Universität Konstanz
Representation of classical solutions to a linear wave equation with pure delay
Denys Khusainov Michael Pokojovy Elvin Azizbayov
Konstanzer Schriften in Mathematik Nr. 322, November 2013
ISSN 1430-3558
© Fachbereich Mathematik und Statistik Universität Konstanz
Konstanzer Online-Publikations-System (KOPS)
URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-247552
with Pure Delay
Denys Ya. Khusainov
∗
, Mihael Pokojovy
†
, Elvin I. Azizbayov
‡
Otober3, 2013
Abstrat
Forawaveequationwithpuredelay,westudyaninhomogeneousinitial-boundaryvalueproblem
in abounded 1D domain. Under smoothnessassumptions, weproveuniqueexisteneof lassial
solutions foranygivennite time horizon and givetheir expliit representation. Continuousde-
pendeneonthedatainaweakextrapolatednormisalsoshown.
1 Introdution
The wave equation is a typial linear hyperboli seond-order partial dierential equationwhih nat-
urally arises when modeling phenomena ofontinuum mehanis suh assound, light, water or other
kind ofwavesinaoustis,(eletro)magnetis, elastiityanduiddynamis, et. (f. [6, 13℄). Provid-
ingaratheradequatedesriptionofphysialproesses,partialdierential equations,or equationswith
distributed parameters ingeneral,have foundnumerousappliations inmehanis, mediine,eology,
et. Introduing after-eets suh as delay into these equations has gained a lot of attention over
severalpastdeades. See e.g.,[2 ,3,7,8℄. Mathematialtreatment ofsuhsystemsrequiresadditional
arefulnesssine distributed systemswithdelayoften turnout to be even ill-posed(f. [4 ,5, 12℄).
Inthepresent paper,weonsideraninitial-boundaryvalueproblemfor agenerallinearwaveequation
withpuredelayand onstant oeientsina bounded intervalsubjetto non-homogeneous Dirihlet
boundary onditions. To solve the equation, we employ Fourier's separation method as well as the
speial funtions referred to as delay sine and osine funtions whih were introdued in [9 , 10℄. We
prove the existene of a unique lassial solution on any nite time interval, show its ontinuous
dependene on the dataina very weakextrapolated norm, give its representation asa Fourier series
and prove itsabsolute anduniform onvergene under ertainonditions onthedata.
2 Equation with pure delay
For
T > 0
,l > 0
, we onsider the following linear wave equation in a bounded interval(0, l)
with asingle delaybeinga seondorderpartial dierene-dierential equation foran unknown funtion
η
∂ tt η(t, x) = a 2 ∂ xx η(t − τ, x) + b∂ x η(t − τ, x) + dη(t − τ, x) + g(t, x)
for(t, x) ∈ (0, T ) × (0, l)
(2.1)∗
DepartmentofCybernetis,KyivNationalTarasShevhenkoUniversity,Ukraine
†
DepartmentofMathematisandStatistis,UniversityofKonstanz,Germany
‡
DepartmentofMehanisandMathematis,BakuStateUniversity,Azerbaijan
η(t, 0) = θ 1 (t), η(t, l) = θ 2 (t)
fort > −τ,
η(t, x) = ψ(t, x)
for(t, x) ∈ (−τ, 0) × (0, l).
(2.2)Sine we are interested in studying lassial solutions, the following ompatibility onditions are re-
quired toassurefor smoothnessofthesolution on theboundary oftime-spae ylinder
ψ(t, 0) = θ 1 (t), ψ(t, l) = θ 2 (t)
fort > −τ.
Denition 2.1. Under a lassial solution to the problem (2.1 ), (2.2) we understand a funtion
η ∈ C 0 [−τ, T ] × [0, l]
whih satises
∂ tt η, ∂ tx η, ∂ xx η ∈ C 0 [−τ, 0] × [0, l]
aswell as
∂ tt η, ∂ tx η, ∂ xx η ∈ C 0 [0, T ] × [0, l]
and,being pluggedinto Equations (2.1 ),(2.2 ), turnsthem into identity.
