Electronic Journal of Qualitative Theory of Differential Equations 2012, No. 7, 1-13; http://www.math.u-szeged.hu/ejqtde/
Observation problems posed for the Klein-Gordon equation
ANDRÁS SZIJÁRTÓ 1
, JENHEGEDS
(SZTE,Bolyai Institute,Szeged, Hungary)
szijartomath.u-szeged.hu, hegedusjmath.u-szeged.hu
Abstrat.Transversal vibrations
u = u(x, t)
ofa stringof lengthl
withxed endsare onsidered, where
u
is governed by the Klein-Gordonequationu tt (x, t) = a 2 u xx (x, t) + cu(x, t), (x, t) ∈ [0, l] × R , a > 0, c < 0.
Suient onditions are obtained that guarantee the solvability of eah of four
observation problems with given state funtions
f, g
at two distint time instants−∞ < t 1 < t 2 < ∞
. The essential onditions are the following: smoothness off, g
as elements of a orresponding subspae
D s+i (0, l)
(introdued in [2℄) of a Sobolevspae
H s+i (0, l)
,wherei = 1, 2
dependingonthe typeofthe observation problem,and the representability oft 2 − t 1 as a rational multiple of
2l
a
. The reonstrution of the unknown initialdata(u(x, 0), u t (x, 0))
astheelementsofD s+1 (0, l) × D s (0, l)
are givenby means of the method of Fourier expansions.
2010 AMS Subjet Classiations: Primary 35Q93, 81Q05;Seondary 35L05,
35R30,42A20
Key words: Observation problems, Klein-Gordonequation, generalizedsolutions,
method of Fourierexpansions.
1. BACKGROUND AND KNOWN RESULTS
In ontrol theory - whih is losely related to the subjet of this paper - numer-
ous monographies and artiles dealt with the aessability of a nal state (position
and speed) of osillations (in partiular string osillations) in the time interval
0 ≤ t ≤ T < ∞
; see for example, [1℄ - [10℄. Although, only the short ommunia-tion [11℄ dealt with observability of the string osillations on the interval
0 ≤ x ≤ l
,and ittreated just the ase whenthe observation instants
t 1 and t 2 are small, namely
0 ≤ t 1 ≤ t 2 ≤ 2l
a
, wherea
is the speed of the wave propagation. Furthermore, it is assumed in [11℄ that the initial data are known on some subinterval[h 1 , h 2 ] ⊂ [0, l]
.We reonstruate the initial data in eah of the four observation problems related to
the Klein-Gordon equation for arbitrarylarge
t 1 and t 2. Our preassumptions are only
that(t 2 − t 1 ) a
(t 2 − t 1 ) a
2l
isrationaland thegiven statefuntions aresmooth enough. Theasesf, g ∈ D s with arbitrarys ∈ R
are also admitted.
1
Correspondingauthor
Let
Ω = {(x, t) : 0 < x < l, t ∈ R }
. Consider the problem(atrst inthe lassialsense)ofthe vibrating
[0, l]
stringwith xedends whenthereisanelasti withdrawing fore proportional to the transversal deetionu(x, t)
of the pointx
of the string attheinstantdenoted by
t
.Thisphenomenon isdesribedbythe Klein-Gordonequation asfollows:(1) u tt (x, t) = a 2 u xx (x, t) + cu(x, t), (x, t) ∈ Ω, a, c ∈ R , 0 < a, 0 > c,
with the initialonditions
(2) u(x, 0) = ϕ(x), u t (x, 0) = ψ(x), 0 ≤ x ≤ l,
and the homogeneousboundaryonditions of the rst kind
(3) u(0, t) = 0, u(l, t) = 0, t ∈ R .
