.
Multiplying (4.4) by 2a−1 𝜃 S ̃̄(·)b(u, 𝜔), integrating over (0, L), taking the real part, then using (4.9) and the fact that
||𝑓 2 ||L2 (0,L) = o(1), we get
{ L } L ( )
ℜ | |2
2i𝜆a −1
𝜃vSb(·)
̃ (u, 𝜔)dx − a−1 𝜃 |Sb(·)
̃ (u, 𝜔)| dx
∫0 ∫0 | | x
{ L } (4.49)
+ℜ −1
2a ̃ (u, 𝜔)dx
c(·)𝜃zSb(·) = o(|𝜆| ). −𝓁
∫0
̃ (u, 𝜔)) and the fact that b̃0 = a − b0 g

Using the definition of (Sb(·) ̃ , we obtain
{ L } { L } { 𝛽 }
ℜ 2i𝜆a −1
𝜃vSb(·)
̃ (u, 𝜔)dx =ℜ 2i𝜆 𝜃vux dx −ℜ 2i𝜆a b0 g̃
−1
𝜃vux dx
∫0 ∫0 ∫0
{ 𝛽 ∞ }
+ℜ 2i𝜆a−1 b0 𝜃v g(s)𝜔x (·, s)dsdx .
∫0 ∫0
Inserting (4.48) in the above equation and using the fact that v is uniformly bounded in L2 (0, L), ||𝑓x1 ||L2 (0,L) = o(1), we
get
{ } L ( { 𝛽 }
L )
ℜ 2i𝜆a −1
𝜃vSb(·)
̃ (u, 𝜔)dx =− 𝜃 |v| x dx + ℜ 2a b0 g̃
2 −1
𝜃vvx dx
∫0 ∫0 ∫0
{ 𝛽 ∞ }
+ ℜ 2i𝜆a−1 b0 𝜃v g(s)𝜔x (·, s)dsdx + o(|𝜆|−𝓁 ).
∫0 ∫0
Now, inserting the above equation in (4.49), we obtain
(
L ) { L }
| |2
− 𝜃 |v|2 + a−1 |Sb(·)
̃ (u, 𝜔)| dx + ℜ 2a−1 ̃ (u, 𝜔)dx
c(·)𝜃zSb(·)
∫0 | | x ∫0
{ 𝛽 } { 𝛽 } (4.50)
∞ ( )
= − ℜ 2a b0 g̃
−1
𝜃vvx dx − ℜ 2i𝜆a b0−1
𝜃v g(s)𝜔x (·, s)dsdx + o |𝜆|−𝓁 .
∫0 ∫0 ∫0
i𝜆𝜔x (·, s) = 𝜔xs (·, s) − vx − 𝜆−𝓁 𝑓x5 (·, s) in (0, 𝛽) × (0, ∞). (4.51)
Inserting (4.51) in the right hand side of (4.50), then using integration by parts with respect to s with the help of
hypotheses (H) and the fact that 𝜔(·, 0) = 0, we get
L( ) { L }
| |2
− 𝜃 |v|2 + a−1 |Sb(·)
̃ (u, 𝜔)| dx + ℜ 2a−1 ̃ (u, 𝜔)dx
c(·)𝜃zSb(·)
∫0 | | x ∫0
{ 𝛽 } { 𝛽 } (4.52)
∞ ∞ ( )
= − ℜ 2a b0
−1
𝜃v −g (s)𝜔x (·, s)dsdx − ℜ 2a 𝜆 b0
′ −1 −𝓁
𝜃v g(s)𝑓x (·, s)dsdx + o |𝜆|−𝓁 .
5
∫0 ∫0 ∫0 ∫0
Using Cauchy–Schwarz inequality, the fact that v is uniformly bounded in L2 (0, L), the definition of g and (4.8), we
obtain
⎧ { } ( 𝓁
)
ℜ −1 b ∫ 𝛽 𝜃v ∫ ∞ −g′ (s)𝜔 (·, s)dsdx = o |𝜆|− 2 ,
⎪ − 2a 0 x
⎨ { −1 −𝓁 }
0 0
𝛽 ∞ ( −𝓁 ) (4.53)
⎪ ℜ 2a 𝜆 b0 ∫0 𝜃v ∫0 g(s)𝑓x (·, s)dsdx = o |𝜆| .
5
⎩
Inserting (4.53) in (4.52), then using integration by parts and the fact that 𝜃(0) = 𝜃(L) = 0, we obtain
L ( ) { L } ( )
| |2 𝓁
𝜃 ′ |v|2 + a−1 |Sb(·)
̃ (u, 𝜔)| dx + ℜ 2a−1 ̃ (u, 𝜔)dx
c(·)𝜃zSb(·) = o |𝜆|− 2 . (4.54)
∫0 | | ∫0
AKIL ET AL. 6973
FIGURE 3 Geometric description of the functions h4 and h5

