Stability of solutions for some classes of

nonlinear damped wave equations

Genni Fragnelli
Dipartimento di Ingegneria dellInformazione,
Universit`a di Siena,
Via Roma 56, 53100 Siena - Italy
tel. +39 0577 234622, email: fragnelli@dii.unisi.it
Dimitri Mugnai
Dipartimento di Matematica e Informatica
Universit`a di Perugia
Via Vanvitelli 1, 06123 Perugia - Italy
tel. +39 075 5855043, e-mail: mugnai@dipmat.unipg.it

Abstract
We consider two classes of semilinear wave equations with nonnegative
damping which may be of type onoff or integrally positive. In both
cases we give a sufficient condition for the asymptotic stability of the
solutions. In the case of integrally positive damping we show that such a
condition is also necessary.

Keywords: damped nonlinear wave equations, integrally positive damping,

onoff damping.
2000AMS Subject Classification: 35L70, 93D20, 25B35.

Introduction

We are concerned with some classes of nonlinear abstract damped wave equations, whose prototype is the usual wave equation in a bounded domain
RN , N 1,

in (0, +) ,
utt = u h(t)ut + f (u)
u(t, x) = 0
in (0, +) ,
(1.1)

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

Research supported by the M.I.U.R. project Metodi Variazionali ed Equazioni Differenziali Nonlineari.

though we can handle equations in a more general Banach setting like u00 +
B(t)u0 + Au = f (u), with B and A suitably given (see Section 2 for the precise
setting).
This problem has been already investigated by many authors in the case of
ordinary differential equations or systems of ordinary differential equations (i.e.
when u depends only on t) when f is linear (for example see [4], [7]) and also
when f is nonlinear ([13], [15], [16]).
In the case of hyperbolic partial differential equations like (1.1), the problem
has been studied when f is linear ([3], [9],. . . ), when f is nonlinear but with
linear growth ([19],. . . ), and when f is nonlinear with superlinear growth ([11]
in the case of constant damping, [14] for more general cases).
Concerning the damping h, different assumptions are alternatively made:
on-off ([6]), increasing ([1]), bounded - in many sense - ([3], [17]), integrally
positive ([19]), etc. . . . In particular, onoff dampers are suitable to describe a
wide variety of communication network models (circuits which can be switched
on or off), as well as systems where a control depending on time is necessary.
Very interesting results in the special case f 0 and damping of type on-off
can be found in [6], also when the term a(t)ut in (1.1) is replaced by a(t)g(ut ),
where g is a nonlinear function with linear growth (see also [10]). For the
case f 0 we also mention a logarithmic decay estimate proved in [2] when
the term a(t)ut in (1.1) is replaced by (1 + t) a(x)g(ut ), with a bounded and
strictly positive on a subdomain of and g possibly having superlinear growth
at infinity.
In [19] the author shows that, if f has linear growth and h is integrally
positive (see Definition 1 below), then any solution u(u0 , u1 ) of (2.5) converges
to 0 in the norm kukL2 + kut kL2 if and only if
Z
Z t
H(t)
e
eH(s) ds dt = +,
(1.2)
0

Rt

where H(t) = 0 h(s)ds. Actually the proof contains some gaps, in particular in
proving that the L2 norm of u0 converges to 0 as t . However we recover
that result in Section 3.
In this paper we show that when h is integrally positive, condition (1.2) is
sufficient to prove global stability for problem (1.1), also when f is superlinear
and satisfies a sign condition. More precisely, in Theorem 3.1 we prove that
condition (1.2) is sufficient under the assumption
uf (u) 0

for every u R,

which was already assumed, for example, in [14] and in [17]. Note that such a
condition is trivially verified when f (u) = |u|p u, p > 0, case which was studied,
for example, in [11], where a global existence result is proved for h =constant.
Under a natural non supercritical growth condition on f , we also show that
(1.2) is also a necessary condition for global stability to hold (see Theorem 3.2).
We remark that the requirement uf (u) 0 is a bit stronger that sf (s) < 1 s2 ,
s 6= 0, which is essentially assumed in [19], and also in [8] when the damping
2

is concentrated in a subset of and Neumanntype conditions are assumed on

the boundary as usual, here 1 denotes the first eigenvalue of on H01 ().
For damping of type onoff, in [6] the following case is considered: let
(an , bn )n be a sequence of open disjoint intervals of (0, ) such that an
and suppose there exists Mn mn > 0 such that
mn h(t) Mn

t (an , bn ).

(1.3)

If f 0, then any solution u(u0 , u1 ) of (2.5) converges to 0 in the norm kukL2 +

kut kL2 if


mn (bn an ) min (bn an )2 ,

n=1

1
1 + mn M n


= .

