Ownership and Copyright (Springer-Verlag Limited

Math. Control Signals Systems (2002) 15: 229–255
Mathematics of Control,
Signals, and Systems
Decay Rates for a Beam with Pointwise Force and
Moment Feedback*
Kais Ammari,
y
Zhuangyi Liu,
z
and Marius Tucsnak
y
Abstract. We consider the Rayleigh beam equation and the Euler–Bernoulli
beam equation with pointwise feedback shear force and bending moment at the
position x in a bounded domain ð0; pÞ with certain boundary conditions. The en-
ergy decay rate in both cases is investigated. In the case of the Rayleigh beam, we
show that the decay rate is exponential if and only if x=p is a rational number
with coprime factorization x=p ¼ p=q, where q is odd. Moreover, for any other
location of the actuator we give explicit polynomial decay estimates valid for reg-
ular initial data. In the case of the Euler–Bernoulli beam, even for a nonhomo-
geneous material, exponential decay of the energy is proved, independently of the
position of the actuator.
Key words. Pointwise control, Exponential decay, Polynomial decay, Observ-
ability inequality.
1. Introduction
Pointwise stabilization of beam systems is a subject widely studied in the literature
(see, for instance, [7]–[9], [17] and [20]). To our knowledge, the stabilizers used are
either of force feedback type or of moment feedback type acting at the joints. It is
known that the stability properties of such a system depend on the location of the
joints in an unrobust way. For instance, for a beam hinged at both ends and of
length p, in order to have strong stabilization the abscissa of the joint has to be
irrational and for exponential stabilizability this abscissa has to satisfy even more
restrictive conditions (see, for instance, [2] for an Euler–Bernoulli beam). In this
paper we investigate the stabilization e¤ect of applying both force and moment
feedbacks simultaneously at the joint. As far as we know, this problem has not yet
been tackled in the literature. Most of the results we obtain are robust with respect
to the location of the joint.
First we consider the following initial and boundary value problem for a
* Date received: October 30, 2000. Date revised: December 20, 2001.
y
Institut Elie Cartan, De´partement de Mathe´matiques, Universite´ de Nancy I, F-54506 Vandoeuvre
le`s Nancy Cedex, France. {ammari,tucsnak}@iecn.u-nancy.fr.
z
Department of Mathematics and Statistics, University of Minnesota, Duluth, Minnesota 55812-
2496, U.S.A. zliu@d.umn.edu.
229
homogenous Rayleigh beam with force and moment damping at x ¼ x:
q
2
u
qt
2
À
q
4
u
qx
2
qt
2
þ
q
4
u
qx
4
þ
qu
qt
ðx; tÞd
x
À
q
2
u
qt qx
ðx; tÞ
dd
x
dx
¼ 0; 0 < x < p; t > 0;
ð1:1Þ
uð0; tÞ ¼ uðp; tÞ ¼ 0;
q
2
u
qx
2
ð0; tÞ ¼
q
2
u
qx
2
ðp; tÞ ¼ 0; t > 0; ð1:2Þ
uðx; 0Þ ¼ u
0
ðxÞ;
qu
qt
ðx; 0Þ ¼ u
1
ðxÞ; 0 < x < p: ð1:3Þ
We also consider a nonhomogenous Euler–Bernoulli beam:
rðxÞ
q
2
u
qt
2
À
q
2
qx
2
pðxÞ
q
2
u
qx
2
_ _
þ
qu
qt
ðx; tÞd
x
À
q
2
u
qx qt
ðx; tÞ
dd
x
dx
¼ 0; 0 < x < p; t > 0;
ð1:4Þ
uð0; tÞ ¼ uðp; tÞ ¼
q
2
u
qx
2
ð0; tÞ ¼ 0;
q
2
u
qx
2
ðp; tÞ ¼ 0; t > 0; ð1:5Þ
uðx; 0Þ ¼ u
0
ðxÞ;
qu
qt
ðx; 0Þ ¼ u
1
ðxÞ; 0 < x < p: ð1:6Þ
Here d
x
is the Dirac mass concentrated in the point x A ð0; pÞ, u is the transverse
displacement of the beam and the coe‰cient functions rðxÞ, pðxÞ A C
1
ð0; pÞ sat-
isfy rðxÞ br
0
> 0, pðxÞ bp
0
> 0.
Remark 1.1. We denote by ½H
2
ð0; pÞ XH
1
0
ð0; pފ
Ã
the dual space of H
2
ð0; pÞ X
H
1
0
ð0; pÞ with respect to the pivot space L
2
ð0; pÞ. If u is smooth on ½ð0; pÞnfxgŠ Â
½0; TŠ, then u satisfies (1.1) in the space Cð0; T; ½H
2
ð0; pÞ XH
1
0
ð0; pފ
Ã
Þ if and only
if
q
2
u
qt
2
À
q
4
u
qx
2
qt
2
þ
q
4
u
qx
4
¼ 0; x A ð0; pÞnfxg; t > 0;
and
½uðÁ ; tފ
x
¼ 0;
qu
qx
ðÁ ; tÞ
_ _
x
¼ 0;
q
2
u
qx
2
ðÁ ; tÞ
_ _
x
¼
q
2
u
qx qt
ðx; tÞ;
q
3
u
qx
3
ðÁ ; tÞ
_ _
x
¼ À
qu
qt
ðx; tÞ
_
¸
¸
¸
¸
_
¸
¸
¸
¸
_
; Et > 0;
where ½ f Š
x
denotes the jump of the function f at the point x. A similar assertion
holds if u is smooth on ½ð0; pÞnfxgŠ  ½0; TŠ and u satisfies (1.4) so that the terms
are in the space Cð0; T; ½H
2
ð0; pÞ XH
1
0
ð0; pފ
Ã
Þ. Thus, (1.1) (respectively (1.4))
is equivalent to the equations modeling the vibrations of two Rayleigh beams (re-
spectively two Euler–Bernoulli beams) with a dissipative joint at x ¼ x, where
both feedback force and feedback moment are applied. The results stated in this
remark, which can be checked by standard calculations, are not used in the re-
maining part of the paper, and we omit their proof.
230 K. Ammari, Z. Liu, and M. Tucsnak
The purpose of this paper is to give a complete characterization of points x for
which the solutions of (1.1)–(1.3) (respectively (1.4)–(1.6)) are exponentially stable
in the energy space, and give an explicit decay rate for regular initial data when
we do not have exponential stability in the energy space. A significant improve-
ment of our new stabilization strategy is that we have avoided the well-known prob-
lems concerning the existence of nodal points. The method used to study (1.1)–
(1.3) is based on some trace regularity results which reduce stability to observ-
ability inequalities for the corresponding undamped problem. As far as we know,
in the case of unbounded feedback, this method is new. Both the regularity and
the observability results are based on Fourier analysis, so we can tackle only the
constant coe‰cients case. In order to study (1.4)–(1.6), we use a frequency domain
method and combine a contradiction argument with the multiplier technique to
carry out a special analysis for the resolvent.
The paper is organized as follows. The statements of the main results are given
in the following section. Sections 3 and 4 are devoted to some trace regularity and
observability results needed in what follows. The proof of the main results is given
in Sections 5 and 6.
2. Statement of the Main Results
We define the energy of a solution u of (1.1)–(1.3) and of a solution u of (1.4)–
(1.6) at the time instant t by
EðuðtÞÞ ¼
1
2
_
p
0
qu
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
qx
2
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dx ð2:1Þ
and
~
EEðuðtÞÞ ¼
1
2
_
p
0
pðxÞ
q
2
u
qx
2
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ rðxÞ
qu
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dx; ð2:2Þ
respectively. We can easily check that every su‰ciently smooth solution of (1.1)–
(1.3) (respectively (1.4)–(1.6)) satisfies the energy identity
Eðuðt
2
ÞÞ À Eðuðt
1
ÞÞ ¼ À
_
t
2
t
1
qu
qt