We reall, that the funtion
u
is said to be a lassial solution of this problem, ifu ∈ C 2 (Ω)
and onditions(1) − (3)
are satised.It is wellknown that if
(4) ϕ ∈ C 2 [0, l], ψ ∈ C 1 [0, l]
andϕ(0) = ϕ(l) = ϕ ′′ (0) = ϕ ′′ (l) = ψ(0) = ψ(l) = 0,
then the Fourier method gives the lassial solution
u
of the problem(1) − (3)
posedforthe Klein-Gordonequation, whih isof the following form:
(5) u(x, t) =
∞
X
n=1
[α n cos (tω n ) + β n sin (tω n )] sin( nπ
l x), (x, t) ∈ Ω,
where
(6) ω n =
r ( nπ
l a) 2 − c, n ∈ N ,
(7) ϕ(x) = u(x, 0) =
∞
X
n=1
α n sin( nπ
l x) ⇒ α n = 2 l
Z l
0
ϕ(x) sin( nπ
l x)dx, n ∈ N ,
(8) ψ(x) = u t (x, 0) =
∞
X
n=1
ω n β n sin( nπ
l x) ⇒ β n = 1 ω n
2 l
Z l
0
ψ(x) sin( nπ
l x)dx, n ∈ N .
The uniqueness of the solution isa onsequene of the lawof onservation of energy.
Tohaveawiderlass offuntionsfor
ϕ, ψ
andf, g
,weshallonsider ertaingener-alizedsolutionsoftheproblem
(1)−(3)
.Namely,byusingthesuggestionsofthereferee,we introdue the spaes
D s (0, l)
,s ∈ R
mentioned in the abstrat (see [2℄). Given an arbitrary real numbers
, on the linear spanD
of the funtionssin nπ l x
,n = 1, 2, ...,
onsider the followingEulidean norm:
∞
X
n=1
c n sin( nπ l x)
s
:=
∞
X
n=1
n 2s |c n | 2
! 1 2 .
Completing
D
withrespettothisnorm,weobtainaHilbertspaeD s.Oneanreadily
verify that for
s ≥ 0
,D s is a losedsubspae of the Sobolevspae H s (0, l)
, namely
D s = {u ∈ H s (0, l) : u (2i) (0) = u (2i) (l) = 0, i = 0, 1, ..., [(s − 1)/2]}.
If we identify
D 0 = L 2 (0, l)
with its dual, thenD −s is the dual spae of D s. Some
of the results of [2℄ (see Setion 1.1-1.3) and [10℄ say that for arbitrary
s ∈ R
with(ϕ, ψ) ∈ D s+1 × D s the generalized mixed problem (1) − (3)
has a unique solution u
satisfying
u ∈ C( R , D s+1 ) ∩ C 1 ( R , D s ) ∩ C 2 ( R , D s−1 )
givenbytheFourierseries
(5)
withoeientsα n , β n denedby (7)
and(8)
.Hereand
belowallFourierexpansions for
ϕ, ψ, f
,g
andu
are understood inthe spaesD s (0, l)
.2. NEW RESULTS
Denition 1. The observation problem posed for the Klein-Gordon equation is the
following. The initial funtions
ϕ
,ψ
are unknown, but suh funtionsf(x)
andg(x)
are given for whih one of the followingfour onditions holds:
(9) u(x, t 1 ) = f(x), u(x, t 2 ) = g(x), 0 ≤ x ≤ l;
(10) u t (x, t 1 ) = f (x), u(x, t 2 ) = g(x), 0 ≤ x ≤ l;
(11) u(x, t 1 ) = f (x), u t (x, t 2 ) = g(x), 0 ≤ x ≤ l;
(12) u t (x, t 1 ) = f(x), u t (x, t 2 ) = g(x), 0 ≤ x ≤ l.
Here
u
is the solution of the generalized problem(1) − (3)
, and the given funtionsf, g
are said to be the partial state of the string at distint time instantst 1 and t 2,
−∞ < t 1 < t 2 < ∞
. Now the problem is to nd the initialfuntionsϕ
,ψ
interms off(x)
,g (x)
.(13) f ∈ D s+2 , g ∈ D s+2 , where s ∈ R ,
(14) t 2 − t 1 = p
q 2l
a ,
where
p, q
are positive integers and they are relative primes. Inaddition, suppose that(15) sin
(t 2 − t 1 ) r
( nπ
l a) 2 − c
6= 0, ∀n ∈ N .
Then the observation problem
(1) − (3)
under ondition(9)
has a unique solution for(ϕ, ψ) ∈ D s+1 × D s. They are represented by their Fourier expansions in the proof below.