(1.4)

In this result the fact that f 0 is essential in the proof of stability. In the
nonlinear case under consideration, with the assumption
Z s
sf (s)
f () d 0 s R,
0

we show that (1.4), which was essentially already introduced in [13] for systems of ordinary differential equations, is again sufficient for the stability (see
Theorem 4.1). As for the case f 0 in [6], we still dont know if (1.4) is also
necessary for stability to hold.
However, in the case of integrally positive damping, we a complete characterization of stability for signed non supercritical nonlinearities.

Acknowledgments
This work was started while the authors were visiting Prof. Laszlo Hatvani at
the Bolyay Institute of the University of Szeged. The first author acknowledges
financial support from GNAMPA. The second author acknowledges financial
support from CNR, in the framework of the bilateral project CNRMTA 132.07.
The authors wish to thank Prof. Hatvani for some interesting discussions on
this subject and for bringing [19] to their knowledge.

The abstract setting

We will use an abstract setting which is a bit less general than the one in [6], but
more natural for our setting. Let us consider a second order evolution problem
of the form
 00
t>0
u + B(t)u0 + Au = f (u)
(2.5)
u(0) = u0 V, u0 (0) = u1 H.
Here H denotes a real Hilbert space with scalar product h, iH and norm k kH ,
A : D(A) H H is a linear selfadjoint coercive operator on H with dense
3

domain and V = D(A1/2 ) with norm kvkV = kA1/2 vkH such that
V , H H 0 , V 0

(2.6)

with dense embeddings.

Concerning the timedependent operator B, in Section 3 we assume it is actually a nonnegative function which can be 0 in a set of measure 0 (see Definition
1 below), while for the results of Section 4 we let it be a positive nonlinear
operator (see below for the precise assumptions), for which we assume that
B L ((0, ); Lip(H, H 0 )).
Finally, on the nonlinearity f we assume alternatively
sf (s) 0
which implies
Z

s R,

f () d 0

F (s) :=

(2.7)

s R,

or
sf (s) F (s) 0

s R.

(2.8)

Remark 2.1. In both cases, f (u) = |u|p u, p > 0, is the prototype function.
By solution of (2.5), we mean a function u L2 (0, ; V ) such that u0
L (0, ; H) and
2

u00 L2 (0, ; V 0 ), Au L2 (0, ; V 0 )

Bu0 L2 (0, ; V 0 ), f (u) L2 (0, ; V 0 ),
with u(0) = u0 , u0 (0) = u1 , and such that
hu00 , iV 0 ,V + hBu0 , iV 0 ,V + hAu, iV 0 ,V = hf (u), iV 0 ,V

(2.9)

for every L2 (0, ; V ) and for a.e. t > 0.

Remark 2.2. In the case of problem (1.1) the condition f (u) L2 (0, ; V 0 )
is automatically satisfied when f (u) = |u|p u, p > 0 and V = H01 (), provided
N = 1, 2 or p 4/(N 2) if N 3.
For any solution u of problem (2.5) we denote by Eu , or simply by E if there
is no need to specify u, the energy associated to such a solution:
E(t) =

1
1 0
ku (t)k2H + ku(t)k2V dx F(u),
2
2

(2.10)

where F(u) is the realvalued functional such that F(0) = 0 and F 0 (u)() =
hf (u), iV 0 ,V . Of course in the case of problem (1.1) we have H = L2 (),
V = H01 () and
Z
F(u) =

F (u) dx.

The following result, proved in [6] when f 0, still holds in the nonlinear
case.
4

Lemma 2.1. For any solution u of (2.5) the associated energy Eu is absolutely
continuous on (0, ) and
Eu0 (t) = hB(t)u0 (t), u0 (t)iH

a.e. in (0, ).

(2.11)

In our setting we will also need the following obvious corollary.

Lemma 2.2. If in addition hB(t)w, wiH 0 for a.e. t 0 and for every
w H, then for any solution u of (2.5) we have that Eu is non increasing.
Remark 2.3. In the case of problem (1.1) equality (2.11) reads
Z
E 0 (t) =
h(t)u2t dx,

which is non positive if h 0 a.e. in .

The integrally positive case

Let us start with the following definitions.

Definition 1. A function h : [0, +) [0, +) is said integrally positive if for
every a > 0 there exists > 0 such that
Z t+a
h(s) ds
t > 0.
t

Definition 2. Solutions of (2.5) are said uniformly bounded in D(A) D(A1/2 )

if for any B1 > 0 there exists B2 > 0 such that
if (u0 , u1 ) D(A) D(A1/2 ) and kAu0 kH + kA1/2 u1 kH B1 ,
and if u(t, u0 , u1 ) denotes the solution of (2.5) with initial condition (u0 , u1 ),
then t > 0
f (u(t, u0 , u1 )) H and
kAu(t, u0 , u1 )kH + kA1/2 u0 (t, u0 , u1 )kH + kf (u(t, u0 , u1 ))kH B2 .
Let us remark that such a definition is a natural modification of the one
introduced in [19], due to the presence of the requirement on kf (u)kH .
For the following result we concentrate on (1.1), where A = , D(A) =
H 2 () H01 (), D(A1/2 ) = H01 () and H = L2 (), and we recall that kuk2
is a norm on H 2 () H01 () which is equivalent to the usual one (for example,
see [12]). After setting k k2 = k kL2 () , we can state our first fundamental
result.
Theorem 3.1. Assume (1.2) and (2.7). If h is integrally positive and solutions
of (1.1) are uniformly bounded in H 2 () H01 () H01 (), then every solution
of (1.1) satisfies
kDu(t)k2 + ku0 (t)k2 0
5

as t +.