ðx; sÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
qt qx
ðx; sÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
ds; ð2:3Þ
respectively
~
EEðuðt
2
ÞÞ À
~
EEðuðt
1
ÞÞ ¼ À
_
t
2
t
1
qu
qt
ðx; sÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
qt qx
ðx; sÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
ds; ð2:4Þ
for all t
2
> t
1
b0 and, therefore, the energy is a nonincreasing function of the
time variable t. Denote
V ¼ H
2
ð0; pÞ XH
1
0
ð0; pÞ:
We define the concept of a finite energy solution of (1.1)–(1.3) and of (1.4)–(1.6).
This concept will be frequently used in what follows.
Decay Rates for a Beam with Pointwise Force and Moment Feedback 231
Definition 2.1. A function
u A Cð0; T; VÞ XC
1
ð0; T; H
1
0
ð0; pÞÞ
is called a finite energy solution of (1.1)–(1.3) if ðqu=qxÞðx; ÁÞ A H
1
ð0; TÞ, u satisfies
(1.3) and (1.1) is satisfied in L
2
ð0; T; V
Ã
Þ, where V
Ã
is the dual space of V with
respect to the pivot space L
2
ð0; pÞ.
Definition 2.2. A function
u A Cð0; T; VÞ XC
1
ð0; T; L
2
ð0; pÞÞ
is called a finite energy solution of (1.4)–(1.6) if ðqu=qxÞðx; ÁÞ A H
1
ð0; TÞ and (1.4)
is satisfied in L
2
ð0; T; V
Ã
Þ.
The result below concerns the well-posedness of the solutions of (1.1)–(1.3) and
the behavior of EðuðtÞÞ when t ! y. The first assertion in the proposition below
was proved, for instance, in the survey paper [22] by using the properties of a gen-
eral class of ‘‘conservative systems.’’ The second part of the proposition is proved
in Section 5.
Proposition 2.3. Let
u
0
u
1
_ _
A V Â H
1
0
ð0; pÞ. Then the following assertions hold:
1. Problem (1.1)–(1.3) admits a unique finite energy solution such that
kuðx; ÁÞk
2
H
1
ð0; TÞ
þ
qu
qx
ðx; ÁÞ
_
_
_
_
_
_
_
_
2
H
1
ð0; TÞ
aCðku
0
k
2
H
2
ð0; pÞ
þ ku
1
k
2
H
1
ð0; pÞ
Þ; ð2:5Þ
where the constant C > 0 depends only on x and T. Moreover, u satisfies the
energy estimate (2.3).
2. We have lim
t!þy
EðuðtÞÞ ¼ 0, for any initial data
u
0
u
1
_ _
in V Â H
1
0
ð0; pÞ.
The next result concerns the well-posedness of (1.4)–(1.6).
Proposition 2.4. Let
u
0
u
1
_ _
A V Â L
2
ð0; pÞ. Then problem (1.4)–(1.6) admits a
unique finite energy solution such that
kuðx; ÁÞk
2
H
1
ð0; TÞ
þ
qu
qx
ðx; ÁÞ
_
_
_
_
_
_
_
_
2
H
1
ð0; TÞ
aCðku
0
k
2
H
2
ð0; pÞ
þ ku
1
k
2
L
2
ð0; pÞ
Þ; ð2:6Þ
where the constant C > 0 depends only on x and T. Moreover, u satisfies the energy
estimate (2.4).
The main result of this paper concerns the precise asymptotic behavior of the
solutions of (1.1)–(1.3) and (1.4)–(1.6). Denote
Y ¼
u
v
_ _
A V Â V j u
jð0; xÞ
A H
3
ð0; xÞ and u
jðx; pÞ
A H
3
ðx; pÞ
_ _
ð2:7Þ
232 K. Ammari, Z. Liu, and M. Tucsnak
and
DðAÞ ¼
u
v
_ _
A Y;
d
2
u
dx
2
ð0Þ ¼
d
2
u
dx
2
ðpÞ ¼ 0;
d
2
u
dx
2
_ _
x
¼
dv
dx
ðxÞ;
d
3
u
dx
3
_ _
x
¼ ÀvðxÞ
_ _
;
ð2:8Þ
where ½gŠ
x
denotes the jump of the function g at the point x. The corresponding
operator A is defined on DðAÞ by
A
u
v
_ _
¼
v
ÀR
d
4
u
dx
4
_ _
À vðxÞRðd
x
Þ þ
dv
dx
ðxÞR
dd
x
dx
_ _
_
_
_
_
;
where the operator R is defined by
R ¼ I À
d
2
dx
2
_ _
À1
: ð2:9Þ
It is well known that R is an isomorphism from H
À1
ð0; pÞ onto H
1
0
ð0; pÞ.
If
u
0
u
1
_ _
A Y, we denote
u
0
u
1
_ _ _
_
_
_
_
_
_
_
2
Y
¼ ku
0
k
2
H
3
ð0; xÞ
þ ku
0
k
2
H
3
ðx; pÞ
þ ku
1
k
2
H
2
ð0; pÞ
: ð2:10Þ
Concerning the system (1.1)–(1.3) our main result is the following.
Theorem 2.5.
1. Let u A Cð0; T; VÞ XC
1
ð0; T; H
1
0
ð0; pÞÞ be a finite energy solution of the sys-
tem (1.1)–(1.3) (the existence and the uniqueness of such a solution follows
from Proposition 2.3). Then u satisfies the estimate
EðuðtÞÞ aMe
Àot
ðku
0
k
2
V
þ ku
1
k
2
L
2
ð0; pÞ
Þ; Et b0; ð2:11Þ
for some constants M; o > 0 (depending only on x), if and only if x=p is a
rational number with coprime factorization
x
p
¼
p
q
; where q is odd: ð2:12Þ
2. For any x A ð0; pÞ, there exists a constant C > 0 (independent of x) such that
for all t b0 we have
EðuðtÞÞ a
C
t þ 1
u
0
u
1
_ _ _
_
_
_
_
_
_
_
2
Y
; E
u
0
u
1
_ _
A DðAÞ: ð2:13Þ
For the system (1.4)–(1.6) our main result is the following.
Theorem 2.6. Let u A Cð0; T; VÞ XC
1
ð0; T; L
2
ð0; pÞÞ be a finite energy solution
of (1.4)–(1.6) (the existence and the uniqueness of such a solution follows from Prop-
osition 2.4). Then, for all x A ð0; pÞ, there exist constants M; o > 0 (depending only
Decay Rates for a Beam with Pointwise Force and Moment Feedback 233
on x), such that
~
EEðuðtÞÞ aMe
Àot
ðku
0
k
2
V
þ ku
1
k
2
L
2
ð0; pÞ
Þ; Et b0: ð2:14Þ
In other terms, for all x A ð0; pÞ, (1.4)–(1.6) define an exponentially stable system in
V Â L
2
ð0; pÞ.
3. Regularity Inequalities
Consider the open-loop system associated to (1.1)–(1.3):
q
2
c
qt
2
À
q
4
c
qx
2
qt
2
þ
q
4
c
qx
4
¼ v
1
ðtÞd
x
þ v
2
ðtÞ
dd
x
dx
; 0 < x < p; t > 0; ð3:1Þ
cð0; tÞ ¼ cðp; tÞ ¼ 0;
q
2
c
qx
2
ð0; tÞ ¼
q
2
c
qx
2
ðp; tÞ ¼ 0; t > 0; ð3:2Þ
cðx; 0Þ ¼ 0;
qc
qt
ðx; 0Þ ¼ 0; 0 < x < p: ð3:3Þ
The finite energy solutions of (3.1)–(3.3) are defined as follows.
Definition 3.1. Let v
1
; v
2
A L
2
ð0; TÞ. Then a function
c A Cð0; T; VÞ XC
1
ð0; T; H
1
0
ð0; pÞÞ XH
2
ð0; T; L
2
ð0; pÞÞ ð3:4Þ
is called a finite energy solution of (3.1)–(3.3) if c satisfies (3.3) and (3.1) is
satisfied in L
2
ð0; T; V
Ã
Þ.
The following existence and trace regularity result for the system (3.1)–(3.3) is
essential for the proof of Theorem 2.5.
Proposition 3.2. Suppose that v
1
; v
2
A L
2
ð0; TÞ. Then the system (3.1)–(3.3) admits
a unique finite energy solution c. Moreover, ðqc=qxÞðx; ÁÞ A H
1
ð0; TÞ and there
exists C > 0 (depending only on T) such that
kcðx; ÁÞk
2
H
1
ð0; TÞ
þ
qc
qx
ðx; ÁÞ
_
_
_
_
_
_
_
_
2
H
1
ð0; TÞ
aCðkv
1
k
2
L
2
ð0; TÞ
þ kv
2
k
2
L
2
ð0; TÞ
Þ: ð3:5Þ
Before proving Proposition 3.2, we consider the following homogeneous undamped
problem:
q
2
j
qt
2
À
q
4
j
qx
2
qt
2
þ
q
4
j
qx
4
¼ 0; 0 < x < p; t > 0; ð3:6Þ
jð0; tÞ ¼ jðp; tÞ ¼ 0;
q
2
j
qx
2
ð0; tÞ ¼
q
2
j
qx
2
ðp; tÞ ¼ 0; t > 0; ð3:7Þ
jðx; 0Þ ¼ u
0
ðxÞ;
qj
qt
ðx; 0Þ ¼ u
1
ðxÞ; 0 < x < p: ð3:8Þ
234 K. Ammari, Z. Liu, and M. Tucsnak
Lemma 3.3. For any initial data
u
0
u
1
_ _
A V Â H
1
0
ð0; pÞ there exists a unique
finite energy solution of (3.6)–(3.8) (in the sense of Definition 3.1). Moreover,
ðqj=qxÞðx; ÁÞ A H
1
ð0; TÞ and there exists C > 0, depending only on T, such that
kjðx; ÁÞk
2
H
1
ð0; TÞ
þ
qj
qx
ðx; ÁÞ
_
_
_
_
_
_
_
_
2
H
1
ð0; TÞ
aCðku
0
k
2
H
2
ð0; pÞ
þ ku
1
k
2
H
1
ð0; pÞ
Þ: ð3:9Þ
Proof. It is easy to see, by the semigroup method [18], that the problem (3.6)–
(3.8) is well-posed in the energy space V Â H
1
0
ð0; pÞ. In order to prove (3.9) we
put
u
0
ðxÞ ¼