[Colour figure can be viewed at wileyonlinelibrary.com]
Next, multiplying (4.6) by 2h𝑦x , integrating over (0, L), taking the real part, then using the fact that yx is uniformly
bounded in L2 (0, L) and ||𝑓 4 ||L2 (0,L) = o(1), we obtain
{ L } L { L }
ℜ 2i𝜆 𝜃z𝑦x dx − 𝜃(|𝑦x |2 )x dx − ℜ 2 c(·)𝜃v𝑦x dx = o(|𝜆|−𝓁 ). (4.55)
∫0 ∫0 ∫0
i𝜆𝑦x = −zx − 𝜆−𝓁 𝑓x3 . (4.56)
Inserting (4.56) in (4.55), then using the fact that z is uniformly bounded in L2 (0, L) and ||𝑓x3 ||L2 (0,L) = o(1), we obtain
L { L }
− 𝜃(|z|2 + |𝑦x |2 )x dx − ℜ 2 c(·)𝜃v𝑦x dx = o(|𝜆|−𝓁 ). (4.57)
∫0 ∫0
Using integration by parts in (4.57) and the fact that 𝜃(0) = 𝜃(L) = 0, we obtain
L { L }
𝜃 (|z| + |𝑦x | )x dx − ℜ
′ 2 2
2 c(·)𝜃v𝑦x dx = o(|𝜆|−𝓁 ). (4.58)
∫0 ∫0
Finally, adding (4.54) and (4.58), we obtain the desired estimation (4.47). The proof is thus complete.
( )
Let 0 < 𝜀 < min 𝛼, 𝛽−𝛼
5
, we fix cut-off functions h4 , h5 ∈ C1 ([0, L]) (see Figure 3) such that 0 ≤ h4 (x) ≤ 1, 0 ≤ h5 (x) ≤ 1,
for all x ∈ [0, L] and
{ {
1 if x ∈ [0, 𝛼 + 2𝜀], 0 if x ∈ [0, 𝛼 + 2𝜀],
h4 (x) = 0 if x ∈ [𝛽 − 3𝜀, L], and h5 (x) = 1 if x ∈ [𝛽 − 3𝜀, L],
and set maxx∈[0,L] |h′4 (x)| = Mh′4 and maxx∈[0,L] |h′5 (x)| = Mh′5 .
( )
. Under the hypotheses (H), the solution U = (u, v, 𝑦, z, 𝜔(·, s))⊤ ∈ D() of
System (4.3)–(4.7) satisfies the following estimations
𝛼+2𝜀 ( ) 𝛽−𝜀
|v|2 + |𝑦x |2 + |z|2 dx ≤ K1 |a − 1||𝜆| |ux ||𝑦x |dx + o(1), (4.59)
∫0 ∫𝛼
L L ( ) 𝛽−𝜀
a |ux |2 dx + |v|2 + |𝑦x |2 + |z|2 dx ≤ K2 |a − 1||𝜆| |ux ||𝑦x |dx + o(1), (4.60)
∫𝛽 ∫𝛽−3𝜀 ∫𝛼
4 ( ) 4
where K1 = |c0 |
1 + (𝛽 − 3𝜀) Mh′4 and K2 = |c0 |
(1 + (L − 𝛼 + 2𝜀)Mh′5 ).
6974 AKIL ET AL.
Proof. First, using the result of Lemma 4.6 with 𝜃 = xh4 and the definition of (Sb(·)
̃ (u, 𝜔)) and c(·), we obtain
𝛼+2𝜀 ( 2 ) 𝛼+2𝜀
| |2
|v| + |𝑦x |2 + |z|2 dx = − a−1 |Sb̃0 (u, 𝜔)| dx
∫0 ∫0 | |
𝛽−3𝜀 ( ) ( )
| | 2
− h4 + xh′4 |v|2 + a−1 |Sb̃ (u, 𝜔)| + |𝑦x |2 + |z|2 dx
∫𝛼+2𝜀 | 0 |
{ 𝛽−3𝜀 } { 𝛽−3𝜀 } ( )
𝓁
− ℜ 2a c0−1 ̄
xh4 zSb̃ (u, 𝜔)dx + ℜ 2c0 xh4 v𝑦x dx + o |𝜆|− 2 .
∫𝛼 0 ∫𝛼
Using Cauchy–Schwarz inequality in the above equation, we obtain
𝛼+2𝜀 ( 2 ) 𝛼+2𝜀
| |2
|v| + |𝑦x |2 + |z|2 dx ≤ a−1 |Sb̃0 (u, 𝜔)| dx
∫0 ∫0 | |
( ( ) 𝛽−3𝜀 )
| |2
+ 1 + (𝛽 − 3𝜀)Mh′4 |v|2 + a−1 |Sb̃ (u, 𝜔)| + |𝑦x |2 + |z|2 dx
∫𝛼+2𝜀 | 0 |
( 𝛽−3𝜀 ) 12 ( 𝛽−3𝜀 ) 12
+ 2c0 (𝛽 − 3𝜀)a−1 |z|2 dx |Sb̃ (u, 𝜔)|2 dx
∫𝛼 ∫𝛼 0
( 𝛽−3𝜀 ) 12 ( 𝛽−3𝜀 ) 12
+ 2c0 (𝛽 − 3𝜀) |v| dx
2
|𝑦x | dx .
2
∫𝛼 ∫𝛼
Thus, from the above inequality, Lemmas 4.1–4.4 and the fact that yx , z are uniformly bounded in L2 (0, L), we
obtain (4.59). Next, using the result of Lemma 4.6 with 𝜃 = (x − L)h5 and the definition of (Sb(·) ̃ (u, 𝜔)) and c(·), we
obtain
L L ( ) 𝛽
| |2
a |ux |2 dx + |v|2 + |z|2 + |𝑦x |2 dx = −a−1 |Sb̃0 (u, 𝜔)| dx
∫𝛽 ∫𝛽−3𝜀 ∫𝛽−3𝜀 | |
𝛽−3𝜀 ( ) ( )
| |2
− h5 + (x − L)h′5 |v|2 + a−1 |Sb̃ (u, 𝜔)| + |𝑦x |2 + |z|2 dx
∫𝛼+2𝜀 | 0 |
{ 𝛽−3𝜀 } { 𝛽−3𝜀 }
− ℜ 2a−1 c0 (x − L)h5 zS̄b̃ (u, 𝜔)dx + ℜ 2c0 (x − L)h5 v𝑦x dx
∫𝛼+2𝜀 0 ∫𝛼+2𝜀
{ 𝛽 ( ∞ ) }
− ℜ 2a b0 c0−1
(x − L)z −̃gux + g(s)𝜔x (·, s)ds dx
∫𝛽−3𝜀 ∫0
{ 𝛾 } { 𝛾 }
− ℜ 2c0 (x − L)zux dx + ℜ 2c0 (x − L)v𝑦x dx .
∫𝛽−3𝜀 ∫𝛽−3𝜀
Using Cauchy–Schwarz inequality in the above equation, Lemmas 4.1–4.4 and the fact that yx , z are uniformly
bounded in L2 (0, L), we obtain
L L ( )
a |ux |2 dx + |v|2 + |z|2 + |𝑦x |2 dx
∫𝛽 ∫𝛽−3𝜀
𝛽−𝜀
(4.61)
4 ( )
≤ 1 + (L − 𝛼 − 2𝜀)Mh′5 |a − 1||𝜆| |ux ||𝑦x |dx +  + o(1),
|c0 | ∫𝛼
where
{ 𝛾 } { 𝛾 }
 ∶= ℜ 2c0 (x − L)v𝑦x dx −ℜ 2c0 (x − L)zux dx . (4.62)
∫𝛽−3𝜀 ∫𝛽−3𝜀
From (4.3) and (4.5), we have
ux = i𝜆−1 vx + i𝜆−(𝓁+1) 𝑓x1 and 𝑦x = i𝜆−1 zx + i𝜆−(𝓁+1) 𝑓x3 . (4.63)