Theorem 2. Suppose that
(16) f ∈ D s+1 , g ∈ D s+2 , where s ∈ R ,
ondition
(14)
holds and(17) cos
(t 2 − t 1 ) r
( nπ
l a) 2 − c
6= 0, ∀n ∈ N .
Thenthe observation problem
(1) − (3)
underondition(10)
has a unique solution for(ϕ, ψ) ∈ D s+1 × D s. They are represented by their Fourier expansions in the proof below.
Theorem 3. Suppose that
(18) f ∈ D s+2 , g ∈ D s+1 , where s ∈ R ,
and onditions
(14)
and(17)
hold. Thenthe observation problem(1) − (3)
under on-dition
(11)
has a uniquesolution for(ϕ, ψ) ∈ D s+1 × D s. They are represented bytheir Fourier expansionsin the proof below.
Theorem 4. Suppose that
(19) f ∈ D s+1 , g ∈ D s+1 , where s ∈ R ,
and onditions
(14)
and(15)
hold. Thenthe observation problem(1) − (3)
under on-dition
(12)
has a uniquesolution for(ϕ, ψ) ∈ D s+1 × D s. They are represented bytheir Fourier expansionsin the proof below.
Lemma 1. If ondition
(14)
holds, then there existN ∈ N
andm ∈ R
suh that1
|sin(ω n (t 2 − t 1 ))| < n
m , ∀n > N.
Proof. First,we deal with the denominator of the left-hand side of the inequality
(20) sin(ω n (t 2 − t 1 )) = sin
(t 2 − t 1 ) nπ
l a + (t 2 − t 1 ) h
ω n − nπ l a i
=
= sin
(t 2 − t 1 ) nπ
l a + (t 2 − t 1 ) ω n 2 − ( nπ l a) 2 ω n + nπ l a
= sin
(t 2 − t 1 ) nπ
l a + (t 2 − t 1 ) −c ω n + nπ l a
.
It follows from the ondition
(14)
that(t 2 − t 1 ) nπ
l a = p q 2nπ,
and that ittakeson atmost
q
dierent values(
mod2π)
asn
varies. Letz n := (t 2 − t 1 ) nπ
l a
andd 1 := min
n, sin z n 6=0 {|sin (z n )|}.
Dueto the absolutevalue bars, there isa real number
d 2 suh that
sin(d 2 ) = d 1 , 0 < d 2 ≤ π 2 .
It iseasy tosee, that
(t 2 − t 1 ) −c
ω n + nπ l a = O( 1
n )
asn → ∞.
Therefore, there existonstants
N ∈ N
,m ∈ R + suh that
(21) πm
2n <
(t 2 − t 1 ) −c ω n + nπ l a
< d 2
2
andm
n < sin d 2
2
, ∀n > N.
So, if
sin
(t 2 − t 1 ) nπ l a
6= 0
, thensin
(t 2 − t 1 ) nπ
l a + (t 2 − t 1 ) −c ω n + nπ l a
>
sin
d 2 − d 2 2
= sin d 2
2
> m n ,
whenever
n > N
,by virtue of(21)
.On the other hand, if
sin
(t 2 − t 1 ) nπ l a
= 0
, thensin
(t 2 − t 1 ) nπ
l a + (t 2 − t 1 ) −c ω n + nπ l a
=
sin
(t 2 − t 1 ) −c ω n + nπ l a
>
> 2 π
(t 2 − t 1 ) −c ω n + nπ l a
> m
n , ∀n > N,
due to
(21)
and the inequality(22) |sin t| > 2
π |t| ,
if0 < |t| < π 2 .
Combining the two ases just above, weget that
sin
(t 2 − t 1 ) nπ
l a + (t 2 − t 1 ) −c ω n + nπ l a
> m
n , ∀n > N.
Lemma 2. If ondition
(14)
holds, then there existN ∈ N
andm ∈ R
suh that1
|cos(ω n (t 2 − t 1 ))| < n
m , ∀n > N.
Proof. Similarlyto
(20)
inthe proof of Lemma 1,now we obtainthatcos(ω n (t 2 − t 1 )) = cos
(t 2 − t 1 ) nπ
l a + (t 2 − t 1 ) −c ω n + nπ l a
.
Let
z n := (t 2 − t 1 ) nπ
l a
andd 1 := min
n, cos z n 6=0 {|cos z n |}.