kb1
a
k
sinðkxÞ; u
1
ðxÞ ¼

kb1
b
k
sinðkxÞ; ð3:10Þ
with ðk
2
a
k
Þ; ðkb
k
Þ A l
2
ðRÞ. A simple calculation shows that
jðx; tÞ ¼

kb1
a
k
cos
k
2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ k
2
p t
_ _
þ
b
k
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ k
2
p
k
2
sin
k
2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ k
2
p t
_ _
_ _
sinðkxÞ; ð3:11Þ
which implies that, for all T > 0, we have
_
T
0
qj
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
j
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt aC

kb1
k
2
a
2
k
k
4
1 þ k
2
þ b
2
k
_ _
aC

kb1
k
2
½a
2
k
k
2
þ b
2
k
Š;
where C > 0 is a constant, depending only on T. This clearly yields (3.9). 9
It can be easily checked that c is a finite energy solution of (3.1)–(3.3) if and
only
q
2
c
qt
2
þR
q
4
c
qx
4
_ _
¼ v
1
ðtÞRðd
x
Þ þ v
2
ðtÞR
dd
x
dx
_ _
in Cð0; T; H
1
0
ð0; pÞÞ; ð3:12Þ
cð0; tÞ ¼ cðp; tÞ ¼ 0; t > 0; ð3:13Þ
cðx; 0Þ ¼ 0;
qc
qt
ðx; 0Þ ¼ 0; 0 < x < p; ð3:14Þ
where the operator R is defined by (2.9). To prove (3.5), we need three technical
results. The first one gives the boundedness of some functions which will be used
in what follows.
Lemma 3.4. Let a > 0, x A ð0; pÞ and denote C
a
¼ fl A Cj Re l ¼ ag. Then the
Decay Rates for a Beam with Pointwise Force and Moment Feedback 235
functions H
i
: C
a
!C, i A f1; 2; 3; 4g, defined by
H
1
ðlÞ ¼ À
sinhðlxÞ
sinhðlpÞ
_
p
0
sinh½lðp À yފRðd
x
ÞðyÞ dy
þ
_
x
0
sinh½lðx À yފRðd
x
ÞðyÞ dy;
H
2
ðlÞ ¼ À
sinhðlxÞ
sinhðlpÞ
_
p
0
sinh½lðp À yފR
dd
x
dy
_ _
ðyÞ dy
þ
_
x
0
sinh½lðx À yފR
dd
x
dy
_ _
ðyÞ dy;
H
3
ðlÞ ¼ Àl
coshðlxÞ
sinhðlpÞ
_
p
0
sinh½lðp À yފRðd
x
ÞðyÞ dy
þ l
_
x
0
cosh½lðx À yފRðd
x
ÞðyÞ dy;
H
4
ðlÞ ¼ Àl
coshðlxÞ
sinhðlpÞ
_
p
0
sinh½lðp À yފR
dd
x
dy
_ _
ðyÞ dy
þ l
_
x
0
cosh½lðx À yފR
dd
x
dy
_ _
ðyÞ dy;
are bounded on C
a
.
Proof. The functions H
i
ðlÞ, for i ¼ 1; 2; 3; 4, are continuous on C
a
. Then, to
prove that these functions are bounded on C
a
, it is su‰cient to prove that they are
bounded when l A C
a
, Im l ! y. A simple calculation shows that
Rðd
x
Þ ¼
À
sinhðx À pÞ sinhðxÞ
sinhðpÞ
; x A ð0; xÞ;
À
sinhðxÞ sinhðx À pÞ
sinhðpÞ
; x A ðx; pÞ;
_
¸
¸
¸
_
¸
¸
¸
_
R
dd
x
dx
_ _
¼
coshðx À pÞ sinhðxÞ
sinhðpÞ
; x A ð0; xÞ;
coshðxÞ sinhðx À pÞ
sinhðpÞ
; x A ðx; pÞ:
_
¸
¸
¸
_
¸
¸
¸
_
ð3:15Þ
It follows that
H
1
ðlÞ ¼
sinhðlxÞ sinhðx À pÞ
sinhðlpÞ sinhðpÞ
_
x
0
sinh½lðp À yފ sinhðyÞ dy
þ
sinhðlxÞ sinhðxÞ
sinhðlpÞ sinhðpÞ
_
p
x
sinh½lðp À yފ sinhðy À pÞ dy
À
sinhðx À pÞ
sinhðpÞ
_
x
0
sinh½lðx À yފ sinhðyÞ dy;
236 K. Ammari, Z. Liu, and M. Tucsnak
H
2
ðlÞ ¼ À
sinhðlxÞ coshðx À pÞ
sinhðlpÞ sinhðpÞ
_
x
0
sinh½lðp À yފ sinhðyÞ dy
À
sinhðlxÞ coshðxÞ
sinhðlpÞ sinhðpÞ
_
p
x
sinh½lðp À yފ sinhðy À pÞ dy
þ
coshðx À pÞ
sinhðpÞ
_
x
0
sinh½lðx À yފ sinhðyÞ dy:
Moreover, by using the monotonicity of the functions sinh, cosh combined with
the fact that, for any z A C, we have jsinhðzÞj ajcoshðRe zÞj, we deduce that
sup
l A C
a
jH
1
ðlÞj a
coshðaxÞ sinhðp À xÞ sinhðxÞ
sinhðpÞ sinhðapÞ
Á fx coshðapÞ þ ðp À xÞ cosh½aðp À xފ þ x sinhðapÞg
and
sup
l A C
a
jH
2
ðlÞj ax
coshðaxÞ coshðp À xÞ sinhðxÞ
sinhðpÞ sinhðapÞ
fsinhðapÞ þ sinhðapÞg
þ ðp À xÞ
cosh½aðp À xފ sinhðp À xÞ coshðaxÞ coshðxÞ
sinhðpÞ sinhðapÞ
:
Thus H
1
and H
2
are bounded on C
a
. In order to show that H
3
and H
4
are also
bounded on C
a
we notice that the definition of H
3
and H
4
, combined with (3.15),
gives
H
3
ðlÞ ¼ l
coshðlxÞ sinhðx À pÞ
sinhðlpÞ sinhðpÞ
_
x
0
sinh½lðp À yފ sinhðyÞ dy
þ l
coshðlxÞ sinhðxÞ
sinhðlpÞ sinhðpÞ
_
p
x
sinh½lðp À yފ sinhðy À pÞ dy
À l
sinhðx À pÞ
sinhðpÞ
_
x
0
cosh½lðx À yފ sinhðyÞ dy
and
H
4
ðlÞ ¼ Àl
coshðlxÞ coshðx À pÞ
sinhðlpÞ sinhðpÞ
_
x
0
sinh½lðp À yފ sinhðyÞ dy
À l
coshðlxÞ coshðxÞ
sinhðlpÞ sinhðpÞ
_
p
x
sinh½lðp À yފ sinhðy À pÞ dy
þ l
coshðx À pÞ
sinhðpÞ
_
x
0
cosh½lðx À yފ sinhðyÞ dy:
Decay Rates for a Beam with Pointwise Force and Moment Feedback 237
Thus, by integration by parts, we obtain
H
3
ðlÞ ¼
coshðlxÞ sinhðx À pÞ
sinhðlpÞ sinhðpÞ
_
x
0
cosh½lðp À yފ coshðyÞ dy
þ
coshðlxÞ sinhðxÞ
sinhðlpÞ sinhðpÞ
_
p
x
cosh½lðp À yފ coshðy À pÞ dy
À
sinhðx À pÞ
sinhðpÞ
_
x
0
sinh½lðx À yފ coshðyÞ dy
and
H
4
ðlÞ ¼ À
coshðlxÞ coshðx À pÞ
sinhðlpÞ sinhðpÞ
_
x
0
cosh½lðp À yފ coshðyÞ dy
À
coshðlxÞ coshðxÞ
coshðlpÞ coshðpÞ
_
p
x
cosh½lðp À yފ coshðy À pÞ dy
þ
coshðx À pÞ
sinhðpÞ
_
x
0
sinh½lðx À yފ coshðyÞ dy þ
coshðlxÞ cosh½lðp À xފ
sinhðlpÞ
:
Then
sup
l A C
a
jH
3
ðlÞj ax
coshðaxÞ coshðx À pÞ
sinhðapÞ sinhðpÞ
coshðapÞ coshðxÞ
þ ðp À xÞ
coshðaxÞ coshðxÞ
sinhðapÞ sinhðpÞ
cosh½aðp À xފ coshðx À pÞ
þ x
coshðx À pÞ
sinhðpÞ
coshðaxÞ coshðxÞ
and
sup
l A C
a
jH
4
ðlÞj ax
coshðaxÞ coshðx À pÞ
sinhðapÞ sinhðpÞ
coshðapÞ coshðxÞ
þ ðp À xÞ
coshðaxÞ coshðxÞ
sinhðapÞ coshðpÞ
cosh½aðp À xފ coshðx À pÞ
þ x
coshðx À pÞ
sinhðpÞ
coshðaxÞ coshðxÞ þ
coshðaxÞ cosh½aðp À xފ
sinhðapÞ
: 9
The following result gives a regularity property for the string equation. This re-
sult, combined with the fact that the operator Rðq
4
=qx
4
Þ is ‘‘similar’’ (in a sense
which will be made precise later) to the second-order derivative operator, will be
used in the proof of Proposition 3.2.
Lemma 3.5. Suppose that v
1
; v
2
A L
2
ð0; TÞ and that c
2
A Cð0; T; VÞ X
C
1
ð0; T; L
2
ð0; pÞÞ satisfies the conditions
q
2
c
2
qt
2
À
q
2
c
2
qx
2
¼ v
1
ðtÞRðd
x
Þ þ v
2
ðtÞR
dd
x
dx
_ _
; 0 < x < p; t > 0; ð3:16Þ
238 K. Ammari, Z. Liu, and M. Tucsnak
c
2
ð0; tÞ ¼ c
2
ðp; tÞ ¼ 0; t > 0; ð3:17Þ
c
2
ðx; 0Þ ¼ 0;
qc
2
qt
ðx; 0Þ ¼ 0; 0 < x < p; ð3:18Þ
where the operator R is defined in (2.9). Then ðqc
2
=qxÞðx; ÁÞ A H
1
ð0; TÞ and there
exists C > 0, depending only on T, such that
kc
2
ðx; ÁÞk
2
H
1
ð0; TÞ
þ
qc
2
qx
ðx; ÁÞ
_
_
_
_
_
_
_
_
2
H
1
ð0; TÞ
aCðkv
1
k
2
L
2
ð0; TÞ
þ kv
2
k
2
L
2
ð0; TÞ
Þ: ð3:19Þ
Proof. As (3.16) is time reversible, after extending v
1
; v
2
by zero for t A Rn½0; TŠ,
we can solve (3.16)–(3.18), for t A R. In this way, it follows that c
2
can be extended
to a function, denoted also by c
2
, such that
c
2
A C
B
ðR; VÞ XC
1
B
ðR; H
1
0
ð0; pÞÞ;
c
2
ðtÞ ¼ 0; Et a0;
_
ð3:20Þ
and c
2
satisfies (3.16)–(3.17) for all t A R (above we denoted by C
k
B
ðR; XÞ the
space of C
k
functions from R into X, which are bounded on R, together with
their derivatives up to the order k). Let
´
c
2
c
2
ðlÞ, where l ¼ g þ ih, g > 0 and h A R,
be the Laplace (with respect to t) transform of c
2
. Since c
2
satisfies (3.20), in
order to prove (3.19) it su‰ces to prove that the functions t ! e
Àgt
c
2
ðx; tÞ and
e
Àgt
ðqc
2
=qxÞðx; tÞ belong to H
1
ðRÞ and that there exist two constants M
1
; M
2
> 0
such that
ke
Àg:
c
2
ðx; ÁÞk
2
H
1
ðÀy; þyÞ
aM
1
ðkv
1
ðÁÞk
2
L
2
ðÀy; þyÞ
þ kv
2
ðÁÞk
2
L
2
ðÀy; þyÞ
Þ
and
e
Àg:
qc
2
qx
ðx; ÁÞ
_
_
_
_
_
_
_
_
2
H
1
ðÀy; þyÞ
aM
2
ðkv
1
ðÁÞk
2
L
2
ðÀy; þyÞ
þ kv
2
ðÁÞk
2
L
2
ðÀy; þyÞ
Þ:
Since the function t ! e
Àgt
ðqc
2
=qxÞðx; tÞ is in L
2
ðRÞ, by the Parseval identity (see
for instance p. 212 of [11]), it su‰ces to prove that the functions
h ! ðg þ ihÞ
^
cc
2
ðx; g þ ihÞ; ðg þ ihÞ
q
^
cc
2
qx
ðx; g þ ihÞ
belong to L
2
ðRÞ, for some g > 0, and that there exist two constants M
3
; M
4
> 0
such that
_
þy
Ày
jðg þ ihÞ
^
cc
2
ðx; g þ ihÞj
2
dh aM
3
_
þy
Ày
ðj^ vv
1
ðg þ ihÞj
2
þ j^ vv
2
ðg þ ihÞj
2
Þ dh ð3:21Þ
and
_
þy
Ày
jðg þ ihÞ
q
^
cc
2
qx
ðx; g þ ihÞj
2
dh aM
4
_
þy
Ày
ðj^ vv
1
ðg þ ihÞj
2
þ j^ vv
2
ðg þ ihÞj
2
Þ dh:
ð3:22Þ
Decay Rates for a Beam with Pointwise Force and Moment Feedback 239
It can be easily checked that, for all x A ð0; pÞ, the function
´
c
2
c
2
satisfies
l
2
´
c
2
c
2
ðx; lÞ À
q
2
^
cc
2
qx
2
ðx; lÞ ¼ ^ vv
1
ðlÞRðd
x
ÞðxÞ þ ^ vv
2
ðlÞR
dd
x
dx
_ _
ðxÞ; Re l > 0;
ð3:23Þ
^
cc
2
ð0; lÞ ¼
^
cc
2
ðp; lÞ ¼ 0; Re l > 0: ð3:24Þ
Then
^
cc
2
ðx; lÞ ¼ À
sinhðlxÞ
l sinhðlpÞ
_
p
0
sinh½lðp À yފ ^ vv
1
ðlÞRðd
x
Þ þ ^ vv
2
ðlÞR
dd
x
dx
_ _ _ _
dy
þ
1
l
_
x
0
sinh½lðx À yފ ^ vv
1
ðlÞRðd
x
Þ þ ^ vv
2
ðlÞR
dd
x
dx
_ _ _ _
dy;
x A ð0; pÞ; Re l > 0; ð3:25Þ
where Rðd
x
Þ and Rðdd
x
=dxÞ are given by (3.15).
The relations above imply that, for every l A C, Re l > 0, we have
l
´
c
2
c
2
ðx; lÞ ¼ H
1
ðlÞ^ vv
1
ðlÞ þ H
2
ðlÞ^ vv
2
ðlÞ; ERe l > 0; ð3:26Þ
and
l
q
´
c
2
c
2
qx
ðx; lÞ ¼ H
3
ðlÞ^ vv
1
ðlÞ þ H
4
ðlÞ^ vv
2
ðlÞ; ERe l > 0; ð3:27Þ
where H
1
; H
2
; H
3
; H
4
are the functions introduced in Lemma 3.4. By applying
Lemma 3.4 we obtain the existence of M
3
; M
4
> 0 such that (3.21) and (3.22)
hold. 9
Remark 3.6. The inequality
kc
2
ðx; ÁÞk
2
H
1
ð0; TÞ
aCðkv
1
k
2
L
2
ð0; TÞ
þ kv
2
k
2
L
2
ð0; TÞ
Þ
can also be obtained by semigroup techniques combined with a trace theorem.
The next result will be used in order to reduce the proof of Proposition 3.2 to a
regularity property for a string equation.
Lemma 3.7. Let R be the operator defined in (2.9). Then the linear operator
L ¼ ÀR
q
4
qx
4
_ _
À
q
2
qx
2
is bounded from V onto V, i.e. we have that L A LðVÞ.
Proof. Let u A V. Equivalently u can be written as
uðxÞ ¼