AKIL ET AL. 6975
Inserting (4.63) in (4.62), then using the fact that v and z are uniformly bounded in L2 (0, L) and ||𝑓x1 ||L2 (0,L) = o(1),
||𝑓x3 ||L2 (0,L) = o(1), we obtain
{ 𝛾 } { 𝛾 }
=ℜ 2c0 i𝜆 −1
(x − L)vzx dx −ℜ 2c0 i𝜆 −1
(x − L)zvx dx + o(|𝜆|−(𝓁+1) ). (4.64)
∫𝛽−3𝜀 ∫𝛽−3𝜀
Using integration by parts to the second term in (4.64), we obtain

{ } {
𝛾 [ ]𝛾 }
=ℜ 2c0 i𝜆 −1
zvdx − ℜ 2c0 i𝜆−1 (x − L)zv 𝛽−3𝜀 + o(|𝜆|−(𝓁+1) ). (4.65)
∫𝛽−3𝜀
From Lemma 4.5, we deduce that

(√ ) (√ )
|v(𝛾)| = O |𝜆| , |v(𝛽 − 3𝜀)| = O |𝜆| , |z(𝛾)| = O(1) and |z(𝛽 − 3𝜀)| = O(1). (4.66)
Using Cauchy–Schwarz inequality, (4.66) and the fact that v, z are uniformly bounded in L2 (0, L), we obtain
{ } {
𝛾 ( ) [ ]𝛾 } ( 1
)
ℜ 2c0 i𝜆−1 zvdx = O |𝜆|−1 = o(1) and − ℜ 2c0 i𝜆−1 (x − L)zv 𝛽−3𝜀 = O |𝜆|− 2 = o(1).
∫𝛽−3𝜀

 = o(1).
Finally, from the above estimation and (4.61), we obtain the desired estimation (4.60). The proof is thus complete.
Proof of Theorem 4.1. The proof of Theorem 4.1 is divided into three steps.
Step 1. Under the hypotheses (H), by taking a = 1 and 𝓁 = 0 in Lemmas 4.1–4.4, we obtain
{
𝛽 ∞ 𝛽 𝛽−𝜀
∫0 ∫0 g(s)|𝜔x (·, s)|2 dsdx = o(1), ∫0 |ux |2 dx = o(1), ∫𝜀 |v|2 dx = o(1),
𝛽−2𝜀 𝛽−3𝜀 (4.67)
∫𝛼+𝜀 |𝑦x |2 dx = o(1) and ∫𝛼+2𝜀 |z|2 dx = o(1).
Step 2. Using the fact that a = 1 and (4.67) in Lemma 4.7, we obtain
{ 𝜀 L L 𝛼+𝜀
∫0 |v|2 dx = o(1), ∫𝛽−𝜀 |v|2 dx = o(1), ∫𝛽 |ux |2 dx = o(1), ∫0 |𝑦x |2 dx = o(1),
L 𝛼+2𝜀 L (4.68)
∫𝛽−2𝜀 |𝑦x | dx = o(1),
2 ∫0 |z| dx = o(1) and
2 ∫𝛽−3𝜀 |z| dx = o(1).
2
Step 3. According to Step 1 and Step 2, we obtain ||U|| = o(1), which contradicts (H1 ). Therefore, (H1 ) holds, and
so by Theorem A3, we deduce that system (2.2)–(2.6) is exponentially stable.
Proof of Theorem 4.2. The proof of Theorem 4.2 is divided into three steps.
Step 1. Under the hypotheses (H) and a ≠ 1, using the fact that yx is uniformly bounded in L2 (0, L) and (4.9) in
estimation (4.20), we get
𝛽−2𝜀 ( 𝓁
) 𝛽−3𝜀 ( 𝓁
)
|𝑦x |2 dx = o |𝜆|− 2 +1 and |z|2 dx = o |𝜆|− 2 +1 .
∫𝛼+𝜀 ∫𝛼+2𝜀
Taking 𝓁 = 2 in the above estimations, we obtain
𝛽−2𝜀 𝛽−3𝜀
|𝑦x |2 dx = o(1) and |z|2 dx = o(1). (4.69)
∫𝛼+𝜀 ∫𝛼+2𝜀
6976 AKIL ET AL.
Taking 𝓁 = 2 in Lemmas 4.1 and 4.2, we obtain
𝛽 ∞ 𝛽 𝛽−𝜀
g(s)|𝜔x (·, s)|2 dsdx = o(𝜆−2 ), |ux |2 dx = o(𝜆−2 ) and |v|2 dx = o(|𝜆|−1 ). (4.70)
∫0 ∫0 ∫0 ∫𝜀
Step 2. Using the fact that a ≠ 1, yx is uniformly bounded in L2 (0, L) and (4.70) in Lemma 4.7, we obtain
𝛼+2𝜀 ( 2 )
|v| + |𝑦x |2 + |z|2 dx = o(1), (4.71)
∫0
L L ( )
a |ux |2 dx + |v|2 + |𝑦x |2 + |z|2 dx = o(1). (4.72)
∫𝛽 ∫𝛽−3𝜀
Using (4.69) and (4.70) in (4.71) and (4.72), we obtain
{ 𝜀 L L 𝛼+𝜀
∫0 |v|2 dx = o(1), ∫𝛽−𝜀 |v|2 dx = o(1), ∫𝛽 |ux |2 dx = o(1), ∫0 |𝑦x |2 dx = o(1),
L 𝛼+2𝜀 L (4.73)
∫𝛽−2𝜀 |𝑦x | dx = o(1),
2 ∫0 |z| dx = o(1) and
2 ∫𝛽−3𝜀 |z| dx = o(1).
2
Step 3. According to Step 1 and Step 2, we obtain ||U|| = o(1), which contradicts (H1 ). This implies that
( )
sup ||(i𝜆I − )−1 || = O 𝜆2 .
𝜆∈R
Finally, according to Theorem A4, we obtain the desired result. The proof is thus complete.
5 LACK OF EXPONENTIAL STABILITY WITH G LO BAL PAST H ISTO RY