Dueto the absolutevalue bars, there isa real number
d 2 suh that
cos(d 2 ) = d 1 , 0 ≤ d 2 < π 2 .
Similarly to
(21)
in the proof of Lemma 1, there exist onstantsN ∈ N
andm ∈ R +
suh that
(23) πm 2n <
(t 2 − t 1 ) −c ω n + nπ l a
<
π 2 − d 2
2
andm n < cos
π 2 + d 2
2
, ∀n > N.
In this manner, if
cos
(t 2 − t 1 ) nπ l a
6= 0
, we obtainagain thatcos
(t 2 − t 1 ) nπ
l a + (t 2 − t 1 ) −c ω n + nπ l a
>
cos
d 2 +
π 2 − d 2
2
= cos π
2 + d 2
2
> m n ,
whenever
n > N
,by virtue of(23)
.On the other hand, inthe ase when
cos
(t 2 − t 1 ) nπ l a
= 0
, we getcos
(t 2 − t 1 ) nπ
l a + (t 2 − t 1 ) −c ω n + nπ l a
=
sin
t 2 − t 1 ) −c ω n + nπ l a
>
> 2 π
(t 2 − t 1 ) −c ω n + nπ l a
> m
n , ∀n > N,
due to
(22)
and(23)
.cos
(t 2 − t 1 ) nπ
l a + (t 2 − t 1 ) −c ω n + nπ l a
> m
n , ∀n > N.
4.PROOFSOF THE THEOREMS
1 − 4
Proof of Theorem1. Sineanyofthesolutions
u
ofproblem(1)−(3)
hasrepresentation(5)
with some oeientsα n , β n ; n ∈ N
, the observation problem an be redued to the problemof the appropriatehoies ofα n and β n suhthat (9)
is satised. Forthis
(9)
is satised. Forthisreason,we substitute
t 1 andt 2 into(5)
, and use the two onditionsin (9)
. Asa result,
(5)
, and use the two onditionsin(9)
. Asa result,weget the followingneessary onditions for
α n , β n:
(24) f(x) = u(x, t 1 ) =
∞
X
n=1
[α n cos(ω n t 1 ) + β n sin(ω n t 1 )] sin( nπ
l x), x ∈ [0, l],
(25) g (x) = u(x, t 2 ) =
∞
X
n=1
[α n cos(ω n t 2 ) + β n sin(ω n t 2 )] sin( nπ
l x), x ∈ [0, l],
where
ω n isdened in(6)
.
The assumption
(13)
guarantees that the oeients of the sine Fourier expansions ofthe funtions
f(x), g(x)
are unambiguously determined and omparing these Fourier series with(24)
and(25)
, forα n , β n we get the followingonditions:
(26)
α n cos(ω n t 1 ) + β n sin(ω n t 1 ) = 2 l
Z l
0
f(x) sin( nπ
l x)dx, n ∈ N , α n cos(ω n t 2 ) + β n sin(ω n t 2 ) = 2
l Z l
0
g(x) sin( nπ
l x)dx, n ∈ N .
The linearsystem
(26)
an be uniquely solved forthe unknown oeientsα n and β n
due toassumption
(15)
:(27)
α n = sin(ω n t 2 ) 2 l R l
0 f (x) sin( nπ l x)dx − sin(ω n t 1 ) 2 l R l
0 g(x) sin( nπ l x)dx
sin(ω n (t 2 − t 1 )) ,
β n = − cos(ω n t 2 ) 2 l R l
0 f (x) sin( nπ l x)dx + cos(ω n t 1 ) 2 l R l
0 g(x) sin( nπ l x)dx
sin(ω n (t 2 − t 1 )) .
So the unknown initial funtions
ϕ
andψ
are uniquely determined and found in theform of
(7)
and(8)
. It remains to show thatϕ, ψ
are from the lassesD s+1 , D s,
respetively, i.e. toshow that the followinginequality holds:
(28) max{kϕk 2 s+1 , kψk 2 s } < ∞.
D n := 2 l
Z l
0
f (x) sin( nπ l x)dx, E n := 2
l Z l
0
g(x) sin( nπ l x)dx.