Remark 2.2. The previous denition does not impose any ontinuity of time derivatives in
t = 0
.If the ontinuity or even smoothness are desired, additional ompatibility onditions on the data,
inluding
g
,arerequired.Let
k · k k,2 := k · k H k,2 ((0,l))
,k ∈ N 0
, denote the standard Sobolev norm (f. [1℄) andk · k −k,2 :=
k · k H −k,2 ((0,l))
denote the norm of orresponding negative Sobolev spae. We introdue the normk · k X :=
s ∞
P
k=0
k · k 2 −k,2
and dene the Hilbert spaeX
as a ompletion ofL 2 (0, l)
with respet
to
k · k X
. Obviously,X ֒ → D (0, l) ′
, i.e.,
X
an be ontinuously embedded into the spae of distributions.With this notation, we easily see that
A := a 2 ∂ x 2 + b∂ x + d
(with∂ x
denoting the distributional derivative)is abounded linearoperatoronX
sinekAk L(X) = sup
kuk X =1
kAuk X = sup
kuk X =1
v u u t
∞
X
k=0
ka 2 ∂ x 2 u + b∂ x u + duk 2 −k,2
≤ sup
kuk X =1
∞
X
k=0
a 2 kuk −k−2,2 + bkuk −k−1,2 + dkuk −k,2
≤ sup
kuk X =1
(a 2 + b + c)kuk X = a 2 + b + d.
Theorem 2.3. Thereexistsaonstant
C > 0
,dependentonly ona, b, d, l, τ, T
,suhthattheestimatet∈[0,T] max kη(·, t)k 2 X + kη t (·, t)k 2 X
≤ C kψ(0, ·)k 2 X + kψ t (0, ·)k 2 X + C
Z 0
−τ
kψ(s, ·)k 2 X + kψ t (s, ·)k 2 X dt+
C Z T
0
kg(s, ·)k 2 X + |θ 1 (s)| 2 + |θ 2 (s)| 2 ds
holds truefor anylassial solutionof Equations(2.1 ), (2.2 ).
Proof. Let
η
be the lassialsolution to Equations(2.1 ), (2.2 ). We denew(t, x) :=
( η(t, x)
for(t, x) ∈ [−τ, 0] × [0, l],
η(t, x) − θ 1 (t) − x l (θ 2 (t) − θ 1 (t))
for(t, x) ∈ (0, T ] × [0, l].
Then
w
satiseshomogeneous Dirihlet boundary onditions and solvestheequation∂ tt w(t, ·) = Aw(t − τ, ·) + f (t, ·)
(2.3)inthe extrapolated spae
X
withf(t, ·) = g(t, ·) + b (θ 2 (t) − θ 1 (t)) + θ 1 (t) + dx l (θ 2 (t) − θ 1 (t)) .
We multiply the equation with
w t (t, ·)
in thesalar produt ofX
and use Young's inequality to gettheestimate
∂ t kw t (t, ·)k 2 X = hAw(t − τ, ·), w t (t, ·)i X + hf (t, ·), w t (t, ·)i X
≤ kw(t − τ, ·)k 2 X +
1 + kAk 2 L(X )
kw t (t, ·)k 2 X + kf (t, ·)k 2 X .
(2.4)
As in[11℄, we introdue thehistory variable
z(s, t, x) := w(t − s, x)
for(s, t, x) ∈ [0, τ ] × [0, T ] × [0, l]
and obtain
z t (s, t, x) + z s (s, t, x) = 0
for(s, t, x) ∈ (0, τ ) × (0, T ) × (0, l).
Multiplying these identities with
w(t, ·)
inX
and performing apartial integration, we nd∂ t Z τ
0
kz(s, t, ·)k 2 X ds = − Z τ
0
∂ s kz(s, t, ·)k 2 X ds = kw(t, ·)k 2 X − kw(t − τ, ·)k 2 X .
(2.5)Addingequations (2.4 ) and (2.5 ) tothetrivial identity
∂ t kw(t, ·)k 2 X ≤ kw(t, ·)k 2 X + kw t (t, ·)k 2 X ,
weobtain
∂ t
kw(t, ·)k 2 X + kw t (t, ·)k 2 X + Z 0
−τ
kw(s, t, ·)k 2 X ds
≤ 2 + kAk L(X)
kw(t, ·)k 2 X +kw t (t, ·)k 2 X +kf (t, ·)k 2 X .