DA MPING IN CASE OF DIFFERENT SPEED PROPAGATION WAVES (aeq1)
This section is independent from the previous ones; here we prove the lack of exponential stability with global past history
damping and global coupling. For this aim, we consider the following system:
⎧ u − au + ∫ ∞ g(s)u (x, t − s)ds + 𝑦 = 0, (x, t) ∈ (0, L) × (0, ∞),

⎪ tt xx 0 xx t
⎪ 𝑦tt − 𝑦xx − ut = 0, (x, t) ∈ (0, L) × (0, ∞),
⎨ u(0, t) = u(L, t) = 𝑦(0, t) = 𝑦(L, t) = 0, t > 0, (5.1)
⎪ (u(x, −s), ut (x, 0)) = (u0 (x, s), u1 (x)), (x, s) ∈ (0, L) × (0, ∞),
⎪ (𝑦(x, 0), 𝑦t (x, 0)) = (𝑦0 (x), 𝑦1 (x)), x ∈ (0, L),
⎩
and the general integral term represent a history term with the relaxation function g that is supposed to satisfy the
following hypotheses
{
g ∈ L1 ([0, ∞)) ∩ C1 ([0, ∞)) be a strictly positive function such that
∞ (HG )
g(0) ∶= g0 > 0, ∫0 g(s)ds ∶= g̃ , ã ∶= a − g̃ > 0, and g′ (s) ≤ −mg(s), for somem > 0.
Now, as in Dafermos,24 we introduce the following auxiliary change of variable
𝜔(x, s, t) ∶= u(x, t) − u(x, t − s), (x, s, t) ∈ (0, L) × (0, ∞) × (0, ∞). (5.2)
Then system (5.1) becomes

∞
̃ xx −
utt − au g(s)𝜔xx (·, s, t) + 𝑦t = 0, (x, t) ∈ (0, L) × (0, ∞), (5.3)
∫0
𝑦tt − 𝑦xx − ut = 0, (x, t) ∈ (0, L) × (0, ∞), (5.4)

AKIL ET AL. 6977
𝜔t (·, s, t) + 𝜔s (·, s, t) − ut = 0, (x, s, t) ∈ (0, L) × (0, ∞) × (0, ∞), (5.5)

{
u(0, t) = u(L, t) = 𝑦(0, t) = 𝑦(L, t) = 0, t > 0,
𝜔(·, 0, t) = 0, (x, t) ∈ (0, L) × (0, ∞), (5.6)
𝜔(0, s, t) = 𝜔(L, s, t) = 0, (s, t) ∈ (0, ∞) × (0, ∞),
and the following initial conditions

