• Aucun résultat trouvé

Summability of solutions of the heat equation with inhomogeneous thermal conductivity in two variables

N/A
N/A
Protected

Academic year: 2021

Partager "Summability of solutions of the heat equation with inhomogeneous thermal conductivity in two variables"

Copied!
21
0
0

Texte intégral

(1)

HAL Id: hal-00364722

https://hal.archives-ouvertes.fr/hal-00364722

Submitted on 26 Feb 2009

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Summability of solutions of the heat equation with inhomogeneous thermal conductivity in two variables

Werner Balser, Michèle Loday-Richaud

To cite this version:

Werner Balser, Michèle Loday-Richaud. Summability of solutions of the heat equation with inhomo- geneous thermal conductivity in two variables. Advances in Dynamical Systems and Applications, Dehli : Research India Publ., 2009, 4 (2), pp.159-177. �hal-00364722�

(2)

inhomogeneous thermal ondutivity in two variables

Werner BALSER

Abteilung Angewandte Analysis

Universitat Ulm, D-89069 ULM, Germany

Email: balsermathematik.uni-ulm.de

Mihele LODAY-RICHAUD

LAREMA, Universite d'Angers, 2 boulevard Lavoisier

49 045 ANGERS edex 01, Frane

Email: mihele.lodayuniv-angers.fr

February 26, 2009

Abstrat

WeinvestigateGevreyorderand1-summabilitypropertiesoftheformalsolution

of a general heat equation in two variables. In partiular, we give neessary and

suÆientonditionsforthe1-summabilityofthesolutioninagivendiretion. When

restritedto thease of onstantsoeÆients, theseonditions oinide withthose

givenbyD.A.Lutz,M.Miyake,R.Shafkeina1999 artile([LMS99 ℄),andwethus

providea newproofof theirresult.

Keywords: Heatequation, Gevreyseries, 1-summability.

AMS lassiation: 35C10, 35C20, 35K05,40-99, 40B05.

Contents

1 The problem 2

2 Gevrey properties 4

3 1-summability 7

(3)

4 Initial onditions 15

4.1 Casea(z)=a2C

. . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

4.2 Casea(z)=bz; b2C

. . . . . . . . . . . . . . . . . . . . . . . . . . 17

1 The problem

A formalsolutionofthe lassialheatinitialonditions problem

(1)

(

t

u

2

z u=0

u(0;z)='(z)

inone dimensionalspatialvariablez reads intheform

b

u(t;z) = exp t 2

z t

'(z)

= X

j0 t

j

j!

' (2j)

(z)

providedthatall derivatives' (2j)

exist 1

. When'2O(D

p

) isholomorphiinadis

D

with enter 0 and radius and hene satises, for any r < , estimates of the

form

' (2j)

(z)

CK 2j

(1+2j)!;

forall j 0 and positive onstantsC and K, on D

r

then, u(t;b z) 2O(D

)[[t℄℄ is a

seriesofGevreytypeoforder1intforallz2D

(inshort,a1-Gevrey series). The

Gevreyestimates areloally uniformwithrespettoz inD

. These onditionsare

optimalasshown bythe followingexample: Letonsider'(z)= 1

1 z

= X

n0 z

n

so

that ' (2j)

(0) =(2j)!. The orresponding solution bu(t;z) is of exat Gevrey order

1 and, in partiular, is divergent. It turns out that it is atually 1-summable in

all diretion but R +

in the sense of Denition3.1 below, that is, 1-summable in t

uniformallywithrespetto znear 0.

In 1999, D. Lutz, M. Miyake and R. Shafke ([LMS99 ℄) gave neessary and

suÆient onditions on ' for ub to be 1-summable in a given diretion argt = .

Various works have been done towards the summability of divergent solutions of

partialdierential equations with onstant oeÆients ( [Bal99 ℄, [Miy99 ℄, [BM99℄,

[Bal04 ℄,...) or variable oeÆients ([H99℄, [Ou02 ℄, [PZ97 ℄, [Mk08℄, [Mk09℄,...)