Thus,we have shown
∂ t E(t) ≤ 2 + kAk L(X)
E(t) + kf (t, ·)k 2 X ,
(2.6)where
E(t) := kw(t, ·)k 2 X + kw t (t, ·)k 2 X + Z 0
−τ
kz(s, t, ·)k X ds.
FromEquation(2.6) we onlude
E(t) ≤ E(0) + 2 + kAk L(X ) Z t
0
E(s)ds + Z t
0
kf (t, ·)k 2 X ds.
Using nowtheintegralform ofGronwall's inequality, we obtain
E(t) ≤ E(0) + Z t
0
kf (t, ·)k 2 X ds + Z t
0
e ( 2+kAk L(X) ) (t−s) E(0) +
Z s 0
kf (ξ, ·)k 2 X dξ
ds
≤ C ˜
E(0) + Z T
0
kf (s, ·)k 2 X ds
(2.7)
for ertain
C > ˜ 0
. Taking into aountc 1 kw(t, ·)k 2 X + |θ 1 (t)| 2 + |θ 2 (t)| 2
≤ kη(t, ·)k 2 X ≤ C 1 kw(t, ·)k 2 X + |θ 1 (t)| 2 + |θ 2 (t)| 2 , c 2 kf (t, ·)k 2 X + |θ 1 (t)| 2 + |θ 2 (t)| 2
≤ kg(t, ·)k 2 X ≤ C 2 kf (t, ·)k 2 X + |θ 1 (t)| 2 + |θ 2 (t)| 2
for some onstants
c 1 , c 2 , C 1 , C 2 > 0
and exploitingthe denitionofE(t)
,theproof is adiret onse-quene of Equation(2.7 ).
(ψ, g, θ 1 , θ 2 ) 7→ η
is well-dened,linear andontinuous inthenorms fromTheorem 2.3 .
Remark 2.5. It was essential to onsider the weak spae
X
. If the spae orresponding to the usual wave equationisused, i.e.,(η, η t ) ∈ H 0 1 (0, l)
× L 2 (0, l)
,there follows from[5℄ thatEquation
(2.1 ), (2.2 ) isan ill-posed problemdue to the lakof ontinuous dependene on thedata even in the
homogeneous ase.
Next, we want to establish onditions on the data allowing for the existene of a lassial solution.
Performing thesubstitution
ξ(t, x) := e − 2a b 2 x η(t, x)
for(t, x) ∈ [−τ, T ] × [0, l]
(2.8)with a new unknown funtion
ξ
(p. [11℄), the initial boundary value problem (2.1 ), (2.2) an bewritten inthefollowing simpliedform withaself-adjoint operator ontheright-handside
∂ tt ξ(t, x) = a 2 ∂ xx ξ(t − τ, x) + cξ(t − τ, x) + f (t, x)
for(t, x) ∈ (0, T ) × (0, l)
(2.9)with
c := d − 4a b 2 2
omplemented bythefollowing boundaryand initial onditionsξ(t, 0) = µ 1 (t), ξ(t, l) = µ 2 (t)
fort > −τ
withµ 1 (t) := θ 1 (t), µ 2 (t) := e 2a b 2 l θ 2 ,
(2.10)ξ(t, x) = ϕ(t, x)
for(t, x) ∈ (−τ, 0) × (0, l)
withϕ(t, x) := e 2a b 2 x ψ(t, x)
(2.11)and
f (t, x) := e 2a b 2 x g(t, x)
for(t, x) ∈ [0, T ] × [0, l].
The solutionwill bedetermined intheform
ξ(t, x) = ξ 0 (t, x) + ξ 1 (t, x) + G(t, x).
Here,
G
is an arbitrary funtion with∂ tt G, ∂ tx G, ∂ xx G ∈ C 0 [−τ, T ] × [0, l]
satisfying theboundary
onditions
G(t, 0) = µ 1 (t), G(t, l) = µ 2 (t).
Assuming
µ 1 , µ 2 ∈ C 2 [−τ, T ]
,we let
G(t, x) := µ 1 (t) + x l (µ 2 (t) − µ 1 (t))
for(t, x) ∈ [−τ, T ] × [0, l].