{
u(·, −s) = u0 (·, s), ut (·, 0) = u1 (·), (x, s) ∈ (0, L) × (0, ∞),
𝑦(·, 0) = 𝑦0 (·), 𝑦t (·, 0) = 𝑦1 (·), x ∈ (0, L), (5.7)
𝜔(·, s, 0) = u0 (·, 0) − u0 (·, s), (x, s) ∈ (0, L) × (0, ∞).
The energy of system (5.3)–(5.7) is given by
L( ) L ∞
1 1
EG (t) = ̃ x |2 + |𝑦t |2 + |𝑦x |2 dx +
|ut |2 + a|u g(s)|𝜔x (·, s, t)|2 dsdx. (5.8)
2 ∫0 2 ∫0 ∫0
Under the hypotheses (HG ) and by letting U = (u, v, 𝑦, z, 𝜔) be a regular solution of system (5.3)–(5.7), then we get with
the help of (5.6) that
L ∞
d 1
EG (t) = g′ (s)|𝜔x (·, s, t)|2 dsdx ≤ 0,
dt 2 ∫0 ∫0
which implies that the system (5.3)–(5.7) is dissipative in the sense that its energy is nonincreasing with respect to time.
Now, we define the following Hilbert space G by
( )2
G = H01 (0, L) × L2 (0, L) × L2g ((0, ∞); H01 (0, L)),
and it is equipped with the following inner product
( ) L ( ) L ∞
U, U 1  = ̃ x u1x + vv1 + 𝑦x 𝑦1x + zz1 dx +
au g(s)𝜔x (·, s)𝜔1x (·, s)dsdx,
G ∫0 ∫0 ∫0
where U = (u, v, 𝑦, z, 𝜔(·, s))⊤ ∈ G and U 1 = (u1 , v1 , 𝑦1 , z1 , 𝜔1 (·, s))⊤ ∈ G . We define the linear unbounded operator
G ∶ D(G ) ⊂ G → G by
{ ⊤
}
( U = (u, v, 𝑦, z, 𝜔(·, s))
) ∈ G | 𝑦 ∈ H (0, L) ∩ H0 (0, L), v, z ∈ H0 (0, L)
2 1 1
D(G ) = ∞ ,
̃aux + ∫0 g(s)𝜔x (·, s)ds x ∈ L2 (0, L), 𝜔s (·, s) ∈ L2g ((0, ∞); H01 (0, L)), 𝜔(·, 0) = 0.
and
⎛ u ⎞ ⎛( ∞
v ) ⎞
⎜ v ⎟ ⎜ au ̃ x + ∫0 g(s)𝜔x (·, s)ds x − z ⎟
G ⎜ 𝑦 ⎟ = ⎜ z ⎟.
⎜ z ⎟ ⎜ 𝑦 + v ⎟
⎜ 𝜔(·, s) ⎟ ⎜ xx ⎟
⎝ ⎠ ⎝ −𝜔s (·, s) + v ⎠
Now, if U = (u, ut , 𝑦, 𝑦t , 𝜔(·, s))⊤ , then system (5.3)–(5.7) can be written as the following first order evolution equation
Ut = G U, U(0) = U0 , (5.9)
where U0 = (u0 (·, 0), u1 , 𝑦0 , 𝑦1 , 𝜔0 (·, s))⊤ ∈ G .

2
Theorem 5.1. Under the hypotheses (HG ). If a ≠ 1, then for any 0 < 𝜖 < 2, we can not expect the energy decay rate t− 2−𝜖
for every U0 ∈ D(G ).
6978 AKIL ET AL.
Proof. Following Huang29 and Pruss30 (see also Theorem A3), it is sufficient to show the existence of sequences
(𝜆n )n ⊂ R∗+ with 𝜆n → ∞, (Un )n ⊂ D(G ) and (Fn )n ⊂ G such that (i𝜆n I − ) Un = Fn is bounded in G and
lim 𝜆−2+𝜖
n ||Un ||G = ∞. (5.10)
n→∞
For this aim, take ( ( ) )

n𝜋x
Fn = 0, 0, 0, sin , 0 and Un = (un , i𝜆n un , 𝑦n , i𝜆n 𝑦n , 𝜔n )
L
such that
⎧ 𝜆n = n𝜋 − L 2
such that n2 > 2𝜋 2L(a−1) ,
⎪ L 2n𝜋(a−1)
⎨ ( ) ( ) ( ) (5.11)
⎪ un (x) = An sin n𝜋x , 𝑦n (x) = Bn sin n𝜋x , 𝜔n (x, s) = An (1 − e−i𝜆n s ) sin n𝜋x ,
⎩ L L L
where An and Bn are complex numbers depending on n and determined explicitly in the sequel. Note that this choice
is compatible with the boundary conditions. So, its is clear that 𝜆n > 0, limn→∞ 𝜆n = ∞, Fn is uniformly bounded in 
and Un ∈ D(G ). Next, detailing i𝜆n Un − G Un = Fn , we get
{ ( )
iAn L2 𝜆 + 𝜆2 L2 − 𝜋 2 n2 Bn = −L2 ,
( 2 2 ) (5.12)
n 𝜋 (a − g𝜆n ) − 𝜆2 L2 An + iL2 𝜆Bn = 0,
∞
where g𝜆n = ∫0 g(s)e−i𝜆n s ds. From the first equation of (5.12), we get
i i(L2 𝜆2 − 𝜋 2 n2 )Bn
An = + . (5.13)
𝜆 L2 𝜆
Inserting (5.13) in the second equation of (5.12), we get

( 2 2 )
𝜆 L − (a − g𝜆n )n2 𝜋 2 L2
Bn = .
−n4 (a − g𝜆n )𝜋 4 + L2 𝜋 2 n2 𝜆2 (a + 1 − g𝜆n ) + L4 (𝜆2 − 𝜆4 )
Consequently, the solution of (5.12) is given by

( )
i i(L2 𝜆2 − 𝜋 2 n2 )Bn B2,n
An = + and Bn = B1,n 1 + , (5.14)
𝜆 L2 𝜆 𝜆n g𝜆n + B3,n
where
⎧ B1,n = 2 2L 2 2 , B2,n = 2 2 L2 𝜆2 2 2 ,
2 4 3
⎪ (n 𝜋 −L 𝜆 ) n 𝜋 (𝜆 L −n 𝜋 )
⎨
⎪ B3,n = (−𝜋 an +L 2n 2𝜆 (a+1)𝜋
2 +L4 (𝜆2 −𝜆4 ) 𝜆
4 4 2 2 2
)
.
⎩ n 𝜋 (n2 𝜋 2 −L2 𝜆2 )
Now, inserting 𝜆n given in (5.11) in the above equation, then using asymptotic expansion, we get
1−a
B1,n = a − 1 + O(n−2 ), B2,n = 𝜋n + O(n−1 ), B3,n = O(n−1 ). (5.15)
L
On the other hand, using hypotheses (HG ) and integration by parts, we obtain
∞
𝜆n g𝜆n = − ig0 − i g′ (s)e−i𝜆n s ds.
∫0
It is clear from Riemann–Lebesgue lemma that the second term in the above equation goes to zero as 𝜆n → ∞. Thus,
we obtain
𝜆n g𝜆n = −ig0 + o(1). (5.16)
AKIL ET AL. 6979
Substituting (5.15) and (5.16) in (5.14), we get