1

Wedenotebu,withahat,toemphasizethepossibledivergeneof theseriesu.b

(4)

in two variables. In [Mk05 ℄, S. Malek has investigated the ase of linear partial

dierentialequationswithonstant oeÆientsinmore variables.

Inthisartileweareinterestedintheverygeneralheatinitialonditionsproblem

withinhomogeneous thermalondutivityand internalheat generation

(2)

(

t

u a(z) 2

z

u=q(t;z) a(z)2O(D

)

u(0;z)='(z) 2O(D

):

The heat equation desribes heat propagation under thermodynamis and Fourier

laws. The oeÆient a(z), named thermal diusivity, is related to the thermal

ondutivity by the formula a =

where is the apaity and the density

of the medium. We assume that a(z) and '(z) are analyti on a neighborhood

of z = 0. The internal heat input q may be smooth or not. An important ase

is the ase with no internal heat generationorresponding to a homogeneous heat

equation:

(3)

(

t

u a(z) 2

z

u=0 a(z)2O(D

)

u(0;z)='(z)2O(D

):

Inaseofan isotropiandhomogeneousmedium,;; andheneaareonstants.

An adequatehoie of unitsallows thento assume a=1and the equation redues

to thereferene heatequation

t

u

2

z u=0.

Atually,fornotationalonveniene, weonsidertheproblemintheform

(4) 1 a(z) 1

t

2

z

b u=

b

f(t;z) ; a(z)2O(D

) and

b

f(t;z)2O(D

)[[t℄℄

where 1

t b

u standsforthe anti-derivative R

t

0 b

u(s;z)ds of ubwith respetto t whih

vanishesat t=0.

Problem (4) isequivalent to

(

t b

u a(z) 2

z b u=

t b

f(t;z)

b

u(0;z) = b

f(0;z):

and heneto Problem (2)byhoosingq(t;z)=

t b

f(t;z) and '(z)= b

f(0;z).

Moreover, Problem (4) redues to the homogeneous ase (3) if and only if the

b

(5)

Fromnow, we denoteD=1 a(z) 1

t

2

z

and, given a seriesub2O(D

)[[t℄℄, we

denote

b

u(t;z)= X

j0 t

j

j!

u

j;

(z)= X

n0 b u

;n (t)

z n

n!

= X

j;n0 b u

j;n t

j

j!

z n

n!

Sine O(D

)[[t℄℄;

t

;

z

isadierentialalgebraanda(z)2O(D

) theoperator

Dats insideO(D

)[[t℄℄. More preisely,we an state:

Proposition 1.1 The map

D: O(D

)[[t℄℄ !O(D

)[[t℄℄

isa linear isomorphism.

Proof. The operator D is linear. A seriesu(t;b z) = X

j0 t

j

j!

b u

j;

(z) is a solutionof

Problem (4)ifand onlyif

(5) ub

j;

(z)= b

f

j;

(z)+a(z)ub 00

j 1;

(z) forallj 0 startingfromub

1;

(z)0:

Consequently,toany b

f(t;z)2O(D

[[t℄℄thereisauniquesolutionbu(t;z)2O(D

[[t℄℄,

whih provesthat Dis bijetive. 2

In Setion 2 we show that the inhomogenuity b

f(t;z) and the unique solution

b

u(t;z) aretogether 1-Gevrey.

InSetion3weproveneessaryandsuÆientonditionsforubtobe1-summable

in a given diretion argt = . The onditions are valid in the ase when either

a(0) 6= 0 or a 0

(0) 6= 0. When a(z) = O(z 2

) an easy ounter-example shows that

even therationalityof b

f(t;z) isinsuÆient.

InSetion4wedisusstheaessibilityofourneessaryandsuÆientonditions.

Indeed, the onditions aregiven not onlyin terms of the data b

f butalso interms

ofthe rst twoterms ub

;0 and bu

;1

of thesolutionubitself.

Inthepartiularasea=1ouronditionsoinidewiththoseof[LMS99 ℄. Wethus

providea newproofof theresult of[LMS99℄.