Sine
(f, g) ∈ D s+2 × D s+2, we havethe following inequality:
(29)
∞
X
n=1
n 2s+4 max{|D n | 2 , |E n | 2 } < ∞.
Byusing Lemma 1,for every
n > N
we get|α n | =
sin(ω n t 2 )D n − sin(ω n t 1 )E n
sin(ω n (t 2 − t 1 ))
<
n m D n
+
n m E n
,
|β n | =
− cos(ω n t 2 )D n + cos(ω n t 1 )E n
sin(ω n (t 2 − t 1 ))
<
n m D n
+
n m E n
,
whih meansthat
(30) max{|α n |, |β n |} < c 1 n max{|D n |, |E n |} n ∈ N ,
with asuitable onstant
c 1.
Let
M ≥ 1
be aonstant suh thatω n < Mn, ∀n ∈ N
. Combining(29)
,(30)
and thedenition of the norm
k.k s we get the desired inequality (28)
:
max{kϕk 2 s+1 , kψk 2 s } = max{
∞
X
n=1
n 2s+2 |α n | 2 ,
∞
X
n=1
n 2s |ω n β n | 2 } ≤
≤
∞
X
n=1
M 2 n 2s+2 max{|α n | 2 , |β n | 2 } < c 2 1 M 2
∞
X
n=1
n 2s+4 max{|D n | 2 , |E n | 2 } < ∞.
Remark 1. In the lassial ase when the given state funtions are ontinuously dif-
ferentiable, aording to Theorem 1, the initialfuntions are also ontinuously dier-
entiable.More preisely, if
u(x, t 1 ) = f (x) ∈ C 4 [0, l], u(x, t 2 ) = g(x) ∈ C 4 [0, l], f, g| 0,l = f ′′ , g ′′ | 0,l = 0,
then
f, g ∈ D 4 and the observation problemhas a unique lassialsolution
u(x, 0) = ϕ(x) ∈ D 3 ⊂ C 2 , u t (x, 0) = ψ(x) ∈ D 2 ⊂ C 1 .
Remark2. Takingintoaount
(20)
,ondition(15)
anbewrittenintothefollowingform:
(31) sin ((t 2 − t 1 )ω n ) = sin (t 2 − t 1 ) nπ
l a + (t 2 − t 1 ) −c
p ( nπ l a) 2 − c + nπ l a
! 6= 0
forall
n ∈ N
.AnalysingtheproofofLemma1,itiseasytoseethattheaboveondition isertainly satisedfor alln
large enough, sayn > N
.Ifwe want toget aneasily veriableondition insteadof
(15)
, whih is not neessarythen
(32) (t 2 − t 1 ) −c
p ( π l a) 2 − c + π l a < π q
is suh a suient ondition. We justify this laim as follows. The rst term in the
argument of the sine funtion in
(31)
is either0 (
mod2π)
, or its distane is at leastπ
q
fromitszeroes, and the seond term inthe argument of the sine funtionin(31)
ispositiveandmonotone dereasingfuntionof
n
.So, ifweassumethat theseond termisalready smaller than
π
q
forn = 1
, whihis atually the ase in(32)
,then ondition(31)
issatised for eahn ≥ 1
.Nevertheless, wean see fromthis simplerondition
(32)
,thatifthe parameters|c|
and
a
inequation(1)
aresuhthateitherc
issmallora
isgreatenough, thenondition(31)
is always satised. Similar observations an be made in the following Theorems2 − 4
.Proof of Theorem 2. In an analogous way asin the proof of Theorem 1, nowwestart
with the following equalities:
f (x) = u t (x, t 1 ) =
∞
X
n=1
[−α n ω n sin(ω n t 1 ) + β n ω n cos(ω n t 1 )] sin( nπ
l x), x ∈ [0, l],
g(x) = u(x, t 2 ) =
∞
X
n=1
[α n cos(ω n t 2 ) + β n sin(ω n t 2 )] sin( nπ
l x), x ∈ [0, l].
Hene we get the followingneessary onditions for the oeients
α n, β n:
−α n ω n sin(ω n t 1 ) + β n ω n cos(ω n t 1 ) = 2 l
Z l
0
f (x) sin( nπ
l x)dx, n ∈ N , α n cos(ω n t 2 ) + β n sin(ω n t 2 ) = 2
l Z l
0
g(x) sin( nπ
l x)dx, n ∈ N .