( )
i(a − 1)2
An = O(1) and Bn = − + o(1) n𝜋.
g0 L
Therefore, from the above equation and (5.16), we get
( ) ( ) ( )
n𝜋x (a − 1)2 n𝜋x
zn (x) = i𝜆n Bn sin = + o(1) n 𝜋
2 2
sin .
L g0 L2 L
Consequently,
( L ) 12 √ ( )
L (a − 1)2
|zn | dx
2
∼ + o(1) n2 𝜋 2 .
∫0 2 g0 L2
Since
( L ) 12 √ ( )
L (a − 1)2
||Un || ≥ |zn | dx
2
∼ + o(1) n2 𝜋 2 ∼ 𝜆2n ,
∫0 2 g0 L2
then for all 0 < 𝜖 < 2, we have
𝜆−2+𝜖
n ||Un ||1 ∼ 𝜆𝜖n → ∞ as n → ∞,
2
hence, we get (5.10). Consequently, we cannot expect the energy decay rate t− 2−𝜖 . The proof is thus complete.
Remark 5.1. In Almeida et al.22 and Cordeiro,23 the authors proved the lack of exponential stability of a coupled wave
equations system with past history damping by taking a particular relaxation function g(s) = e−𝜇s such that s ∈ R+
and 𝜇 > 1.
6 CO N C LUSION AN D FU T U R E WORKS
We have studied the stabilization of a locally coupled wave equations with local viscoelastic damping of past history type
acting only in one equation via non smooth coefficients. We proved the strong stability of the system by using Arendt–Batty
criteria. We established the exponential stability of the solution if and only if the waves have the same speed propagation
(i.e., a = 1). In the case a ≠ 1, we proved that the energy of our system decays polynomially with the rate t−1 . Lack of
exponential stability result has been proved in case that the speed of waves propagation are different (i.e. a ≠ 1). According
to Theorem 5.1, we can conjecture that the energy decay rate t−1 is optimal but this question remains open. Moreover, it
would be interesting to
(1) study system (1.1) in the multidimensional case,
(2) obtain the decay rates of system (1.1) for a much wider class of relaxation functions g, like in Guesmia10,15 and using
the recent results from,31
(3) study system (1.1) with local internal past history damping, in other words, by only assuming that b is positive on a
non empty subinterval of (0, L) that could be away from the boundary.
ACKNOWLEDGEMENT
The authors would like to thank the referees for their valuable comments and helpful suggestions.
ORCID
Mohammad Akil https://orcid.org/0000-0002-2994-5337
REFERENCES
1. Russell DL. A general framework for the study of indirect damping mechanisms in elastic systems. J Math Anal Appl. 1993;173(2):339-358.
2. Alabau F. Stabilisation frontière indirecte de systèmes faiblement couplés. Comptes Rendus de l'Acad des Sci - Ser I - Math.
1999;328(11):1015-1020.
6980 AKIL ET AL.
3. Alabau F, Cannarsa P, Komornik V. Indirect internal stabilization of weakly coupled evolution equations. J Evol Equa. 2002;2(2):127-150.
4. Alabau-Boussouira F. Indirect boundary stabilization of weakly coupled hyperbolic systems. SIAM J Control Optim. 2002;41(2):511-541.
5. Alabau-Boussouira F, Léautaud M. Indirect stabilization of locally coupled wave-type systems. ESAIM Control, Optim Calc Var.
2011;18(2):548-582.
6. Liu Z, Rao B. Frequency domain approach for the polynomial stability of a system of partially damped wave equations. J Math Anal Appl.
2007;335(2):860-881.
7. Zhang X, Zuazua E. Polynomial decay and control of a 1-D hyperbolic-parabolic coupled system. J Differential Equations.
2004;204(2):380-438.
8. Liu Z, Zheng S. On the exponential stability of linear viscoelasticity and thermoviscoelasticity. Quart Appl Math. 1996;54(1):21-31.
9. Rivera JEM, Sare HDF. Stability of Timoshenko systems with past history. J Math Anal Appl. 2008;339(1):482-502.
10. Guesmia A. Asymptotic stability of abstract dissipative systems with infinite memory. J Math Anal Appl. 2011;382(2):748-760.
11. Matos LPV, Júnior DSA, Santos ML. Polynomial decay to a class of abstract coupled systems with past history. Differ Integ Equa.
2012;25(11-12):1119-1134.
12. Fatori L, Monteiro R, Fernández Sare H. The Timoshenko system with history and Cattaneo law. Appl Math Comput. 2014;228:128-140.
13. Santos M, Soufyane A, Junior D. Asymptotic behavior to Bresse system with past history. Quart Appl Math. 2014;73:23-54.