(2.12)• ξ 0
solvesthe homogeneous equation∂ tt ξ 0 (t, x) = a 2 ∂ xx ξ 0 (t − τ, x) + cξ 0 (t − τ, x)
(2.13)subjetto homogeneousboundary andnon-homogeneous initial onditions
ξ 0 (t, 0) ≡ 0, ξ 0 (t, l) = 0
in(−τ, T ),
ξ 0 (t, x) = Φ(t, x)
for(t, x) ∈ (−τ, 0) × (0, l)
withΦ(t, x) := ϕ(t, x) − G(t, x).
(2.14)In partiular,withthe funtion
G
seleted asinEquation(2.12 ), we obtainΦ(t, x) = ϕ(t, x) − µ 1 (t) − x l (µ 2 (t) − µ 1 (t))
for(t, x) ∈ [−τ, 0] × [0, l].
(2.15)• ξ 1
solvesthe non-homogeneous equation∂ tt ξ 1 (t, x) = a 2 ∂ xx ξ 1 (t − τ, x) + cξ 1 (t − τ, x) + F (t, x)
for(t, x) ∈ (0, T ) × (0, l)
(2.16)with
F(t, x) := a 2 ∂ xx G(t − τ, x) + cG(t − τ, x) − ∂ tt G(t, x)
for(t, x) ∈ [0, T ] × [0, l]
(2.17)subjetto homogeneousboundary andinitial onditions. For
G
fromEquation (2.12 ),we haveF(t, x) = f (t, x)+c µ 1 (t − τ ) + x l (µ 2 (t − τ ) − µ 1 (t − τ ))
− µ ¨ 1 (t) + x l (¨ µ 2 (t) − µ ¨ 1 (t))
.
(2.18)In this setion, we obtain a formal solution to the initial-boundary value problem (2.13 ) with initial
and boundary onditionsgiveninEquations(2.10),(2.11 ). Weexploit Fourier's separationmethod to
determine
ξ 0
intheprodutformξ 0 (t, x) = T (t)X(x)
. Afterplugging thisansatzinto Equation(2.13),wend
X(x) ¨ T (t) = a 2 X ′′ (x)T (t − τ ) + cX(x)T (t − τ ).
Hene,
X(x)
T ¨ (t) − cT (t − τ )
= a 2 X ′′ (x)T (t − τ ).
By formallyseparating the variables,we dedue
X ′′ (x)
X(x) = T ¨ (t) − cT (t − τ )
a 2 T (t − τ ) = −λ 2 .
Thus,theequation anbe deoupled asfollows
T ¨ (t) + a 2 λ 2 − c
T (t − τ ) = 0, X ′′ (x) + λ 2 X(x) = 0.
(3.1)These arelinearseondorder ordinary(delay) dierential equations withonstant oeients.
Due to thezero boundary onditions for
ξ 0
,the boundary onditions for theseond equation in(3.1 )will also be homogeneous, i.e.,
X(0) = 0, X(l) = 0.
Therefore, we obtain a Sturm &Liouville problem admitting non-trivial solutions only for the eigen-
numbers
λ 2 = λ 2 n = πn l n
for
n ∈ N
and theorresponding eigenfuntions
X n (x) = sin πn l x
for
n ∈ N .
Assuming
πa l
2
− c > 0,
wedenote
ω n = q
πn a
2
− c
forn ∈ N
and onsiderthe rst equationin(3.1 ), i.e.,
T ¨ (t) + ω 2 n T (t − τ ) = 0
forn ∈ N .
(3.2)The initial onditions foreah of theequations in(3.2 ) an beobtained byexpanding theinitial data
into a Fourier series withrespet totheeigenfuntion basisof theseondequation in(3.1 )
Φ(t, ·) =
∞
X
n=1
Φ n (t) sin πn l x
with
Φ n (t) = 2 l
Z l 0
(ϕ(t, s) − G(t, s)) sin πn l s ds,
∂ t Φ(t, ·) =
∞
X
n=1
˙Φ n (t) sin πn l x
with
Φ ˙ n (t) = 2 l
Z l 0
(∂ t ϕ(t, s) − ∂ t G(t, s)) sin πn l s ds
(3.3)
for
t ∈ [−τ, T ]
. Let us further determine the solution of theCauhy problem assoiated witheah oftheequations in(3.2 ) subjetto theinitial onditionsfrom Equation(3.3 ).
with pure delay obtained in [9 ℄. The authors onsidered a linear homogeneous seond order delay
dierential equation
¨
x(t) + ω 2 x(t − τ ) = 0
fort ∈ (0, ∞), x(t) = β(t)
fort ∈ [−τ, 0].