2 Gevrey properties

In this artile, we onsider t as the variable and z as a parameter. The lassial

(6)

Denition 2.1 (1-Gevrey series) Aseries bu(t;z)= X

j0 t

j

j!

b u

j;

(z)2O(D

)[[t℄℄ is

of Gevrey type of order 1 if there exist 0 <r ; C >0;K >0 suh that for all

j0 and jzjr we have

jbu

j;

(z)jCK j

(1+2j):

In other words, u(t;b z) is 1-Gevrey in t, uniformally in z on a neighbourhood of

z=0.

We denote O(D

)[[t℄℄

1

the subsetof O(D

)[[t℄℄ made of the series whih are of

Gevreytypeof order 1.

Proposition 2.2 O(D

)[[t℄℄

1

;

t

;

z

isa dierentialalgebra stableunder 1

t and

1

z .

Proof. The proof is similarto theone withoutparameter. Stabilityunder

z is

provedusingtheCauhyIntegralFormulaandisguarantedbytheondition\there

exist r :::"inDenition 2.1. 2

It resultsfromthisPropositionthattheoperatorD=1 a(z) 1

t

2

z

atsinside

thespae O(D

)[[t℄℄

1 .

Beause the mainresult of this setion (Theorem 2.5) is set up using Nagumo

normson O(D

) we beginwitha reallof theirdenitionandmainpropertiesand

we refer to [N42 ℄orto [CRSS00℄formore details.

Denition 2.3 (Nagumo norms)

Let f 2 O(D

), p 0; 0 < r and let d

r

(z) = jzj r denote the eulidian

distane of z to the boundary of the dis D

r .

The Nagumo norm kfk

p;r

of f isdened by

kfk

p;r

= sup

jzj<r

f(z)d

r (z)

p

:

Proposition 2.4 (Properties of Nagumo norms)

(i) k:k

p;r

is a norm on O(D

);

(ii) For all z2D ; jf(z)jkfk d(z) p

;

(7)

(iii) kfk

0;r

=sup

z2D

r

jf(z)j is the usualsup-norm on D

r

;

(iv)kfgk

p+q;r kfk

p;r kgk

q;r

;

(v)(most important) kf 0

k

p+1;r

e(p+1)kfk

p;r .

Notethatthesameindexr oursonbothsidesoftheinequality(v). Onegetsthus

an estimateof thederivativef 0

intermsof f withouthavingto shrinkthe domain

D

r .

Theorem 2.5 Themap

D: (

O(D

)[[t℄℄

1

! O(D

)[[t℄℄

1

b

u(t;z) 7!

b

f(t;z)=Dbu(t;z)

isa linear isomorphism.

Proof. It results from Proposition 2.2 that D O(D

)[[t℄℄

1

O(D

)[[t℄℄

1 and

from Proposition1.1 that Dis linear and injetive. We areleft to prove that Dis

also surjetive.

Let b

f(t;z)= X

j0 t

j

j!

b

f

j;

(z)2O(D

)[[t℄℄

1

. The oeÆients b

f

j;

(z) satisfy

8

>

>

<

>

>

:

b

f

j;

(z)2O(D

)forall j0:

There exist0<r; C>0; K>0 suh thatforall j0and jzjr

j b

f

j;

(z)jCK j

(1+2j)!

and we look forward to similar onditions on the oeÆients bu

j;

(z) of bu(t;z) =

X

j0 t

j

j!

b u

j;

(z).

Fromthe reurrenerelation(5)therelation

b u

j;

(z)

(1+2j)

= b

f

j;

(z)

(1+2j)

+a(z) b u 00

j 1;

(z)

(1+2j)

starting from ub

1;

(z) 0 holds for all j 0. Applying the Nagumo norms of

indies(2j;r) and properties (iv)and (v)of Proposition2.4weget

kbu

j;

(z)k

2j;r

(1+2j)

k b

f

j;

(z)k

2j;r

(1+2j)

+ ka(z)k

0;r kbu

00

j 1;

(z)k

2j;r

(1+2j)

00

+ ka(z)k

0;r e

2 kbu

j 1;

(z)k

2j 2;r

(8)

Denote g

j

= k

b

f

j;

(z)k

2j;r

(1+2j)

and=ka(z)k

0;r e

2

andonsiderthenumerialsequene

(

v

1

=0

v

j

=g

j +v

j 1

forall j0:

By onstrution, kbu

j;

(z)k

2j;r

(1+2j) v

j

forallj 0.