Thelinear equations justreeived an beuniquely solved for the unknown oeients
α n and β n,due to assumption (17)
:
(17)
:α n = − sin(ω n t 2 ) 2 l R l
0 f (x) sin( nπ l x)dx + cos(ω n t 1 )ω n 2 l
R l
0 g(x) sin( nπ l x)dx
ω n cos(ω n (t 2 − t 1 )) ,
β n = cos(ω n t 2 ) 2 l R l
0 f (x) sin( nπ l x)dx + sin(ω n t 1 )ω n 2 l
R l
0 g(x) sin( nπ l x)dx ω n cos(ω n (t 2 − t 1 )) .
So the unknown initial funtions
ϕ
andψ
are uniquely determined and found in theform of
(7)
and(8)
. It remains to show thatϕ, ψ
are from the lassesD s+1 , D s,
respetively. To this eet, it isenough toshow that
(28)
holds.Again, let
D n := 2 l
Z l
0
f (x) sin( nπ l x)dx, E n := 2
l Z l
0
g(x) sin( nπ l x)dx.
Sine
(f, g) ∈ D s+1 × D s+2, we havethat the inequality (29 ′ )
holds:
(29 ′ )
∞
X
n=1
n 2s+4 max{| 1
n D n | 2 , |E n | 2 } < ∞.
Byusing Lemma 2,for every
n > N
we have|α n | =
− sin(ω n t 2 )D n + cos(ω n t 1 )ω n E n
ω n cos(ω n (t 2 − t 1 ))
<
1 ω n
n m D n
+
n m E n
,
|β n | =
cos(ω n t 2 )D n + sin(ω n t 1 )ω n E n
ω n cos(ω n (t 2 − t 1 ))
<
1 ω n
n m D n
+
n m E n
,
whih meansthat
(30 ′ ) max{|α n |, |β n |} < c 2 n max{| 1
n D n |, |E n |} n ∈ N ,
with asuitable onstant
c 2.
Combining
(29 ′ )
,(30 ′ )
andthedenition ofthenormk.k s wegetthedesiredinequality
(28)
:max{kϕk 2 s+1 , kψk 2 s } = max{
∞
X
n=1
n 2s+2 |α n | 2 ,
∞
X
n=1
n 2s |ω n β n | 2 } ≤
≤
∞
X
n=1
M 2 n 2s+2 max{|α n | 2 , |β n | 2 } < c 2 2 M 2
∞
X
n=1
n 2s+4 max{| 1
n D n | 2 , |E n | 2 } < ∞.
Proof of Theorem3. Thisproofgoesalongthesamelines asthatofTheorem2,exept
that here wehave tointerhange the roles of the oeients
α n and β n.
f (x) = u t (x, t 1 ) =
∞
X
n=1
[−α n ω n sin(ω n t 1 ) + β n ω n cos(ω n t 1 )] sin( nπ
l x), x ∈ [0, l], g(x) = u t (x, t 2 ) =
∞
X
n=1
[−α n ω n sin(ω n t 2 ) + β n ω n cos(ω n t 2 )] sin( nπ
l x), x ∈ [0, l],
whene the neessary onditions for the oeients
α n, β n are the following:
−α n ω n sin(ω n t 1 ) + β n ω n cos(ω n t 1 ) = 2 l
Z l
0
f (x) sin( nπ
l x)dx, n ∈ N ,
−α n ω n sin(ω n t 2 ) + β n ω n cos(ω n t 2 ) = 2 l
Z l
0
g(x) sin( nπ
l x)dx, n ∈ N .
Thelinear equations justreeived an beuniquely solved for the unknown oeients
α n and β n,due to assumption (15)
:
α n = cos(ω n t 2 ) 2 l R l
(15)
:α n = cos(ω n t 2 ) 2 l R l
0 f(x) sin( nπ l x)dx − cos(ω n t 1 ) 2 l R l
0 g(x) sin( nπ l x)dx ω n sin(ω n (t 2 − t 1 )) , β n = sin(ω n t 2 ) 2 l R l
0 f(x) sin( nπ l x)dx − sin(ω n t 1 ) 2 l R l
0 g(x) sin( nπ l x)dx ω n sin(ω n (t 2 − t 1 )) .