(3.4)Theyintrodued two speialfuntionsreferredto asdelayosine and sinefuntions. Exploitingthese
funtions, a uniquesolution to the initial valueproblem(3.4 ) wasobtained.
Denition 3.1. Delayosine isthe funtion given as
cos τ (ω, t) =
0, −∞ < t < −τ,
1, −τ ≤ t < 0,
1 − ω 2 t 2! 2 , 0 ≤ t < τ,
.
.
.
.
.
.
1 − ω 2 t 2! 2 + ω 4 ( t−τ) 4! 4 − · · · + (−1) k ω 2k (t−(k−1)τ) (2k)! 2k , (k − 1)τ ≤ t < kτ
(3.5)
with
2k
-order polynomials on eah of the intervals(k − 1)τ ≤ t < kτ
ontinuously adjusted at the nodest = kτ
,k ∈ N 0
.0 5 10 15
−1.5
−1
−0.5 0 0.5 1 1.5 2 2.5
t co s τ (0 . 5 , t )
τ = 0.1 τ = 0.05 τ = 0 . 5
0 5 10 15
−10
−5 0 5 10 15
t co s τ (1 , t )
τ = 0.1 τ = 0.05 τ = 0 . 5
0 5 10 15
−1000
−500 0 500 1000 1500 2000 2500
t co s τ (2 , t )
τ = 0.1 τ = 0.05 τ = 0 . 5
Figure1: Delayosine funtion
Denition 3.2. Delaysine isthe funtion givenas
cos τ (ω, t) =
0, −∞ < t < −τ,
ω(1 + τ ), −τ ≤ t < 0,
ω(1 + τ ) − ω 3 t 3! 3 , 0 ≤ t < τ,
.
.
.
.
.
.
ω(1 + τ ) − ω 3 t 3! 3 + · · · + (−1) k ω 2k+1 ( t−(k−1)τ) (2k+1)! 2k+1 , (k − 1)τ ≤ t < kτ
(3.6)
with
(2k + 1)
-order polynomials on eah of the intervals(k − 1)τ ≤ t < kτ
ontinuously adjusted at thenodest = kτ
,k ∈ N 0
.0 5 10 15
−2
−1.5
−1
−0.5 0 0.5 1 1.5
t si n τ (0 . 5 , t )
τ = 0.1 τ = 0.05 τ = 0 . 5
0 5 10 15
−15
−10
−5 0 5 10
t si n τ (1 , t )
τ = 0.1 τ = 0.05 τ = 0.5
0 5 10 15
−2000
−1000 0 1000 2000 3000 4000 5000
t si n τ (2 , t )
τ = 0.1 τ = 0.05 τ = 0 . 5
Figure2: Delaysine funtion
There has furtherbeen provedthat delayosine uniquely solvesthelinearhomogeneous seondorder
ordinarydelaydierentialequationwithpuredelaysubjettotheunitinitialonditions
x ≡ 1
in[−τ, 0]
and thedelaysine inits turnsolvesEquation(3.4) subjettotheinitial onditions
x(t) = ω(t + τ )
fort ∈ [−τ, 0]
.Using the fat above, the solution of the Cauhy problem was represented in the integral form. In
partiular,thesolution
x
tothehomogeneousdelaydierentialequation(3.4 )withtheinitialonditionsx ≡ β
in[−τ, 0]
for anarbitraryβ ∈ C 2 ([−τ, 0])
wasshownto be given asx(t) = β(−τ ) cos τ (ω, t) + ω 1 β(−τ ˙ ) sin τ (ω, t) + ω 1
Z 0
−τ
sin τ (ω, t − τ − s) ¨ β(s)ds.