(3.12)

Proof. By Remark (2.3) there exists E 0 such that



Z
1 0
1
2
lim E(t) = lim
ku (t)k2 + kDu(t)k2 dx
F (u) dx = E . (3.13)
t+
t+ 2
2

We want to show that

R s E = 0, so let us assume by contradiction that E > 0.
By (2.7) F (s) = 0 f ( ) d 0 for any s R, so that by (3.13) there exists
L [0, 2E ] such that
lim sup ku0 (t)k22 = L.
(3.14)
t+

Assume by contradiction that L > 0. We must distinguish several cases.

First case: ku0 (t)k22 L t > 0. Then in this way, by (2.11) and Remark 2.3
we get
Z
Z
0
0 < E = E(0) +
E ( ) d = E(0) 2
h( )ku0 ( )k22 d
0 Z
0
(3.15)
h( ) d.
= E(0) 2L
0

Since h is integrally positive, there exists > 0 such that

Z n+1
h( ) d
n N.

(3.16)

Therefore, (3.15) and (3.16) imply

0 < E(0) 2L

= ,

n1

and a contradiction arises.

Second case: ku0 (t)k22 6= L. Set lim inf t+ ku0 (t)k22 = ` [0, L], and assume
first that ` < L. Then there exist two sequences (sn )n and (tn )n such that
1. 0 < sn < tn < sn+1 n N;
2. sn + as n ;
3.

3L + `
L+`
= ku0 (sn )k22 < ku0 (tn )k22 =
2
4

4.

L
2L
ku0 (t)k22
3
3

n N;

t (sn , tn ).

Since solutions of (1.1) are uniformly bounded, there exists M > 0 such that
d 0
ku (t)k22
dt

= 2hu0 (t), u00 (t)i

= 2hu0 (t), Au(t)i 2h(t)ku0 (t)k22 + 2hu0 (t), f (u)i
2ku0 (t)k2 kAu(t)k2 + 2ku0 (t)k2 kf (u)k2 M.
6

Therefore
L`
3L + ` L + `
=

=
4
4
2

tn

sn

d 0
ku (t)k22 dt M (tn sn ),
dt

so that

L`
n N.
4M
In this way, by (2.11) and (3.17) we get
Z
Z
0 < E = E(0) +
E 0 ( ) d E(0) 2
tn sn

(3.17)

h( )ku0 ( )k22 d

n (sn ,tn )

(3.18)
2L
E(0)
3

Z
h( ) d.
n (sn ,sn + L`
4M )

Since h is integrally positive, there exists > 0 such that

Z
h( ) d n N,
(sn ,sn + L`
4M )

so that (3.18) gives again 0 < .

Now assume that there exists limt+ ku0 (t)k22 = L. Then there exists
M > 0 such that
L
t M.
(3.19)
ku0 (t)k22
2
In this way, by (2.11), Remark 2.3 and (3.19) we get
Z
Z
0
0 < E = E(0) +
E ( ) d CM 2
h( )ku0 ( )k22 d
0

(3.20)

h( ) d,
M

where CM is a constant depending on E(0) and M . Since h is integrally positive,

there exists > 0 such that
Z
h( ) d n N,
(tn ,tn +L/3M )

so that (3.20) gives 0 < , again a contradiction.

Thus we can conclude that L = 0, i.e.
lim ku0 (t)k2 = 0.

t+

As a consequence, (3.13) implies



Z
1 1/2
kA u(t)k2
F (u(t)) dx = E0 > 0.
lim
t 2

(3.21)

Then there exists T > 0 such that for any t T one has
Z
1 1/2
E0
kA u(t)k2
.
F (u(t)) dx
2
2

Since F (u) 0 u, we get

kA1/2 u(t)k2 E0
Now set
v(t) = hu(t), u0 (t)i =

t > T.

(3.22)

1 d
ku(t)k22 ,
2 dt

(3.23)

so that
v 0 (t)

= ku0 (t)k2 + hu(t), h(t)u0 (t) Au + f (u)i

= ku0 (t)k2 h(t)v(t) kA1/2 uk22 + hu, f (u)i
ku0 (t)k2 h(t)v(t) kA1/2 uk22

by (2.7). Finally, (3.21) and (3.22) imply that there exist T0 > T and > 0
such that
v 0 (t) h(t)v(t) t T0 ,
that is


d
v(t)eH(t) eH(t) t T0 ,
dt
Rt
where H(t) = 0 h( ) d. Integrating between T0 and t gives
Z t
H(t)+H(T0 )
H(t)
v(t) v(T0 )e
e
eH( ) d.
T0

Integrating again between T0 and t, by (3.23) we get

1
ku(t)k22
2

1
ku(T0 )k22
2

+v(T0 )eH(T0 )

eH(s) ds

T0

T0

eH(s)

eH( ) d ds.