14. Jin K-P, Liang J, Xiao T-J. Coupled second order evolution equations with fading memory: Optimal energy decay rate. J Differ Equa.
2014;257(5):1501-1528.
15. Guesmia A. Asymptotic behavior for coupled abstract evolution equations with one infinite memory. Appl Anal. 2015;94(1):184-217.
16. Alabau-Boussouira F, Wang Z, Yu L. A one-step optimal energy decay formula for indirectly nonlinearly damped hyperbolic systems
coupled by velocities. ESAIM: COCV. 2017;23(2):721-749.
17. Abdallah F, Ghader M, Wehbe A. Stability results of a distributed problem involving Bresse system with history and/or Cattaneo law under
fully dirichlet or mixed boundary conditions. Math Meth Appl Sci. 2018;41(5):1876-1907.
18. Cavalcanti MM, Domingos Cavalcanti VN, Silva MAJ, de Souza Franco AY. Exponential stability for the wave model with localized memory
in a past history framework. J Differential Equations. 2018;264(11):6535-6584.
19. Hao J, Wang P. General stability result of abstract thermoelastic system with infinite memory. Bull Malaysian Math Sci Soc.
2019;42(5):2549-2567.
20. Hassan JH, Messaoudi SA. General decay rate for a class of weakly dissipative second-order systems with memory. Math Methods Appl
Sci. 2019;42(8):2842-2853.
21. Jin K-P, Liang J, Xiao T-J. Asymptotic behavior for coupled systems of second order abstract evolution equations with one infinite memory.
J Math Anal Appl. 2019;475(1):554-575.
22. Almeida R, Santos M. Lack of exponential decay of a coupled system of wave equations with memory. Nonlin Anal Real World Appl.
2011;12(2):1023-1032.
23. Cordeiro S, Lobato R, Raposo C. Optimal polynomial decay for a coupled system of wave with past history. Open J Math Anal. 2020;4:49-59.
24. Dafermos CM. Asymptotic stability in viscoelasticity. Arch Rat Mech Anal. 1970;37(4):297-308.
25. Kato T. Perturbation theory for linear operators. Berlin Heidelberg: Springer; 1995.
26. Pazy A. Semigroups of Linear Operators and Applications to Partial Differential Equations, Volume 44 of Applied Mathematical Sciences.
New York: Springer-Verlag; 1983.
27. Arendt W, Batty CJK. Tauberian theorems and stability of one-parameter semigroups. Trans Amer Math Soc. 1988;306(2):837-852.
28. Akil M, Chitour Y, Ghader M, Wehbe A. Stability and exact controllability of a Timoshenko system with only one fractional damping on
the boundary. Asympt Anal. 2019;10:1-60.
29. Huang FL. Characteristic conditions for exponential stability of linear dynamical systems in Hilbert spaces. Ann Differential Equa.
1985;1(1):43-56.
30. Prüss J. On the spectrum of C0 -semigroups. Trans Amer Math Soc. 1984;284(2):847-857.
31. Rozendaal J, Seifert D, Stahn R. Optimal rates of decay for operator semigroups on Hilbert spaces. Adv Math. 2019;346:359-388.
32. Borichev A, Tomilov Y. Optimal polynomial decay of functions and operator semigroups. Math Ann. 2010;347(2):455-478.
33. Batty CJK, Duyckaerts T. Non-uniform stability for bounded semi-groups on Banach spaces. J Evol Equ. 2008;8(4):765-780.
34. Liu Z, Rao B. Characterization of polynomial decay rate for the solution of linear evolution equation. Z Angew Math Phys.
2005;56(4):630-644.
How to cite this article: Akil M, Badawi H, Nicaise S, Wehbe A. Stability results of coupled wave models
with locally memory in a past history framework via nonsmooth coefficients on the interface. Math Meth Appl Sci.
2021;44:6950-6981. https://doi.org/10.1002/mma.7235
AKIL ET AL. 6981
APPENDIX A : S OME NOTIONS AND STABILITY THEOREMS
In order to make this paper more self-contained, we recall in this short appendix some notions and stability results used
in this work.
( )
Definition A1. Assume that A is the generator of C0 semigroup of contractions etA t≥0 on a Hilbert space H. The C0
( tA )
semigroup e t≥0 is said to be
(1) Strongly stable if
lim ||etA x0 ||H = 0, ∀ x0 ∈ H.
t→+∞
(2) Exponentially (or uniformly) stable if there exists two positive constants M and 𝜀 such that
||etA x0 ||H ≤ Me−𝜀t ||x0 ||H , ∀ t > 0, ∀ x0 ∈ H.
(3) Polynomially stable if there exists two positive constants C and 𝛼 such that
||etA x0 ||H ≤ Ct−𝛼 ||Ax0 ||H , ∀ t > 0, ∀ x0 ∈ D(A).