(3.7)Turning bak to the delaydierential equation (3.2) withtheinitial onditions (2.4 ), we obtain their
unique solution inthe form
T n (t) = Φ n (−τ ) cos τ (ω n , t) + ω 1
n ˙Φ n (−τ ) sin τ (ω n , t) + ω 1
n
Z 0
−τ
sin τ (ω n , t − τ − s) ¨ Φ(s)ds.
(3.8)Thus,assuming suient smoothnessof the datato be speied later,thesolution
ξ 0
tothehomoge-neousequation(2.13 )satisfyinghomogeneousboundaryandnon-homogeneousinitialonditions
ξ ≡ Φ
in
[−τ, 0] × [0, l]
reads asξ 0 (t, x) =
∞
X
n=1
Φ n (−τ ) cos τ (ω n , t) + ω 1
n
Φ ˙ n (−τ ) sin τ (ω n , t)+
1 ω n
Z 0
−τ
sin τ (ω n , t − τ − s) ¨ Φ n (s)ds
sin πn l x , Φ n (t) = 2
l Z l
0
(ϕ(t, s) − G(t, s)) sin πn l s
ds
forn ∈ N .
(3.9)
Next,weonsiderthenon-homogeneous equation(2.16 )withtheright-handsidefromEquation(2.18 )
subjetto homogeneousinitial and boundaryonditions
∂ tt ξ 1 (t, x) = a 2 ∂ xx ξ 1 (t − τ, x) + cξ 1 (t − τ, x) + F (t, x)
for(t, x) ∈ (0, T ) × (0, l),
where
F (t, x) = f (t, x) + c µ 1 (t − τ ) + x l (µ 2 (t − τ ) − µ 1 (t − τ ))
− µ ¨ 1 (t) − x l (¨ µ 2 (t) − µ ¨ 1 (t)) .
The solution will be onstruted as asFourier series with respet to theeigenfuntions of the Sturm
&Liouville problemfrom theprevious setion,i.e.,
ξ 1 (t, x) =
∞
X
n=1
T n (t) sin πn l x
.
(4.1)Plugging theansatz from (4.1) into Equation (2.6) and omparingthe time-dependent Fourier oe-
ients, we obtaina systemofountablymany seondorder delaydierential equations
T ¨ n (t) + ω n T n (t − τ ) = F n (t)
fort ∈ (0, T )
withF n (t) = 2 l
Z l 0
F (t, s) sin πn l x
ds.
(4.2)In [9℄,theinitial valueproblemfor thenon-homogeneous delay dierential equation
¨
x(t) + ω 2 x(t − τ ) = f (t)
fort ≥ 0
withhomogeneous initial onditions
x ≡ 0
in[−τ, 0]
wasshownto be uniquely solved byx(t) =
Z t 0
sin τ (ω, t − τ − s)f (s)ds.
(4.3)Exploiting Equation(4.3 ),the equations in(4.2 ) subjetto zeroinitial onditions areuniquely solved
by
T n (t) = Z t
0
sin τ (ω n , t − τ − s)F n (s)ds.
(4.4)Therefore, the non-homogeneous partial delay dierential equation with homogeneous initial and
boundary onditionformallyreads as
ξ 1 (t, x) =
∞
X
n=1
Z t 0
sin τ (ω n , t − τ − s)F n (s)ds
sin πn l x
for
(t, x) ∈ [0, T ] × [0, l].
(4.5)5 General ase solution
The solutioninthe general ase an thus formally be representedasthefollowing series
ξ(t, x) =
∞
X
n=1
Φ n (−τ ) cos τ (ω n , t) + ω n 1 ˙Φ n (−τ ) sin τ (ω n , t)+
1 ω n
Z 0
−τ
sin τ (ω n , t − τ − s) ¨ Φ n (s)ds
sin πn l x +
∞
X
n=1
Z t 0
sin τ (ω n , t − τ − s)F n (s)ds
sin πn l x
+ G(t, x).