T0

Now proceed as in [19].

Remark 3.1. In proving the previous result actually what is really needed is
that
if (u0 , u1 ) D(A) H and kAu0 kH + ku1 kH B1 ,
then t > 0 f (u(t, u0 , u1 )) H and
kAu(t, u0 , u1 )kH + ku0 (t, u0 , u1 )kH + kf (u(t, u0 , u1 ))kH B2 .
However, we preferred to maintain the definition proposed in [19], since it natural to deal with solutions whose time derivative is still in H01 (), as it happens
when it is possible to apply a regularity result.
8

Remark 3.2. In proving the analogue of Theorem 3.1 in [19], the author didnt
take into account the different possibilities about the limit L defined in (3.14).
However, adapting our proof to any function f with sublinear growth and such
that sf (s) < 1 s2 , s 6= 0, like in [19], we can recover the stability result quoted
therein.
As in [19], we prove that condition (1.2) is also necessary for asymptotical
stability to hold, even without the assumption that h is integrally positive and
without the sign assumption on f , though we need f to be non supercritical,
in the usual sense. Moreover, we can even require a weaker a priori bound
condition.
Definition 3. Solutions of (2.5) are said weakly uniformly bounded in D(A1/2 )
H if for any B1 > 0 there exists B2 > 0 such that
if (u0 , u1 ) D(A1/2 ) H and kA1/2 u0 kH + ku1 kH B1 ,
then t > 0
f (u(t, u0 , u1 )) H and
kA1/2 u(t, u0 , u1 )kH + ku0 (t, u0 , u1 )kH + kf (u(t, u0 , u1 ))kH B2 , where
u(t, u0 , u1 ) denotes the solution of (2.5) with initial condition (u0 , u1 ).
Theorem 3.2. Suppose that (2.7) holds and that solutions of (1.1) are weakly
uniformly bounded in D(A1/2 ) L2 (). In addition, assume that
(
[1, 0, ) if N = 1, 2,
such that |f (s)| a + b|s|p s R.
a, b 0, p
[1, NN2 ] if N 3,
If every solution of (1.1) satisfies (3.12), then (1.2) holds.
Proof. Since every solution u of (1.1) are bounded, there exists M > 0 such
that, in particular,
kA1/2 uk2 M.
(3.24)
Moreover, by the H
older and Sobolev inequalities, there exist S1 , Sp > 0 such
that
Z
Z
1/2
|u| dx S1 kA uk2 and
|u|p dx Sp kA1/2 ukp2 .
(3.25)

Assume by contradiction that

Z
Z t
eH(t)
eH(s) ds dt < .
0

Then, for any D, > 0 there exists t0 such that

Z
Z t
D
eH(t)
eH(s) ds dt <
.
4
t0
t0
9

(3.26)

Now take H 2 () H01 () such that kk22 = D and consider the solution
u of (1.1) such that u(t0 ) = . Finally, for t t0 , set w(t) = hu(t), u0 (t)i.
Differentiating, we get
w0 (t)

= ku0 (t)k22 + hu(t), u00 (t)i

= ku0 (t)k22 + hu0 (t), Aui h(t)w(t) + hu(t), f (u)i
= ku0 (t)k22 kA1/2 u(t)k2 h(t)w(t) + hu(t), f (u)i

kA1/2 u(t)k2 h(t)w(t)

(3.27)

(a|u(t) + b|u(t)|p ) dt.

By (3.25) we get
w0 (t) kA1/2 u(t)k2 h(t)w(t) aS1 kA1/2 uk2 bSp kA1/2 ukp2 .

(3.28)

By the Young inequality we can find > 0 such that (3.28) gives
w0 (t) (1 + kA1/2 u(t)kp2 ) h(t)w(t) bSp kA1/2 ukp2 ,
and by (3.24),
w0 (t) ( + M p + bSp M p ) h(t)w(t).