nb1
a
n
sinðnxÞ;
240 K. Ammari, Z. Liu, and M. Tucsnak
with

nb1
n
4
a
2
n
< y. Then
Lu ¼ À
d
2
u
dx
2
ÀR
d
4
u
dx
4
_ _
¼

nb1
n
2
n
2
þ 1
a
n
sinðnxÞ;
which clearly implies that Lu A V and that kLuk
V
akuk
V
, which yields the con-
clusion of the lemma. 9
Proof of Proposition 3.2. Denote
DðA
0
Þ ¼
u
v
_ _
A ½H
3
ð0; pÞ XH
1
0
ð0; pފ  V;
d
2
u
dx
2
ð0Þ ¼
d
2
u
dx
2
ðpÞ ¼ 0
_ _
: ð3:28Þ
The corresponding operator A
0
is defined by
A
0
u
v
_ _
¼
v
ÀR
d
4
u
dx
4
_ _
_
_
_
_
; E
u
v
_ _
A DðA
0
Þ;
where R is defined by (2.9).
We use the method of transposition. Let DðA
0
Þ be the space defined in (3.28)
and denote by D½ðA
0
ފ
Ã
the dual space of DðA
0
Þ with respect to the pivot space
V Â H
1
0
ð0; pÞ. It is well known that A
0
can be extended to a skew-adjoint opera-
tor (denoted also by A
0
),
A
0
: V Â H
1
0
ð0; pÞ ! ½DðA
0
ފ
Ã
;
such that A
0
generates a group of isometries on ½DðA
0
ފ
Ã
, denoted by SðtÞ.
We define the operator
B: R
2
! ½DðA
0
ފ
Ã
; B
k
1
k
2
_ _
¼
0
k
1
Rðd
x
Þ À k
2
R
dd
x
dx
_ _
_
_
_
_
; E
k
1
k
2
_ _
A R
2
:
Then the problem (3.1)–(3.3) can be written as a Cauchy problem in ½DðA
0
ފ
Ã
:
q
qt
cðtÞ
qc
qt
ðtÞ
_
_
_
_
¼ A
0
cðtÞ
qc
qt
ðtÞ
_
_
_
_
À B
v
1
ðtÞ
v
2
ðtÞ
_ _
; Et > 0; ð3:29Þ
cð0Þ ¼
qc
qt
ð0Þ ¼ 0: ð3:30Þ
After a simple calculation we get that the operator B
Ã
: DðA
Ã
0
Þ !R
2
is given by
B
Ã
u
v
_ _
¼
vðxÞ
dv
dx
ðxÞ
_
_
_
_
; E
u
v
_ _
A DðA
Ã
0
Þ:
The relation above and that fact that A
0
is skew-adjoint imply that
B
Ã
S
Ã
ðtÞ
u
0
u
1
_ _
¼
qj
qt
ðx; tÞ
q
2
j
qx qt
ðx; tÞ
_
_
_
_
_
_
_
_
; E
u
0
u
1
_ _
A DðA
0
Þ; ð3:31Þ
Decay Rates for a Beam with Pointwise Force and Moment Feedback 241
with j satisfying (3.6)–(3.8). From (3.9) and (3.31) we deduce that there exists a
constant C > 0 (depending only on T) such that
_
T
0
B
Ã
S
Ã
ðtÞ
u
0
u
1
_ _ _
_
_
_
_
_
_
_
2
R
2
dt aCkðu
0
; u
1
Þk
2
VÂH
1
0
ð0; pÞ
; E
u
0
u
1
_ _
A DðA
0
Þ: ð3:32Þ
According to Theorem 3.1 on p. 173 in [5], inequality (3.32) implies that (3.29)
and (3.30) admit a unique finite energy solution. Moreover, this solution depends
continuously on v
1
; v
2
, i.e. there exsits a constant C > 0 (depending only on T)
such that
c
qc
qt
_
_
_
_
_
_
_
_
_
_
_
_
Cð0; T; VÂH
1
0
ð0; pÞÞ
aðkv
1
k
L
2
ð0; TÞ
þ kv
2
k
L
2
ð0; TÞ
Þ: ð3:33Þ
We have thus proved the conclusion (3.4).
We prove now the trace regularity property (3.5). According to the expression
of R, we have
R
q
4
qx
4
_ _
¼ À
q
2
qx
2
À L;
where L A LðVÞ (see Lemma 3.7). We notice that c can be written as
c ¼ c
1
þ c
2
; ð3:34Þ
where c
2
satisfies (3.16)–(3.18) and c
1
is the solution of
q
2
c
1
qt
2
À
q
2
c
1
qx
2
¼ Lc; 0 < x < p; t > 0; ð3:35Þ
c
1
ð0; tÞ ¼ c
1
ðp; tÞ ¼ 0; t > 0; ð3:36Þ
c
1
ðx; 0Þ ¼ 0;
qc
1
qt
ðx; 0Þ ¼ 0; 0 < x < p: ð3:37Þ
Since Lc A Cð0; T; VÞ, then by the classical theory for evolution equations of hy-
perbolic type [13], we obtain that c
1
A Cð0; T; H
3
ð0; pÞÞ XC
1
ð0; T; H
2
ð0; pÞÞ and
that there exists a constant C ¼ C
T
> 0 such that
c
1
qc
1
qt
_
_
_
_
_
_
_
_
_
_
_
_
Cð0; T; H
3
ð0; pÞÂH
2
ð0; pÞÞ
aC
T
kck
Cð0; T; VÞ
:
This inequality, combined with (3.33) and the standard trace theorem, implies
that
kc
1
ðx; ÁÞk
2
H
1
ð0; TÞ
þ
qc
1
qx
ðx; ÁÞ
_
_
_
_
_
_
_
_
2
H
1
ð0; TÞ
aCðkv
1
k
2
L
2
ð0; TÞ
þ kv
2
k
2
L
2
ð0; TÞ
Þ;
for some constant C depending only on T. The last inequality, combined with
(3.19) and (3.34), implies the conclusion (3.5). 9
242 K. Ammari, Z. Liu, and M. Tucsnak
4. Observability Inequalities
The observability inequality concerning the solutions of (3.6)–(3.8) is given in the
next proposition.
Proposition 4.1. Let T > 0 be fixed. Then the following assertions hold:
1. There exists a constant C
x
> 0 such that all finite energy solutions j of (3.6)–
(3.8) satisfy
_
T
0
qj
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
j
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt bC
x
ðku
0
k
2
H
2
ð0; pÞ
þ ku
1
k
2
H
1
ð0; pÞ
Þ ð4:1Þ
if and only if x=p is a rational number with coprime factorization
x
p
¼
p
q
; where q is odd:
2. There exists a constant C > 0 (independent of x) such that all finite energy
solutions j of (3.6)–(3.8) satisfy
_
T
0
qj
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
j
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt bCðku
0
k
2
H
1
ð0; pÞ
þ ku
1
k
2
L
2
ð0; pÞ
Þ:
Proof. First step. From (3.11) and the Ball–Slemrod generalization of Ingham’s
inequality (see [4]), we obtain that, for all T > 2p, there exists a constant C
T
such
that
_
T
0
qj
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
j
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt
bC
T