( )
To show the strong stability of the C0 semigroup etA t≥0 we rely on the following result due to Arendt-Batty.27
( )
Theorem A2. Assume that A is the generator of a C0 semigroup of contractions etA t≥0 on a Hilbert space H. If A has
no ) imaginary eigenvalues and 𝜎 (A) ∩ iR is countable, where 𝜎 (A) denotes the spectrum of A, then the C0 semigroup
( tApure
e t≥0 is strongly stable.
Concerning the characterisation of exponential stability of C0 semigroup of contraction (etA )t ≥ 0 we rely on the following
result due to Huang29 and Pruss.30
( )
Theorem A3. Let A : D(A) ⊂ H → H generates a C0 semigroup of contractions etA t≥0 on H. Assume that iR ⊂ 𝜌(A).
( )
Then, the C0 semigroup etA t≥0 is exponentially stable if and only if
limsup ||(i𝜆I − A)−1 ||(H) < ∞.

𝜆∈R, |𝜆|→+∞
( )
Also, concerning the characterization of polynomial stability stability of a C0 semigroup of contraction etA t≥0 we
rely on the following result due to Borichev and Tomilov32 (see also Borichev and Tomilov33,34 and the recent paper of
Rozendaal et al.31 )
( )
Theorem A4. Assume that A is the generator of a strongly continuous semigroup of contractions etA t≥0 on . If iR ⊂
𝜌(), then for a fixed 𝓁 > 0, the following conditions are equivalent:
‖ ‖ ( )
sup ‖(i𝜆I − )−1 ‖ = O |𝜆|𝓁 , (A1)
𝜆∈R
‖ ‖()
C
||et U0 ||2 ≤ 2
||U0 ||2D() , ∀t > 0, U0 ∈ D(), for some C > 0. (A2)
t 𝓁

Stability Results of Coupled Wave Models With Locally Memory in A Past History Framework Via Nonsmooth Coefficients On The Interface

Uploaded by

Copyright:

Available Formats

You might also like

Stability Results of Coupled Wave Models With Locally Memory in A Past History Framework Via Nonsmooth Coefficients On The Interface

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Stability Results of Coupled Wave Models With Locally Memory in A Past History Framework Via Nonsmooth Coefficients On The Interface

Uploaded by

Copyright:

Available Formats

10991476, 2021, 8, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/mma.7235 by UFPR - Universidade Federal do Parana, Wiley Online Library on [09/02/2023].

Stability results of coupled wave models with locally

Mohammad Akil1 Haidar Badawi2 Serge Nicaise2 Ali Wehbe3

FIGURE 1 Geometric description of the functions b(x) and c(x)

⎧ u + A u − ∫ ∞ g(s)A u(t − s)ds + 𝛽v = 0,

⎧ utt (t) + Au(t) − ∫ t g(t − s)Au(s)ds + 𝛼u(t) + 𝛽Bv(t) = 𝑓 (u(t)), t > 0,

⎧ utt (t) + A1 u(t) − ∫0∞ g(s)Au(t − s)ds + Bv(t) = 0, t > 0,

Then, system (1.1) becomes

𝑦tt − 𝑦xx − c(·)ut = 0, (x, t) ∈ (0, L) × (0, ∞), (2.3)

𝜔t (·, s, t) + 𝜔s (·, s, t) − ut = 0, (x, s, t) ∈ (0, 𝛽) × (0, ∞) × (0, ∞), (2.4)

With the following boundary conditions

and the following initial conditions

The energy of system (2.2)–(2.6) is given by

E(t) = E1 (t) + E2 (t) + E3 (t), (2.7)

Deriving (2.4) with respect to x, we obtain

The Hilbert space  is equipped with the following inner product

Ut = U, U(0) = U0 , (2.16)

where U0 = (u0 (·, 0), u1 , 𝑦0 , 𝑦1 , 𝜔0 (·, s))⊤ ∈ .

Equivalently, we have the following system

−𝑦xx − c(·)v = 𝑓 4 , (2.22)

𝜔s (·, s) − v = 𝑓 5 (·, s), (2.23)

From (2.19), (2.23), and (2.24), we get

Using integration by parts in (2.26), we obtain

Moreover, from Young's inequality, we have

Inserting (2.27) in the above inequality, we get

and consequently, 𝜔(·, s) ∈ g . Now, see the definition of (Sb(·)

𝑦xx − c(·)𝑓 1 = − 𝑓 4 , (2.29)

u(0) = u(L) = 𝑦(0) = 𝑦(L) = 0, (2.30)

U(x, s, t) = et U0 (x, s) ∈ C0 (R+ , ).

U(x, s, t) = et U0 (x, s) ∈ C0 (R+ , D()) ∩ C1 (R+ , ).

lim ||et U0 || = 0.

Equivalently, we have the following system

𝑦xx + c(·)v = i𝜆z, (3.5)

−𝜔s (·, s) + v = i𝜆𝜔(·, s). (3.6)

𝜔(·, s) = 0 in (0, 𝛽) × (0, ∞). (3.10)

From (3.2), (3.6), and (3.10), we obtain

𝜆2 𝑦 + 𝑦xx + c(·)i𝜆u = 0, in (0, L), (3.13)

u(0) = u(L) = 𝑦(0) = 𝑦(L) = 0. (3.14)

𝑦 = z = 0 in (𝛼, 𝛽). (3.15)

𝑦(𝛼) = 𝑦x (𝛼) = 0. (3.16)

𝜆2 𝑦 + 𝑦xx = 0, in (0, 𝛼), (3.17)

𝑦(0) = 𝑦(𝛼) = 𝑦x (𝛼) = 0. (3.18)

𝑦 = 0 in (0, 𝛼). (3.19)

Therefore, from (3.4) and (3.19), we obtain

𝑦 = z = 0 in (0, 𝛼). (3.20)

̃ (u, 𝜔)) and the fact that b̃0 = a − b0 g

From (3.8), (3.11), and (3.21), we get

Thus, from (3.22) and the fact that U ∈ D(), we obtain

uxx ∈ L2 (0, L) and consequently u ∈ C1 ([0, L]). (3.23)

u(𝛽) = ux (𝛽) = 𝑦(𝛽) = 𝑦x (𝛽) = 0. (3.24)

𝜆2 u + auxx − c0 i𝜆𝑦 = 0, in (𝛽, 𝛾), (3.25)

𝜆2 𝑦 + 𝑦xx + c0 i𝜆u = 0, in (𝛽, 𝛾), (3.26)

u(𝛽) = ux (𝛽) = 𝑦(𝛽) = 𝑦x (𝛽) = 0. (3.27)

Vx = BV, V(𝛽) = 0. (3.28)

V(x) = eB(x−𝛽) V(𝛽). (3.29)

Thus, from (3.29) and the fact that V(𝛽) = 0, we get

V = 0 in (𝛽, 𝛾) and consequently u = ux = 𝑦 = 𝑦x = 0 in (𝛽, 𝛾). (3.30)

So, from (3.2), (3.4), and (3.30), we get