(3.29)

Setting = ( + M p + bSp M p ) (which is independent of t0 ), integrating

(3.29) twice between t0 and t gives
Z t
Z s
1
1
2
2
H(s)
ku(t)k2 ku(t0 )k2
e
eH( ) d ds.
2
2
t0
t0
Finally, by (3.26), we get
D
1
ku(t)k22 ,
2
2
and a contradiction arises.
Remark 3.3. 1. Of course a uniform bound implies a weak uniform bound by
Poincares inequality. However, we preferred to present Theorem 3.2 under the
more general assumption of a weak uniform bound on the set of solutions.
2. In the proof of Theorem 3.2 we maintained the abstract form A for ,
since an analogous version for the abstract problem (2.5) can be provided at
once.
Summing up, in view of Theorems 3.2 and 3.1, we can conclude with the
following result.
Theorem 3.3. Let be a bounded domain of RN , N 1. Assume that (2.7)
holds, that h is integrally positive and that solutions of (1.1) are uniformly
bounded in H 2 () H01 () H01 (); then every solution of (1.1) verifies
kDu(t)k2 + ku0 (t)k2 0
if and only if (1.2) holds.
10

as t +

The onoff case

As in Section 2, we set V = D(A1/2 ) H and by (2.6) there exists 1 > 0 such

that
1
1
kvk2H
kvk2V =
kA1/2 vk2H for any v H.
(4.30)
1
1
We remark that in the standard case of problem (1.1) we have H = L2 (),
V = H01 () and 1 is the first eigenvalue of on H01 (), (4.30) being the
usual Poincares inequality.
Theorem 4.1. Fix T > 0 and assume there exist M, m > 0 such that
mkvk2H hB(t)v, viH

t [0, T ], v H

kB(t)vk2H 0 M hB(t)v, viH

t [0, T ], v H.

(4.31)
(4.32)

Moreover, suppose that (2.6) and (2.8) hold. Then for every (u0 , u1 ) V H
the solution u of problem (2.5) satisfies
Eu (T )

1
3

T
1+
30

Eu (0).

1
4
1

3T 2
32m

(4.33)

MT 2
161

Proof. As in [5] and [6], for any t [0, T ] we set (t) = t2 (T t)2 , so that
0 (t) = 2t(T t)(T 2t) and
|0 (t)| 2T 1/2 (t)

for all t [0, T ],

max (t) =
t[0,T ]

and

(t) dt =
0

(4.34)

T4
16

(4.35)

T5
.
30

(4.36)

Therefore (2.11) gives

T

hBu0 , u0 iH dt 0.

E(0) E(T ) =
0

Multiplying equation (2.5) by u gives

Z

n
o
hu00 + Bu0 , uiV 0 ,V + kA1/2 uk2H f (u)u dt = 0.

Therefore, integrating by parts,

Z
0

kA1/2 uk2H dt =

n
o
h(u)0 , u0 i hBu0 , uiH + f (u)u dt,

11

(4.37)

that is
Z T

1/2

kA

uk2H

n
o
ku0 k2H + 0 hu, u0 i hBu0 , uiH + f (u)u dt.

dt =

Thus for any , > 0 we get

Z T
Z
kA1/2 uk2H dt
0

+ 02 kuk2H +

n
1
kuk2H + kBu0 k2H 0
4

o
1 0 2
ku kH + ku0 k2H + f (u)u dt.
4

By (4.34), (4.35), (4.32) and (4.30) we get

Z T
Z T

MT4
kA1/2 uk2H dt
kA1/2 uk2H +
hBu0 , u0 iH

64
1
0
0
(4.38)
T4 0 2
1
4T 2
ku kH + f (u)u
kA1/2 uk2H + ku0 k2H +
+
1
4
16


dt.

Let us choose and so that

1
4T 2
=
= .
1
1
4
Then (4.38) reads
Z T
kA1/2 uk2H dt

MT4
1
kA1/2 uk2H +
hBu0 , u0 iH + f (u)u
4
161

4T 2 0 2
1
T4 0 2
ku kH
+ kA1/2 uk2H +
ku kH +
4
1
16


dt,

so that
1
2

kA1/2 uk2H dt

0
T

where CT =
1
2

4T
1

o
n
MT4
CT ku0 k2H +
hBu0 , u0 iH + f (u)u dt,
161

T4
16 .

By (4.37) this means

kA1/2 uk2H dt

(4.39)
Z

n
o

MT4 
CT ku0 k2H + f (u)u dt +
E(0) E(T ) .
161
12

By (2.10)
Z
Z T
Z
Z T
1 T
1 T
1/2
2
0 2
kA ukH dt =
E(t) dt
ku kH dt +
F (u) dt
2 0
2 0
0
0
and by Lemma 2.2
Z
1 T
kA1/2 uk2H dt
2 0
(4.40)
Z
E(T )
0

1
(t) dt
2

ku0 k2H dt +

F (u) dt.
0

Therefore (4.36), (4.39), (4.35) and (4.40) give

Z T
T5
[f (u)u F (u)]dt
E(T )
30
0

Z T

T4
MT4 
+ CT +
ku0 k2H dt +
E(0) E(T ) ,
32
161
0
which implies, together with (2.8),


Z T
 5
MT4
T
T4
MT4
+
E(0).
E(T ) CT +
ku0 k2H dt +
30
161
32
161
0
By (4.31) we get

Z T
 5

MT4
T4
T
1
MT4
+
CT +
E(T )
hBu0 , u0 iH dt +
E(0).
30
161
m
32
161
0
By (2.11) this implies


 5

 MT4
MT4
T4 
T
1
+
CT +
E(0) E(T ) +
E(T )
E(0),
30
161
m
32
161
and so


MT4
T4
T5
CT
+
+
+
30
161
m
32m


E(T )

CT
T4
MT4
+
+
m
32m
161


E(0)

and (4.33) follows.