kb1
a
2
k
k
4
1 þ k
2
þ b
2
k
_ _
½k
2
cos
2
ðkxÞ þ sin
2
ðkxފ; ð4:2Þ
where ða
k
Þ and ðb
k
Þ have beeen defined by (3.10). If x satisfies (2.12), then, by
Lemma 3.4 in [20], there exists a constant k
x
> 0 such that
jcosðkxÞj bk
x
; Ek b1: ð4:3Þ
Inequalities (4.2) and (4.3) imply that
_
T
0
qj
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
j
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt bC

kb1
ðk
4
a
2
k
þ k
2
b
2
k
Þ;
for some constant C > 0. It follows that (4.1) holds for all x satisfying (2.12).
On the other hand if x does not satisfy (2.12), then we can again apply Lemma
3.4 from [20] to get the existence of a sequence ðp
m
Þ HN
Ã
, lim
m!y
p
m
¼ y such
that
lim
m!y
cosðp
m
xÞ ¼ 0: ð4:4Þ
Decay Rates for a Beam with Pointwise Force and Moment Feedback 243
If we denote by j
m
the solution of (3.6)–(3.7) with initial data
j
m
ðx; 0Þ ¼ sinðp
m
xÞ;
qj
m
qt
ðx; 0Þ ¼ 0; Ex A ð0; pÞ;
a simple calculation using (4.4) implies that
lim
m!y
_
T
0
fjðqj
m
=qtÞðx; tÞj
2
þ jðq
2
j
m
=qxqtÞðx; tÞj
2
g dt
kj
m
ð0Þk
2
H
2
ð0; pÞ
þ kðqj
m
=qtÞð0Þk
2
H
1
ð0; pÞ
¼ 0:
So, (4.1) is false for any x not satisfying (2.12).
Second step. In order to prove assertion 2 of Proposition 4.1 it is su‰cient to
notice that
_
T
0
qj
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
j
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt
bC
T

kb1
a
2
k
k
4
1 þ k
2
þ b
2
k
_ _
½k
2
cos
2
ðkxÞ þ sin
2
ðkxފ
bC
T

kb1
a
2
k
k
4
1 þ k
2
þ b
2
k
_ _
bC
T

kb1
½a
2
k
k
2
þ b
2
k
Š: 9
5. Proof of Proposition 2.3 and Theorem 2.5
Proof of Proposition 2.3. As we already mentioned in Section 2, the existence
and uniqueness of finite energy solutions of (1.1)–(1.3) was already proved in the
literature (see, for instance, [22]).
In order to prove the estimate (2.3) and the trace regularity property (2.5), it
su‰ces to remark that they hold for regular solutions ði.e.
u
qu=qt
_ _
A Cð0; T; DðAÞ)Þ
and to use the density of DðAÞ in V Â H
1
0
ð0; pÞ.
Finally, since the embedding of DðAÞ in V Â H
1
0
ð0; pÞ is obviously compact, by
LaSalle’s invariance principle (see, for instance, [10]) the strong stability estimate
at the end of Proposition 2.3 holds if the relations
q
2
v
qt
2
À
q
4
v
qx
2
qt
2
þ
q
4
v
qx
4
¼ 0; 0 < x < p; t > 0;
vð0; tÞ ¼ vðp; tÞ ¼
q
2
v
qx
2
ð0; tÞ ¼
q
2
v
qx
2
ðp; tÞ ¼ 0; t > 0;
qv
qt
ðx; tÞ ¼
q
2
v
qt qx
ðx; tÞ ¼ 0; t > 0;
imply that v 10.
244 K. Ammari, Z. Liu, and M. Tucsnak
In order to prove this unique continuation result, we put
vðx; 0Þ ¼

kb1
a
k
sinðkxÞ;
qv
qt
ðx; 0Þ ¼

kb1
b
k
sinðkxÞ; ð5:1Þ
with ðk
2
a
k
Þ; ðkb
k
Þ A l
2
ðRÞ. We have
qv
qt
ðx; tÞ ¼

kb1
Àa
k
k
2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ k
2
p sin
k
2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ k
2
p t
_ _
þ b
k
cos
k
2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ k
2
p t
_ _ _ _
sinðkxÞ ð5:2Þ
and
q
2
v
qt qx
ðx; tÞ ¼

kb1
Àka
k
k
2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ k
2
p sin
k
2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ k
2
p t
_ _
þ kb
k
cos
k
2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þ k
2
p t
_ _ _ _
cosðkxÞ;
ð5:3Þ
which implies, according to the Ball–Slemrod generalization of Ingham’s inequal-
ity, that for T > 2p,
_
T
0
qv
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
v
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt bC

kb1
a
2
k
k
4
1 þ k
2
þb
2
k
_ _
ðk
2
cos
2
ðkxÞ þsin
2
ðkxÞÞ
bC

kb1
a
2
k
k
4
1 þ k
2
þb
2
k
_ _
: ð5:4Þ
Hence, vðx; 0Þ ¼ 0 and ðqv=qtÞðx; 0Þ ¼ 0 for all x A ½0; pŠ which implies that v 10:
Thus we proved the strong stability result. 9
Let u A Cð0; T; VÞ XC
1
ð0; T; H
1
0
ð0; pÞÞ be the solution of (1.1)–(1.3). Then u
can be written as
u ¼ u
1
þ u
2
; ð5:5Þ
where u
1
is the solution of (3.6)–(3.8) and u
2
satisfies
q
2
u
2
qt
2
À
q
4
u
2
qx
2
qt
2
þ
q
4
u
2
qx
4
þ
qu
qt
ðx; tÞd
x
À
q
2
u
qt qx
ðx; tÞ
dd
x
dx
¼ 0 in ð0; pÞ Â ð0; TÞ;
ð5:6Þ
u
2
ð0; tÞ ¼ u
2
ðp; tÞ ¼
q
2
u
2
qx
2
ð0; tÞ ¼
q
2
u
2
qx
2
ðp; tÞ ¼ 0; t A ð0; TÞ; ð5:7Þ
u
2
ðx; 0Þ ¼
qu
2
qt
ðx; 0Þ ¼ 0; x A ð0; pÞ: ð5:8Þ
We note by Proposition 2.3 that ðq
2
u=qxqtÞðx; ÁÞ A L
2
ð0; TÞ, so (5.6) makes sense.
The main ingredient of the proof of Theorem 2.5 is the following result.
Lemma 5.1. Suppose that ðu
0
; u
1
Þ A V Â H
1
0
ð0; pÞ. Then the solutions u of
Decay Rates for a Beam with Pointwise Force and Moment Feedback 245
(1.1)–(1.3) and the solution u
1
of (3.6)–(3.8) satisfy
C
1
_
T
0
qu
1
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
1
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt
a
_
T
0
qu
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt
a4
_
T
0
qu
1
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
1
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt; ð5:9Þ
where C
1
> 0 is a constant independent of
u
0
u
1
_ _
.
Proof. First, we note by (3.9) that ðq
2
u
1
=qtqxÞðx; ÁÞ A L
2
ð0; TÞ. Relation (5.5)
implies that
_
T
0
qu
1
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
1
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt a2
_
_
T
0
qu
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt
þ
_
T
0
qu
2
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
2
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt
_
:
The estimate above combined with inequality (3.5) in Proposition 3.2 implies the
existence of a constant C
1
> 0, independent of
u
0
u
1
_ _
, such that
C
1
_
T
0
qu
1
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
1
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt a
_
T
0
qu
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt:
ð5:10Þ
On the other hand, (5.6) can be rewritten as
q
2
u
2
qt
2
À
q
4
u
2
qx
2
qt
2
þ
q
4
u
2
qx
4
þ
qu
2
qt
ðx; tÞd
x
À
q
2
u
2
qt qx
ðx; tÞ
dd
x
dx
¼ À
qu
1
qt
ðx; tÞd
x
þ
q
2
u
1
qt qx
ðx; tÞ
dd
x
dx
in ð0; pÞ Â ð0; TÞ: ð5:11Þ
If we formally take the duality product (with respect to the pivot space L
2
ð0; pÞ)
of both sides of (5.11) by qu
2
=qt (this can be done rigorously by considering a
regularizing sequence) and then we integrate from 0 to T we obtain that
1
2
qu
2
qt
ðÁ ; TÞ
_
_
_
_
_
_
_
_
2
L
2
ð0; pÞ
þ
1
2
q
2
u
2
qx qt
ðÁ ; TÞ
_
_
_
_
_
_
_
_
2
L
2
ð0; pÞ
þ
1
2
q
2
u
2
qx
2
ðÁ ; TÞ
_
_
_
_
_
_
_
_
2
L
2
ð0; pÞ
þ
_
T
0
qu
2
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
2
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt
¼ À
_
T
0
qu
1
qt
ðx; tÞ
qu
2
qt
ðx; tÞ dt À
_
T
0
q
2
u
1
qt qx
ðx; tÞ
q
2
u
2
qt qx
ðx; tÞ dt:
246 K. Ammari, Z. Liu, and M. Tucsnak
The last relation implies that
_
T
0
qu
2
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
2
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt

_
T
0
qu
1
qt
ðx; tÞ
qu
2
qt
ðx; tÞ dt À
_
T
0
q
2
u
1
qt qx
ðx; tÞ
q
2
u
2
qt qx
ðx; tÞ dt;
which obviously yields
_
T
0
qu
2
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
2
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt a
_
T
0
qu
1
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
1
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt:
Combining (5.5) with the inequality above, we obtain
_
T
0
qu
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt a4
_
T
0
qu
1
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
1
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt:
ð5:12Þ
Therefore, inequality (5.9) follows from (5.10) and (5.12). 9
We are now ready to prove the main results.
Proof of Theorem 2.5. 1. Proof of the first assertion. The energy EðuðtÞÞ of a fi-
nite energy solution of (1.1)–(1.3) decays exponentially if and only if there exists
T > 0 and C > 0 (depending on T) such that
Eðuð0ÞÞ À EðuðTÞÞ bCEðuð0ÞÞ; E
u
0
u
1
_ _
A V Â H
1
0
ð0; pÞ: ð5:13Þ
By using (2.3), the relation above can be rewritten
_
T
0
qu
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt bCEðuð0ÞÞ; E
u
0
u
1
_ _
A V Â H
1
0
ð0; pÞ:
By considering again the decomposition u ¼ u
1
þ u
2
, introduced in (5.5) and by
using Lemma 5.1, the above inequality is equivalent to
_
T
0
qu
1
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
1
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt bCEðuð0ÞÞ; E
u
0
u
1
_ _
A V Â H
1
0
ð0; pÞ:
By Proposition 4.1 the inequality above holds if and only if x=p is a rational
number with coprime factorization
x
p
¼
p
q
;
where q is odd. Thus we have proved the first assertion of Theorem 2.5.
Decay Rates for a Beam with Pointwise Force and Moment Feedback 247
2. Proof of the second assertion. By Proposition 4.1 and Lemma 5.1, the
solution u of (1.1)–(1.3) satisfies the inequality
_
T
0
qu
qt
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
q
2
u
qt qx
ðx; tÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
_ _
dt bK
1
u
0
u
1
_ _ _
_
_
_
_
_
_
_
2
H
1
ð0; pÞÂL
2
ð0; pÞ
;
E
u
0
u
1
_ _
A V Â H
1
0
ð0; pÞ;
for some constant K
1
> 0. By using (2.3) we see that the left-hand side of the last
inequality is equal to Eðuð0ÞÞ À EðuðTÞÞ, so we obtain that there exists a constant
K
1
such that, for all
u
0
u
1
_ _
A V Â H
1
0
ð0; pÞ we have
uðTÞ
u
0
ðTÞ
_ _ _
_
_
_
_
_
_
_
2
VÂH
1
0
ð0; pÞ
a
u
0
u
1
_ _ _
_
_
_
_
_
_
_
2
VÂH
1
ð0; pÞ
À K
1
u
0
u
1
_ _ _
_
_
_
_
_
_
_
2
H
1
ð0; pÞÂL
2
ð0; pÞ
; ð5:14Þ
where we denote by
0
the derivative with respect to time.
By using a simple interpolation inequality (see p. 49 of [16]), the fact that,
according to (2.3), the function
t !
uðtÞ
u
0
ðtÞ
_ _ _
_
_
_
_
_
_
_
2
VÂH
1
0
ð0; pÞ
is nonincreasing and by relation (5.14) we obtain the existence of a constant K
2
> 0
such that
uðTÞ
u
0
ðTÞ
_ _ _
_
_
_
_
_
_
_
2
VÂH
1
0
ð0; pÞ
a
u
0
u
1
_ _ _
_
_
_
_
_
_
_
2
VÂH
1
0
ð0; pÞ
À K
2
uðTÞ
u
0
ðTÞ
_ _ _
_
_
_
_
_
_
_
4
VÂH
1
0
ð0; pÞ
u
0
u
1
_ _ _
_
_
_
_
_
_
_
2
Y
; ð5:15Þ
where Y is given by (2.7).
We follow now the method used in [21]. The estimate (5.15) remains valid in
successive intervals ½kT; ðk þ 1ÞTŠ, so we have
uððk þ 1ÞTÞ
u
0
ððk þ 1ÞTÞ
_ _ _
_
_
_
_
_
_
_
2
VÂH
1
0
ð0; pÞ
a
uðkTÞ
u
0
ðkTÞ
_ _ _
_
_
_
_
_
_
_
2
VÂH
1
0
ð0; pÞ
À K
2
uððk þ 1ÞTÞ
u
0
ððk þ 1ÞTÞ
_ _ _
_
_
_
_
_
_
_
4
VÂH
1
0
ð0; pÞ
uðkTÞ
u
0
ðkTÞ
_ _ _
_
_
_
_
_
_
_
2
Y
:
Since A generates a semigroup of contractions in DðAÞ and the graph norm on
DðAÞ is equivalent to k Á k
Y
, the relation above implies the existence of a constant
248 K. Ammari, Z. Liu, and M. Tucsnak
K
3
> 0 such that
uððk þ 1ÞTÞ
u
0
ððk þ 1ÞTÞ
_ _ _
_
_
_
_
_
_
_
2
VÂH
1
0
ð0; pÞ
a
uðkTÞ
u
0
ðkTÞ
_ _ _
_
_
_
_
_
_
_
2
VÂH
1
0
ð0; pÞ
ÀK
3
uððk þ 1ÞTÞ
u
0
ððk þ 1ÞTÞ
_ _ _
_
_
_
_
_
_
_
4
VÂH
1
0
ð0; pÞ
u
0
u
1
_ _ _
_
_
_
_
_
_
_
2
Y
; E
u
0
u
1
_ _
A DðAÞ:
ð5:16Þ
If we adopt now the notation
E
k
¼
uðkTÞ
u
0
ðkTÞ
_ _ _
_
_
_
_
_
_
_
2
VÂH
1
0
ð0; pÞ
u
0
u
1
_ _ _
_
_
_
_
_
_
_
2
Y
; ð5:17Þ
relation (5.16) gives
E
kþ1
aE
k
À K
3
E
2
kþ1
; Ek b0: ð5:18Þ
By applying Lemma 1.2 from [21] and using relation (5.18) we obtain the
existence of a constant M > 0 such that
uðkTÞ
u
0
ðkTÞ
_ _ _
_
_
_
_
_
_
_
2
VÂH
1
0
ð0; pÞ
a
M
u
0
u
1
_ _ _
_
_
_
_
_
_
_
2
Y
k þ 1
; Ek b0;
which leads to (2.13). 9
6. Proof of Proposition 2.4 and Theorem 2.6
Proof of Proposition 2.4. Define
H¼ V Â L
2
ð0; pÞ
equipped with the inner product
y
1
v
1
_ _
;
y
2
v
2
_ _ _ _
H
¼
_
p
0
pðxÞ
d
2
y
1
dx
2
d
2
y
2
dx
2
þ rðxÞv
1
v
2
_ _
dx: ð6:1Þ
Furthermore, we define an operator Ain Hby
DðAÞ ¼
y
v
_ _
A H
y; v A V; pðxÞ
d
2
y
dx
2
A H
2
ð0; xÞ XH
2
ðx; pÞ
d
2
y
dx
2
ð0Þ ¼
d
2
y
dx
2
ðpÞ ¼ 0;
d
dx
p
d
2
y
dx
2
_ _ _ _
x
¼ ÀvðxÞ
p
d
2
y
dx
2
_ _ _ _
x
¼
dv
dx
ðxÞ
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
¸
_
_
¸
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
¸
_
;
ð6:2Þ
Decay Rates for a Beam with Pointwise Force and Moment Feedback 249
Aðy; vÞ ¼ v; À
1
rðxÞ
d
2
dx
2
pðxÞ
d
2
y
dx
2
_ _
À vðxÞd
x
þ
dv
dx
ðxÞ
dd
x
dx
_ _
; ð6:3Þ
where the derivatives with respect to x are calculated in D
0
ð0; pÞ.
The existence and uniqueness of finite energy solutions of (1.4)–(1.6) can be
obtained by standard semigroup methods.
In order to prove the estimate (2.4), it su‰ces to note that it holds for solutions
u
qu
qt
_ _
A Cð0; T; DðAÞÞ
and to use the density of DðAÞ in V Â L
2
ð0; pÞ. 9
Lemma 6.1. The spectrum of Acontains no point on the imaginary axis.
Proof. Since Ahas compact resolvent, its spectrum sðAÞ only consists of eigen-