(5.1)
Theorem 5.1. Let
T > 0
,τ > 0
andm := ⌈ T τ ⌉
. Further, let the datafuntionsϕ
,µ 1
,µ 2
andf
besuhthattheir Fourier oeients
Φ n
andF n
giveninEquations(3.3 )and(4.5 ) satisfytheonditionsn→∞ lim
|Φ n (−τ )| + | Φ ˙ n (−τ )|
n 2m+3+α = 0, lim
n→∞ max
s∈[−τ,0] | Φ ¨ n |n 2m+3+α = 0,
n→∞ lim max
k=1,...,m max
t∈[(k−1)τ,max{kτ,T }] |F n (t)|n 2k+3+α = 0
(5.2)
for an arbitrary,but xed
α > 0
. Letπ l a 2
> c.
Then thelassialsolutionto problem(2.9 )(2.11 ) anberepresentedasan absolutelyand uniformly
onvergent FourierseriesgiveninEquation(5.1 ). Thelatterseriesisatwieontinuouslydierentiable
funtion withrespetto both variables. Itsderivativesoforderlessorequal twowithrespetto
t
andx
anbeobtainedbyaterm-wisedierentiationoftheseriesandtheresultingseriesarealsoabsolutely and uniformlyonvergent in[0, T ] × [0, l]
.Proof. Weregroup the series fromEquation(5.1) into the following sum
ξ(t, x) = S 1 (t, x) + S 2 (t, x) + S 3 (t, x) + G(t, x),
where
S 1 (t, x) =
∞
X
n=1
A n (t) sin πn l x
, S 2 (t, x) =
∞
X
n=1
B n (t) sin πn l x
, S 3 (t, x) =
∞
X
n=1
C n (t) sin πn l x , A n (t) = Φ n (−τ ) cos τ (ω n , t) + ω 1
n ˙Φ n (−τ ) sin τ (ω n , t), B n (t) = ω 1
n
Z 0
−τ
sin τ (ω n , t − τ − s) ¨ Φ n ds, C n (t) = ω 1
n
Z 0
−τ
sin τ (ω n , t − τ − s)F n (s)ds
and
ω n = q
πn l a 2
− c
forn ∈ N .
1. First,weonsidertheoeient funtions
A n
. Foranarbitraryt ∈ [0, T ]
with(k − 1)τ ≤ t < kτ
,we nd
A n (t) = Φ n (−τ ) cos τ (ω n , t) + ω 1
n ˙Φ n (−τ ) sin τ (ω n , t)
=
1 − πn l a 2 t 3
2! + · · · + (−1) k πn l a 2k (t−(k−1)τ) 2k (2k)!
Φ n (−τ )+
(1 + τ ) − πn l a 2 t 2
3! + · · · + (−1) k πn l a 2k (t−(k−1)τ ) 2k+1 (2k+1)!
Φ n (−τ ).
If
Φ n (−τ )
and˙Φ n (−τ )
,n ∈ N
,aresuhthat the onditionn→∞ lim
|Φ n (−τ )| + | ˙Φ n (−τ )|
n 2k+3+α = 0
holdstrue, the series
S 1
aswellasits derivativesoforderlessorequal 2onvergeabsolutelyanduniformly. Note thatasingledierentiation withrespetto
x
orresponds,roughlyspeaking, to a multipliationwithn
.2. Next, we onsider the oeients
B n
. For an arbitraryt ∈ [0, T ]
with(k − 1)τ ≤ t < kτ
, weperform the substitution
t − τ − s = ξ
and exploitthe meanvaluetheoremto estimate|B n (t)| =
1 ω n
Z t t−τ
sin τ (ω n , ξ) ¨ Φ n (t − τ − ξ)dξ
≤ τ max
−τ≤s≤0 | Φ ¨ n (s)| max
j=k−1,k max
t−τ≤s≤t
(s − τ ) − πn l a 2 s 3 3! + . . . +(−1) j πn l a 2j (s−(j−1)τ) 2j+1
(2j+1)!
.
Applying the theorem on dierentiation under the integral sign to
B n
and taking into aountthat
sin τ πn l a, ·
is twieweakly dierentiable in
[0, ∞)
,namely:sin τ πn l a, ·
∈ W loc 2,∞ (0, ∞)
,
its derivatives are polynomials of order lower than those of
sin τ πn l a, ·
and their onvolution
with
Φ ¨ n
isontinuous, analogous estimates an be obtainedfor˙Φ n
andΦ ¨ n
whih,intheir turn,also followto be ontinuousfuntions.