Theorem 4.1 is the essential tool for the following stability result, which can
be proved extending the method of Smith (see [18]) as already done in [6].
Theorem 4.2. Let (an , bn )n be a sequence of disjoint open intervals in (0, +)
with an + and assume that (1.3), (1.4) and (2.8) hold. Then for every
(u0 , u1 ) D(A1/2 ) H the solution u of problem (2.5) is such that Eu (t) 0
as t +.
The straightforward proof of Theorem 4.2 is now equal to the proof of [6,
Theorem 3.2], since the main tool in the proof is an inequality of the type of
(4.33) proved in Theorem 4.1.
13

5
5.1

Some concrete applications

The integrally positive case

Consider again problem (1.1), where is a bounded domain of RN , N 1.

First, let us briefly show that the set of solutions is weakly uniformly bounded
in H01 () L2 () under the subcritical growth assumption on f of Theorem
3.2, even without any sign condition on f . Indeed, by Lemma 2.2, if u solves
(1.1), we get
Z
1
1
E(t) E(0) = ku1 k22 + ku0 k2H 1 () dx
F (u0 ) dx.
0
2
2

But

b
|F (u0 )| dx a
|u0 | dx +
p+1

|u0 |p+1 dx,

and by (3.25) there exists S > 0

Z
|F (u0 )| dx S(ku0 kH01 () + ku0 kpH 1 () ).
0

Therefore, if ku0 kH01 () + ku1 k2 B1 , then for every t 0

E(0)

B12
+ S(B1 + B1p ) := B2 .
2

Hence for every t 0

ku(t, u0 , u1 )k2H 1 ()
0

+ ku

(t, u0 , u1 )k22

Z
F (u) dx 2E(0) = 2B2 ,

= 2E(t) + 2

that is Definition 3 is verified, as claimed.

Finally, we recall that the request that the set of solutions of (1.1) is uniformly bounded in H 2 () H01 () H01 () is not so strange. Several examples
are considered in [19], and we refer to that cases therein for a sublinear f , simple
recalling the following
Example 5.1 ([19], Example 3.1). Assume that f (s) = s for some constant
< 1 , where 1 is the first eigenvalue of on H01 (). Then the set of
solutions of (1.1) is uniformly bounded in H 2 () H01 () H01 ().
Finally, we show that the set of solutions of (1.1) is uniformly bounded in
H 2 () H01 () H01 () also in more general cases. This result appears in [19]
for f having linear growth, under the additional assumptions that h is bounded
above and below by strictly positive constants (though there is a mistake in
the final step on page 198). We take the latter assumption, but we let f have
superlinear growth.

14

Lemma 5.1. Suppose that N = 1, f is an absolutely continuous function satisfying (2.7) such that that
|f 0 (s)| b1 |s|p1

b1 > 0, p [1, ) :

s R.

Assume that there exist two positive constants < such that h(t) for
any t 0. Finally assume that > 0 there exists > 0 such thatkDuk2
implies kDf (u)k2 kDuk2 . Then the set of solutions of (1.1) is uniformly
bounded in H 2 () H01 () H01 ().
Proof. First let us note that the condition on f 0 implies that |f (s)| b2 |s|p for
any s. Now take B1 > 0 and (u0 , u1 ) H 2 () H01 () H01 () such that
ku0 k2 + kDu1 k2 B1 , and observe that the growth condition on f and the
Sobolev inequality immediately ensure that f (u) L2 () for any t > 0.
By Lemma 2.2 and Remark 2.3 we get
Z
0
2
2
ku (t)k2 + kDu(t)k2 2E(0) + 2
F (u(t)) dx

Z
2E(0) = ku1 k22 + kDu0 k22 2
F (u0 ) dx.

By the Poincare and Sobolev inequality inequalities and the growth condition
on f , we get
Z


2
2
ku1 k2 + kDu0 k2 2
F (u0 ) dx C kDu1 k22 + ku0 k22 + kuk1 + kukp+1
p+1

Applying again the Poincare and Sobolev inequality inequalities, since ku0 k2 +
kDu1 k2 B1 , we get the existence of a constant B > 0 such that
ku0 (t)k22 + kDu(t)k22 B
Now set

Z
V (t) =

t > 0.

(5.41)

u(t)u0 (t) dx.