So the unknown initial funtions
ϕ
andψ
are uniquely determined and found in theform of
(7)
and(8)
. It remains to show thatϕ, ψ
are from the lassesD s+1 , D s,
respetively. To this eet, it isenough toshow that
(28)
holds.Again, let
D n := 2 l
Z l
0
f (x) sin( nπ l x)dx, E n := 2
l Z l
0
g(x) sin( nπ l x)dx.
Sine
(f, g) ∈ D s+1 × D s+1, we havethat the inequality (29 ′′ )
holds:
(29 ′′ )
∞
X
n=1
n 2s+2 max{|D n | 2 , |E n | 2 } < ∞.
Byusing Lemma 1,for every
n > N
we get|α n | =
cos(ω n t 2 )D n − cos(ω n t 1 )E n
ω n sin(ω n (t 2 − t 1 ))
<
1 ω n
n m D n
+
1 ω n
n m E n
,
|β n | =
sin(ω n t 2 )D n − sin(ω n t 1 )E n
ω n sin(ω n (t 2 − t 1 ))
<
1 ω n
n m D n
+
1 ω n
n m E n
,
whih meansthat
(30 ′′ ) max{|α n |, |β n |} < c 4 max{|D n |, |E n |} n ∈ N ,
with asuitable onstant
c 4.
Combining
(29 ′′ )
,(30 ′′ )
andthedenitionofthenormk.k s wegetthedesiredinequality
(28)
:max{kϕk 2 s+1 , kψk 2 s } = max{
∞
X
n=1
n 2s+2 |α n | 2 ,
∞
X
n=1
n 2s |ω n β n | 2 } ≤
≤
∞
X
n=1
M 2 n 2s+2 max{|α n | 2 , |β n | 2 } < c 2 4 M 2
∞
X
n=1
n 2s+2 max{|D n | 2 , |E n | 2 } < ∞.
ACKNOWLEDGEMENTS
We would like to sinerely thank Professor Feren Móriz for his enouragement
andonstanthelponapreviousdraft.Wealsothanktheanonymousreferee forareful
reading of the previous draft and his/her suggestions whih resulted in the present
version of this paper.
REFERENCES
[1℄ Lions,J.L.,Contrleoptimaldesystemes gouvernéspar deséquationsauxdérivées
partielles (Étudesmathématiques), Paris: Dunod,1968.
[2℄ Komornik,V., Exat ontrollability andstabilization. The multiplier method., Re-
searh inApplied Mathematis 36. Chihester: Wiley. Paris: Masson, 1994.
[3℄ Komornik,V.,Exat ontrollability in shorttimefor thewaveequation,Ann. Inst.
Henri Poinaré, Anal.Non Linéaire6, No.2, pp. 153-164,1989.
[4℄ Joó, I., On the vibration of a string., Studia Si. Math. Hungar 22, No. 1-4, pp.
1-9, 1987.
[5℄ Horváth,M.,Vibrating stringswith freeends,Ata Math.Hungar51,No.1-2,pp.
171-180,1988.
[6℄ Il'in, V.A., Boundary ontrol of the vibration proess at two ends, Dokl. Aad.
Nauk369, No.5, pp. 592-596,1999.
[7℄ Moiseev,E.I.,Kholomeeva,A.A.,Optimalboundarydisplaementontrolofstring
vibrations with nonloal oddness ondition of therst kind,DierentialEquations
46,No. 11,pp. 1624-1630, 2010.
UspekhiMat. Nauk 60,No. 6,pp. 89-114, 2005.
[9℄ Lions, J.L., Exat ontrollability, stabilization and perturbations for distributed
systems, SIAM Rev. 30,No. 1,pp. 1-68, 1988.
[10℄ Lions, J.L., Magenes, E., Probl
`
emes aux limites non homog`
enes et appliations I-III,Dunod,Paris, 1968-1970.[11℄ Znamenskaya,L.N.,Stateobservability ofelastistringvibrationsunderthebound-
aryonditionsof therstkind,DierentialEquations46,No.5,pp.748-752,2010.
(Reeived Marh30, 2011)