values of A. We will show that the equation
Az ¼ ibz ð6:4Þ
with z ¼
y
v
_ _
A DðAÞ and b 00 has only the trivial solution.
By taking the inner product of (6.4) with z A Hand using
RehAz; zi
H
¼ ÀjvðxÞj
2
À
dv
dx
ðxÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
; ð6:5Þ
we obtain that vðxÞ ¼ ðdv=dxÞðxÞ ¼ 0. Next, we eliminate v in (6.4) to get a fourth-
order ordinary di¤erential equation with variable coe‰cients:
d
2
dx
2
pðxÞ
d
2
y
dx
2
_ _
À b
2
rðxÞy ¼ 0;
yð0Þ ¼ yðpÞ ¼
d
2
y
dx
2
ð0Þ ¼
d
2
y
dx
2
ðpÞ ¼ 0;
yðxÞ ¼
dy
dx
ðxÞ ¼ 0:
_
¸
¸
¸
¸
¸
¸
¸
_
¸
¸
¸
¸
¸
¸
¸
_
ð6:6Þ
By Lemma 2.3 in [15] the above system only has a trivial solution. 9
Proof of Theorem 2.6. By a classical result (see [12] and [19]) it su‰ces to show
that Asatisfies the following two conditions:
rðAÞ Ifib j b A Rg 1iR ð6:7Þ
and
limsup
jbj!y
kðib ÀAÞ
À1
k < y; ð6:8Þ
where rðAÞ denotes the resolvent set of the operator A.
By Lemma 6.1 the condition (6.7) is satisfied. Suppose that the condition (6.8)
is false. By the Banach–Steinhaus theorem (see [6]), there exist a sequence of real
250 K. Ammari, Z. Liu, and M. Tucsnak
numbers b
n
! y and a sequence of vectors z
n
¼
y
n
v
n
_ _
A DðAÞ with kz
n
k
H
¼ 1
such that
kðib
n
I ÀAÞz
n
k
H
! 0 as n ! y; ð6:9Þ
i.e.
ib
n
y
n
À v
n
1 f
n
! 0 in V; ð6:10Þ
ib
n
rðxÞv
n
þ
d
2
dx
2
pðxÞ
d
2
y
n
dx
2
_ _
1g
n
! 0 in L
2
ð0; pÞ: ð6:11Þ
Our goal is to derive from (6.9) that kz
n
k
H
converges to zero, thus, a contradiction.
The proof is divided into four steps:
First step. We notice that from (6.5) we have
kðib
n
I ÀAÞz
n
k
H
bjRehðib
n
I ÀAÞz
n
; z
n
i
H
j ¼ jv
n
ðxÞj
2
þ
dv
n
dx
ðxÞ
¸
¸
¸
¸
¸
¸
¸
¸
2
: ð6:12Þ
Then, by (6.9),
jv
n
ðxÞj ! 0;
dv
n
dx
ðxÞ
¸
¸
¸
¸
¸
¸
¸
¸
! 0: ð6:13Þ
This further leads to
jb
n
y
n
ðxÞj ! 0; b
n
dy
n
dx
ðxÞ
¸
¸
¸
¸
¸
¸
¸
¸
! 0; ð6:14Þ
due to (6.10) and the trace theorem.
Second step. We now express v
n
as a function of y
n
from (6.10) and substitute it
into (6.11) to get
Àb
2
n
ry
n
þ
d
2
dx
2
p
d
2
y
n
dx
2
_ _
¼ g
n
þ ib
n
f
n
: ð6:15Þ
Next, we take the inner product of (6.15) with qðxÞðdy
n
=dxÞ in L
2
ð0; xÞ where
qðxÞ A C
2
ð½0; xŠÞ and qð0Þ ¼ 0. We obtain that
_
x
0
Àb
2
n
ry
n
þ
d
2
dx
2
p
d
2
y
n
dx
2
_ _ _ _
qðxÞ
dy
n
dx
dx
¼
_
x
0
ðg
n
þ ib
n
f
n
ÞqðxÞ
dy
n
dx
dx
¼
_
x
0
g
n
qðxÞ
dy
n
dx
dx À
_
x
0
q
df
n
dx
b
n
y
n
dx
À
_
x
0
f
n
dq
dx
b
n
y
n
dx þ f
n
ðxÞqðxÞb
n
y
n
ðxÞ: ð6:16Þ
It is clear that the right-hand side of (6.16) converges to zero since f
n
; g
n
converge
to zero in H
2
and L
2
, respectively.
Decay Rates for a Beam with Pointwise Force and Moment Feedback 251
By a straightforward calculation,
Re
_
x
0
Àb
2
n
ry
n
q
dy
n
dx
dx
_ _
¼ À
1
2
rðxÞqðxÞjb
n
y
n
ðxÞj
2
þ
1
2
_
x
0
d
dx
ðrqÞjb
n
y
n
j
2
dx ð6:17Þ
and
Re
_
x
0
d
2
dx
2
p
d
2
y
n
dx
2
_ _
q
dy
n
dx
dx
_ _
¼ Re q
d
dx
p
d
2
y
n
dx
2
_ _
ðx
À
Þ À p
dq
dx
d
2
y
n
dx
2
ðx
À
Þ
_ _
dy
n
dx
ðxÞ
_ _
À
1
2
pðxÞqðxÞ
d
2
y
n
dx
2
ðx
À
Þ
¸
¸
¸
¸
¸
¸
¸
¸
2
þ
_
x
0
1
2
3p
dq
dx
À
dp
dx
q
_ _
d
2
y
n
dx
2
¸
¸
¸
¸
¸
¸
¸
¸
2
þ Re p
d
2
q
dx
2
d
2
y
n
dx
2
dy
n
dx
_ _
_ _
dx:
Notice from (6.11) that ðd
2
=dx
2
Þðpðd
2
y
n
=dx
2
ÞÞ=b
n
is bounded in L
2
ð0; xÞ
which further implies the boundedness of jðd=dxÞðpðd
2
y
n
=dx
2
ÞÞðx
À
Þj=b
n
and
j pðd
2
y
n
=dx
2
Þðx
À
Þj=b
n
. Since
_
x
0
dy
n
dx
¸
¸
¸
¸
¸
¸
¸
¸
2
dx ¼ À
1
ib
n
_
x
0
v
n
d
2
y
n
dx
2
dx À
1
ib
n
_
x
0
ðib
n
y
n
À v
n
Þ
d
2
y
n
dx
2
dx
þ b
n
dy
n
dx
ðxÞ
_ _
1
b
n
y
n
ðxÞ
_ _
;
then, from the boundedness of v
n
, ib
n
y
n
À v
n
, d
2
y
n
=dx
2
, in L
2
ð0; xÞ and (6.14), we
have dy
n
=dx converges to zero in L
2
ð0; xÞ. With these facts and the boundary con-
ditions of y
n
and (6.14) at hand, we simplify (6.16), then take its real parts. This
leads to
_
x
0
r
dq
dx
þ
dr
dx
q
_ _
jb
n
y
n
j
2
dx þ
_
x
0
3p
dq
dx
À
dp
dx
q
_ _
d
2
y
n
dx
2
¸
¸
¸
¸
¸
¸
¸
¸
2
dx
À pðxÞqðxÞ
d
2
y
n
dx
2
ðx
À
Þ
¸
¸
¸
¸
¸
¸
¸
¸
2
! 0: ð6:18Þ
Similarly, we take the inner product of (6.15) with q
1
ðxÞðdy
n
=dxÞ in L
2
ðx; pÞ with
q
1
A C
2
ð½x; pŠÞ and q
1
ðpÞ ¼ 0, then repeat the above procedure. This will give us
_
p
x
r
dq
1
dx
þ
dr
dx
q
1
_ _
jb
n
y
n
j
2
dx þ
_
p
x
3p
dq
1
dx
À
dp
dx
q
1
_ _
d
2
y
n
dx
2
¸
¸
¸
¸
¸
¸
¸
¸
2
dx
þ pðxÞq
1
ðxÞ
d
2
y
n
dx
2
ðx
þ
Þ
¸
¸
¸
¸
¸
¸
¸
¸
2
! 0: ð6:19Þ
Third step. Next, we show that both of jðd
2
y
n
=dx
2
Þðx
À
Þj and jðd
2
y
n
=dx
2
Þðx
þ
Þj
converge to zero. To proceed, we take the inner product of (6.15) with
252 K. Ammari, Z. Liu, and M. Tucsnak
ð1=j
n
Þe
Àj
n
hðxÞ
in L
2
ð0; xÞ, where j
n
¼
ffiffiffiffiffiffiffiffi
jb
n
j
_
, and
hðxÞ ¼
_
x
x
rðsÞ
pðsÞ
_ _
1=4
ds:
This leads to
À
_
x
0
j
3
n
y
n
re
Àj
n
h
dx þ
_
x
0
d
2
dx
2
p
d
2
y
n
dx
2
_ _
1
j
n
e
Àj
n
h
dx ! 0: ð6:20Þ
Performing integration by parts four times to the second term on the left-hand side
of (6.20), we obtain
_
x
0
d
2
dx
2
p
d
2
y
n
dx
2
_ _
1
j
n
e
Àj
n
h
dx
¼
1
j
n
d
dx
p
d
2
y
n
dx
2
_ _
þ p
d
2
y
n
dx
2
dh
dx
þ j
n
dy
n
dx
p
dh
dx
_ _
2
þ jb
n
j y
n
p
dh
dx
_ _
3
_ _
e
Àj
n
h
j
x
0
À
_
x
0
p
d
2
h
dx
2
d
2
y
n
dx
2
þ
d
dx
p
dh
dx
_ _
2
_ _
j
n
dy
n
dx
þ
d
dx
p
dh
dx
_ _
3
_ _
b
n
y
n
e
Àj
n
h
_ _
dx
þ
_
x
0
j
3
n
ðh
0
Þ
4
y
n
pe
Àj
n
h
dx: ð6:21Þ
Since e
Àj
n
h
converges to zero in L
2
ð0; xÞ, the first integral term on the right-hand
side of (6.21) will also converge to zero. We then substitute (6.21) into (6.20) to
cancel the other inner product term. Among the remaining boundary terms, those
at x ¼ 0 all converge to zero due to the exponential decay term e
Àj
n
hð0Þ
and the
boundedness of ðd
2
=dx
2
Þðpðd
2
y
n
=dx
2
Þ=b
n