Proceeding as in [19], we can find

Z


0
V (t) =
u0 (t)2 + u00 (t)u(t) dx,

and using the equation in (1.1)

Z
Z
Z
Z
V 0 (t) =
u0 (t)2 dx
|Du(t)|2 dx h(t)
u0 (t)u(t) dx +
f (u(t))u(t) dx

By (2.7) and the H

older, Young and Poincare inequalities
Z

Z


2
u0 (t)2 dx +
1
|Du(t)|2 dx
V 0 (t) 1 +
2
2
1

15

for any > 0. Therefore



Z t
Z t

2
ku0 ( )k22 d +
kDu( )k22 d.
V (t) V (0) + 1 +
1
2
2
1
0
0
Choosing < 21 we get the existence of a positive constant C such that


Z t
Z t
ku0 ( )k22 d V (t)
t > 0.
kDu( )k22 d C V (0) +
0

But
|V (t)| ku0 (t)k2 ku(t)k2

1 0
ku (t)k2 kDu(t)k2 B1
1

by (5.41), B1 being a positive constant, so that



Z t
Z t
0
2
2
ku ( )k2 d
kDu( )k2 d C1 1 +

t > 0

t > 0

(5.42)

for some positive constant C1 .

Using again Lemma 2.2 and Remark 2.3, we have
Z t
E(t) = E(0)
h( )ku0 ( )k22 d,
0

so that

ku0 ( )k22 d

E(0)
C2

t > 0

for some positive constant C2 . In this way (5.42) implies the existence of C3 > 0
such that
Z
t

kDu( )k22 d C3

t > 0.

(5.43)

Finally, introduce
W (t) =

1
2

|Du0 (t)|2 dx +

1
2

|u(t)|2 dx.

As in [19], we find
t

h( )kDu0 ( )k22 d +

W (t) = W (0)
0

Z tZ
0

Du0 ( ) Df (u( )) dx d,

so that by Cauchys inequality

Z t
Z t
0
2
W (t) W (0)
h( )kDu ( )k2 d +
kDu0 ( )k2 kDf (u( ))k2 d.
0

16

By Youngs inequality, for any > 0

Z t
h( )kDu0 ( )k22 d
W (t) W (0)
Z t 0
Z t
1
+
kDu0 ( )k22 d +
kDf (u( ))k2 d.
2 0
0
By assumption on h we finally get
Z t
Z t
1
W (t) W (0) + ( )
h( )kDu0 ( )k22 d +
kDf (u( ))k2 d.
2 0
0
Choosing < and recalling that ku0 k2 + kDu1 k2 B1 , we get
t

Z
W (t) C + C

kDf (u( ))k2 d

t > 0

(5.44)

for some positive constant C.

By assumption, taking := B in (5.41), from (5.44) we get
Z
W (t) C + C

kDu( )kp2 d

t > 0.

(5.45)

By (5.43) we find D > 0 such that W (t) D for any t, i.e. the set of solutions
is uniformly bounded in H 2 () H01 () H01 (), as claimed.
As a final application, Lemma 5.1 gives:
Proposition 5.1. Suppose that N = 1, f is an absolutely continuous function
satisfying (2.7) such that that
b1 > 0, p [1, ) :

|f 0 (s)| b1 |s|p1

s R.

Finally assume that there exist two positive constants < such that
h(t) for any t 0. Then the set of solutions of (1.1) is uniformly bounded
in H 2 () H01 () H01 ().
Proof. The only thing to prove is that for any > 0 there exists > 0 such
thatkDuk2 implies kDf (u)k2 . But this is just an application of Sobolev
and Poincares inequality. Indeed, if kDuk2 , then
kDf (u)k2 = kf 0 (u)Duk2 b1 k|u|p1 Duk2
CkDukp2 Cp1 kDuk2 = kDuk2 ,
as soon as = Cp1 . Now apply Lemma 5.1.
Summing up, Theorems 3.1, 3.2 and Proposition 5.1 imply the following final
result.
17

Proposition 5.2. Assume N = 1 and suppose that f is an absolutely continuous function satisfying (2.7) and such that
|f 0 (s)| b1 |s|p1

b1 > 0, p [1, ) :

s R.

Moreover, assume that there exist two positive constants < such that
h(t) for any t 0. Then every solution of (1.1) satisfies
kukH01 () + ku0 kL2 () 0

as t +

if and only if (1.2) holds.

5.2

The onoff case

A particular case of the abstract problem considered in the Section 4, is the

following nonlinear wave system in a bounded domain of RN , N 1,:

in (0, +) ,
utt = u h(t)g(ut ) + f (u)
u(t, x) = 0
in (0, +) ,
(W )

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

where g : R R is a C 1 function with g(0) = 0 and 0 < m g 0 (v) M
v R, while u0 H01 () and u1 L2 ().
In this case Theorem 4.1 reads as follows.
Theorem 5.1. Fix T > 0 and assume there exist B, b > 0 such that
0 < m h(t) M

t [0, T ].

(5.46)

Suppose (2.8) holds. Then for every (u0 , u1 ) H01 () L2 () the solution u of
problem (W ) satisfies
E(T )

1
1+

T3
30

4
1

1
3T 2
T2
+ 32m
+M
16

E(0).