Now, ifthe ondition
n→∞ lim max
s∈[−τ,0] | Φ ¨ n |n 2m+3+α = 0
issatised,theseries
S 2
aswellasits derivativesoforderlessorequal2onvergeabsolutelyanduniformly.
3. Finally,we lookat theFourier oeients
C n
. Again, for anarbitrary period of timet ∈ [0, T ]
with
(k − 1)t ≤ t < kτ
,0 ≤ k ≤ m
, we substitutet − τ − ξ = s
. One again, using the meanvaluetheorem, we estimate
|C n (t)| =
1 ω n
Z t t−τ
sin τ πn l a, ξ
F n (t − τ − ξ)dξ
≤ τ max
t−τ≤s≤t | Φ ¨ n (s)| max
j=k−1,k max
t−τ≤s≤t
(s − τ ) − πn l a 2 s 3 3! + . . . +(−1) j πn l a 2j (s−(j−1)τ) 2j+1
(2j+1)!
.
Asbefore,
C n
an beshown to be twie ontinuouslydierentiable. If nown→∞ lim max
k=1,...,m max
t∈[(k−1)τ,max{kτ,T }] |F n (t)|n 2k+3+α = 0
(5.3)is satised, then both
S 3
and its derivatives of order less or equal 2 onverge absolutely and uniformly.Sine all three onditions areguaranteed bythe assumptions of theTheorem due to thefat
k ≤ m
,theproof isnished.
Remark 5.2. From the pratial point of view,the rapid deayondition onthe Fourier oeients
of thedata given in Equation(5.2 ) mean asuiently highSobolev regularityof thedata and orre-
sponding higherorder ompatibilityonditions at theboundaryof
(0, l)
(f. [11℄).Referenes
[1℄ Adams, R.A. Sobolev spaes, Pure and Applied Mathematis, Vol. 65, New York-London: Aa-
demi Press,1975
ofTarasShevhenko National Universityof Kyiv,Series: Cybernetis,12, 2012, pp.414
[3℄ Bátkai, A.,Piazzera, S.Semigroups for DelayEquations,ResarhNotes inMathematis, 10,A.K.
Peters: Wellesley MA,2005
[4℄ Datko,R.Two examplesof ill-posedness withrespetto timedelays revisited.IEEE Trans. Auto-
matiControl, 50,1997, pp.1374 1379
[5℄ Dreher,M., Quintanilla,R., Rake,R.Ill-posedproblems inthermomehanis. Appl. Math.Lett.,
22(9),2009, pp.13741379
[6℄ Ek,Ch.,Garke,H.,Knabber,P.MathematisheModellierung.Springer-VerlagBerlinHeidelberg,
2008
[7℄ Els'gol'ts,L.E.,Norkin,S.B.IntrodutiontotheTheoryandAppliationofDierentialEquations
withDeviating Arguments. MathematisinSiene and Engineering, 105,AademiPress, 1973,
pp.1357
[8℄ Hale,J.K. Theoryof Funtional Dierential Equations, Applied Mathematial SienesSeries, 3,
1977,pp.1365
[9℄ Khusainov, D.Ya., Diblík, J., R·ºi£kova, M., Luká£ová, J. Representation of a solution of the
Cauhyproblemforan osillatingsystemwithpuredelay.NonlinearOsillations,11(2),2008, pp.
276285
[10℄ Khusainov, D.Ya., Ivanov, A.F., Kovarzh, I.V. The solution of wave equation with delay (in
Ukrainian with English summary), Bulletin of Taras Shevhenko National University of Kyiv.
Series: Physisand Mathematis, 4,2006, pp.243248
[11℄ Khusainov, D.Ya., Pokojovy, M., Azizbayov, E, On Classial Solvability for a Linear 1D Heat
EquationwithConstantDelay,toappearin: JournalofComputationalandApplied Mathematis,
2013
[12℄ Rake,R.Instabilityofoupledsystemswithdelay,Commun.Pure.Appl.Anal.,11(5),2012,pp.
17531773
[13℄ Tikhonov, A.N., Samarskii,A.A. Equations ofMathematial Physis. DoverPubliations, 1990,
pp.1765