Þ in L
2
ð0; xÞ; those containing y
n
ðxÞ,
ðdy
n
=dxÞðxÞ also converge to zero due to (6.14). Thus, we simplify (6.20) to
1
j
n
d
dx
p
d
2
y
n
dx
2
_ _
ðx
À
Þ À pðxÞ
rðxÞ
pðxÞ
_ _
1=4
d
2
y
n
dx
2
ðx
À
Þ ! 0: ð6:22Þ
Similarly, we take the inner product of (6.15) with ð1=j
n
Þe
Àj
n
h
1
ðxÞ
in L
2
ðx; pÞ where
h
1
ðxÞ ¼
_
x
x
rðsÞ
pðsÞ
_ _
1=4
ds:
Repeating the above analysis yields
1
j
n
d
dx
p
d
2
y
n
dx
2
_ _
ðx
þ
Þ þ pðxÞ
rðxÞ
pðxÞ
_ _
1=4
d
2
y
n
dx
2
ðx
þ
Þ ! 0: ð6:23Þ
Recall that
d
dx
p
d
2
y
n
dx
2
_ _
ðx
þ
Þ À
d
dx
d
2
y
n
dx
2
_ _
ðx
À
Þ ¼ Àv
n
ðxÞ
converges to zero. Then the di¤erence of (6.22) and (6.23) leads to
d
2
y
n
dx
2
ðx
þ
Þ þ
d
2
y
n
dx
2
ðx
À
Þ ! 0: ð6:24Þ
Decay Rates for a Beam with Pointwise Force and Moment Feedback 253
Also recall that
p
d
2
y
n
dx
2
_ _
ðx
þ
Þ À p
d
2
y
n
dx
2
_ _
ðx
À
Þ ¼
dv
n
dx
ðxÞ
converges to zero. Therefore, we have proved
d
2
y
n
dx
2
ðx
À
Þ ! 0;
d
2
y
n
dx
2
ðx
þ
Þ ! 0: ð6:25Þ
In view of (6.25), we simplify (6.18) and (6.19) to
_
x
0
r
dq
dx
þ
dr
dx
q
_ _
jb
n
y
n
j
2
dx þ
_
x
0
3p
dq
dx
À
dp
dx
q
_ _
d
2
y
n
dx
2
¸
¸
¸
¸
¸
¸
¸
¸
2
dx ! 0 ð6:26Þ
and
_
p
x
r
dq
1
dx
þ
dr
dx
q
1
_ _
jb
n
y
n
j
2
dx þ
_
p
x
3p
dq
1
dx
À
dp
dx
q
1
_ _
d
2
y
n
dx
2
¸
¸
¸
¸
¸
¸
¸
¸
2
dx ! 0: ð6:27Þ
Fourth step. Finally, we choose qðxÞ and q
1
ðxÞ so that rðdq=dxÞ þ ðdr=dxÞq
and 3pðdq=dxÞ À ðdp=dxÞq are strictly positive, and rðdq
1
=dxÞ þ ðdr=dxÞq
1
and
3pðdq
1
=dxÞ À ðdp=dxÞq
1
are strictly negative. This can be done by taking
qðxÞ ¼ e
ax
À 1; q
1
ðxÞ ¼ e
bðpÀxÞ
À 1; ð6:28Þ
with
a ¼ max
kdr=dxk
y
r
0
;
kdp=dxk
y
3p
0
_ _
; b ¼ min
kdr=dxk
y
r
0
;
kdp=dxk
y
3p
0
_ _
:
Therefore, (6.26) and (6.27) imply
kb
n
y
n
k
L
2
ð0; pÞ
! 0; ky
n
k
V
! 0: ð6:29Þ
In view of (6.10), we also get
kv
n
k
L
2
ð0; pÞ
! 0; ð6:30Þ
which clearly contradicts (6.9). 9
Remark 6.2. A di¤erent method, working only in the case rðxÞ ¼ pðxÞ ¼ 1, was
proposed in [1]. This method, combining ideas from [20] and [3], uses the permu-
tation of one of the inputs with the corresponding output of the system.
References
[1] K. Ammari, Asymptotic behavior of coupled Euler–Bernoulli beams with pointwise dissipation,
ESAIM Proceedings, 8 (2000), 1–12.
[2] K. Ammari and M. Tucsnak, Stabilization of Bernoulli–Euler beam by means of a pointwise
feedback force, SIAM J. Control. Optim., 39 (2000), 1160–1181.
[3] K. Ammari and M. Tucsnak, Stabilization of second order equations by a class of unbounded
feedbacks, ESAIM COCV, 6 (2001), 361–386.
[4] J. M. Ball and M. Slemrod, Nonharmonic Fourier series and the stabilization of semilinear
control systems, Comm. Pure Appl. Math., 32 (1979), 555–587.
254 K. Ammari, Z. Liu, and M. Tucsnak
[5] A. Bensoussan, G. Da Prato, M. C. Delfour and S. K. Mitter, Representation and Control of
Infinite Dimensional Systems, Vol. I, Birkha¨user, Basel, 1992.
[6] H. Brezis, Analyse Fonctionnelle, The´orie et Applications, Masson, Paris, 1983.
[7] G. Chen, M. C. Delfour, A. M. Krall and G. Payre, Modeling, stabilization and control of
serially connected beams, SIAM J. Control Optim., 25 (1987), 526–546.
[8] G. Chen, S. G. Krantz, D. W. Ma, C. E. Wayne and H. H. West, The Euler–Bernoulli beam
equation with boundary energy dissipation, in Operator Methods for Optimal Control Problems,
Sung J. Lee, ed., Marcel Dekker, New York, 1988, pp. 67–96.
[9] G. Chen, S. G. Krantz, D. L. Russell, C. E. Wayne, H. H. West and M. P. Coleman, Analysis,
design and behavior of dissipative joints for coupled beams, SIAM J. Control Optim., 28 (1990),
423–437.
[10] C. Dafermos, Asymptotic behavior of solutions of evolution equations, in Nonlinear Evolution
Equations, M. G. Crandall, ed., Academic Press, New York, 1978, pp. 103–123.
[11] G. Doetsch, Introduction to the Theory and Application of the Laplace Transformation, Springer-
Verlag, Berlin, 1974.
[12] F. Huang, Characteristic conditions for exponential stability of linear dynamical systems in
Hilbert space, Ann. Di¤erential Equations, 1 (1985), 43–56.
[13] I. Lasiecka, J.-L. Lions and R. Triggiani, Nonhomogeneous boundary value problems for second
order hyperbolic operators, J. Math. Pures Appl., 65 (1986), 149–192.
[14] P. D. Lax and R. S. Philips, Scattering theory for dissipative hyperbolic systems, J. Funct. Anal.,
14 (1973), 172–235.
[15] W. Leighton and Z. Nehari, On the oscillation of solutions of self-adjoint linear di¤erential
equations of the fourth order, Trans. Amer. Math. Soc., 89 (1958), 325–377.
[16] J. L. Lions and E. Magenes, Proble`mes aux Limites non Homoge´nes et Applications, Vol. 1,
Dunod, Paris, 1968.
[17] K. Liu and Z. Liu, Boundary stabilization of nonhomogeneous beam with rotatory inertia at the
tip, Comp. Appl. Math., 114 (2000), 1–10.
[18] A. Pazy, Semigroups of Linear Operators and Applications to Partial Di¤erential Equations,
Applied Mathematical Sciences, Vol. 44, Springer-Verlag, New York, 1983.
[19] J. Pru¨ ss, On the spectrum of C
0
-semigroups, Trans. Amer. Math. Soc., 248 (1984), 847–857.
[20] R. Rebarber, Exponential stability of beams with dissipative joints: a frequency domain approach,
SIAM J. Control. Optim., 33 (1995), 1–28.
[21] D. L. Russell, Decay rates for weakly damped systems in Hilbert space obtained with control
theoretic methods, J. Di¤erential Equations, 19 (1975), 344–370.
[22] G. Weiss, O. J. Sta¤ans and M. Tucsnak, Well-posed linear systems—a survey with emphasis on
conservative systems, Appl. Math. Comput. Sci., 11 (2001), 101–127.
Decay Rates for a Beam with Pointwise Force and Moment Feedback 255