In the same way as Theorem 4.2 is implied by Theorem 4.1, Theorem 5.1
immediately gives the following fundamental application:
Theorem 5.2. Let (an , bn )n be a sequence of disjoint open intervals in (0, +)
with an + and assume that (1.3), (1.4) and (2.8) hold. Then for every
(u0 , u1 ) H01 () L2 () the solution u of problem (2.5) is such that Eu (t) 0
as t +.
Of course the abstract setting we gave in Section 2 lets us deal with higher
order problem.
the problem
Indeed, consider
in (0, +) ,
utt = 2m u h(t)g(ut ) + f (u)
Cu(t, x) = 0 R2m
in (0, +) ,
(H)

u(0, x) = u0 (x) D(m ), ut (0, x) = u1 (x) x ,

18

where m N, g is as before and C is a boundary operator. For example,

if m = 1 and Cu = (u, u/), being the unit outward normal to , we
have D(m ) = H02 (), while, in the case Cu = (u, u) we have D(m ) =
H01 () H 2 (). Other generalization are easy to do. For this case we have the
following
Theorem 5.3. Let (an , bn )n be a sequence of disjoint open intervals in (0, +)
with an + and assume that (1.3), (1.4) and (2.8) hold. Then for every
(u0 , u1 ) D(m )L2 () the solution u of problem (2.5) is such that Eu (t) 0
as t +.

References
[1] F. Alabau-Boussouira, A general formula for decay rates of nonlinear
dissipative systems, C. R. Math. Acad. Sci. Paris 338 (2004), no. 1, 3540.
[2] M. Bellassoued, Decay of solutions of the wave equation with arbitrary
localized nonlinear damping, J. Differential Equations 211 (2005), no. 2,
303332.
[3] S. Cox - E. Zuazua, The rate at which energy decays in a damped
string, Comm. Partial Differential Equations 19 (1994), no. 1-2, 213-243.
Elbert, Stability of some difference equations. Advances in difference
[4] A.
equations, (Veszprem, 1995), 165187, Gordon and Breach, Amsterdam,
1997.
[5] A. Haraux, On a completion problem in the theory of distributed control
of wave equations, Nonlinear partial differential equations and their applications, Collge de France Seminar, Vol. X (Paris, 19871988), 241271,
Pitman Res. Notes Math. Ser., 220, 1991.
[6] A. Haraux - P. Martinez - J. Vancostenoble, Asymptotic stability
for intermittently controlled second order evolution equations, SIAM J.
Control and Opt., 43 (2005), no. 6, 20892108.
[7] L. Hatvani - T. Krisztin, Necessary and sufficient conditions for intermittent stabilization of linear oscillators by large damping, Differential
Integral Equations 10 (1997), no. 2, 265272.
[8] I. Lasiecka - D. Toundykov, Energy decay rates for the semilinear
wave equation with nonlinear localized damping and source terms, Nonlinear Anal. 64 (2006), no. 8, 17571797.
pez-Go
mez, On the linear damped wave equation, J. Differential
[9] J. Lo
Equations 134 (1997), no. 1, 26-45.
[10] P. Martinez, Precise decay rate estimates for time-dependent dissipative
systems, Israel J. Math. 119 (2000), 291324.
19

[11] S.A. Messaoudi, Energy decay of solutions of a semilinear wave equation,

Int. J. Appl. Math. 2 (2000), no. 9, 10371048.
[12] D. Mugnai, On a reversed variational inequality, Topol. Methods Nonlinear Anal. (2001) Vol. 17, No. 2, pp. 321358.
[13] P. Pucci - J. Serrin, Asymptotic stability for intermittently controlled
nonlinear oscillators, SIAM J. Math. Anal. 25 (1994), no. 3, 815835.
[14] P. Pucci - J. Serrin, Asymptotic stability for nonautonomous dissipative wave systems, Comm. Pure Appl. Math. 49 (1996), no. 2, 177216.
[15] P. Pucci - J. Serrin, Precise damping conditions for global asymptotic
stability for nonlinear second order systems, Acta Math. 170 (1993), no.
2, 275307.
[16] P. Pucci - J. Serrin, Precise damping conditions for global asymptotic
stability for nonlinear second order systems. II, J. Differential Equations
113 (1994), no. 2, 505534.
[17] P. Pucci - J. Serrin, Stability for abstract evolution equations, Partial
differential equations and applications, 279288, Lecture Notes in Pure
and Appl. Math. 177, Dekker, New York, 1996.
[18] R. A. Smith,Asymptotic stability of x00 + a(t)x0 + x = 0, Quart. J. Math.
Oxford Ser. (2) 12 (1961), 123126.
[19] B. Zhang, Asymptotic behavior of solutions of a nonlinear damped wave
equation, Differential Equations Dynam. Systems 2 (1994), no. 3, 173204.

20