Letusboundv

j

asfollows. Byassumption,0g

j

CK j

(1+2j)

(1+2j) r

2j

=C(Kr 2

) j

for all j and the series g(X) = P

j0 g

j X

j

is onvergent. Due to the reurrene

relationdeningthev

j

'stheseriesv(X)= P

j0 v

j X

j

satisfy(1 X)v(X) =g(X).

ItisthenonvergentandthereexistonstantsC 0

>0;K 0

>0suhthatv

j C

0

K 0

j

forall j. Hene,

kbu

j;

(z)k

2j;r C

0

K 0

j

(1+2j) forallj 0:

Wededueasimilarestimateonthesup-normbyshrinkingthedomainD

r

. Indeed,

let 0<r 0

<r. Forall j0and z2D

r 0

,

jbu

j;

(z)j =

bu

j;

(z)d

r (z)

2j 1

d

r (z)

2j

1

(r r 0

) 2j

b u

j;

(z)d

r (z)

2j

Hene,

sup

z2D

r 0

jbu

j;

(z)j

1

(r r 0

) 2j

kbu

j;

k

2j;r

C

0

K 0

(r r 0

) 2

j

(1+2j)

2

3 1-summability

Stillonsidering t as the variable and z as a parameter, one extends the lassial

notionsofsummabilitytofamiliesparameterizedbyzinrequiringsimilaronditions,

theestimates beinghoweveruniformwithrespettotheparameterz. Forageneral

studyof serieswithoeÆients ina Banah spae we refer to [Bal00 ℄. Among the

(9)

we hoose here a generalization of Ramis denitionwhih states that a series b

f is

1-summablein the diretion if there exists a holomorphifuntion f whih is 1-

Gevreyasymptoti to b

f onan open setor

;>

bisetedby withopeninglarger

than (f. [R80 ℄ Def 3.1). There are various equivalent ways of expressing the

1-Gevrey asymptotiity. We hooseto extend theone whih sets onditionson the

suessive derivativesof f (see [Mal95 ℄ p. 171 or[R80 ℄ Thm2.4, forinstane).

Denition 3.1 (1-summability) A series u(t;b z) 2 O(D

)[[t℄℄ is 1-summable in

the diretion argt = if there exist a setor

;>

, a radius 0 < r and a

funtionu(t;z) alled 1-sumof bu(t;z) inthediretion suh that

1. u isdened and holomorphi on

;>

D

r

;

2. For any z 2 D

r

the map t 7! u(t;z) has u(t;b z) = X

j0 t

j

j!

b u

j;

(z) as Taylor

series at0 on

;>

;

3. For any proper 2

subsetor

;>

there exist onstants C > 0;K > 0

suh that for all `0, all t2 andz2D

r

`

t u(t;z)

CK

`

(1+2`)

:

We denoteO(D

)fftgg

1;

thesubsetofO(D

)[[t℄℄made ofall 1-summableseriesin

thediretionargt=. Atually,O(D

)fftgg

1;

is inludedinO(D

)[[t℄℄

1 .

For any xed z 2 D

r

, the 1-summabilty of the series u(t;b z) is the lassial

1-summabilityand Watson Lemma impliestheuniityof its1-sum, ifany.

Proposition 3.2 O(D

)fftgg

1;

;

t

;

z

is a dierential C-al gebra stable under

1

t

and 1

z .

Proof. Letbu(t;z)andbv(t;z)betwo1-summableseriesindiretion. InDenition

3.1weanhoosethesameonstantsr;C ;Kbothforubandv.b Theprodutw(t;z)=

2

Inthis ontexta subsetorof asetor 0

is saidapropersubsetor andonedenotes 0

if

itslosureinC isontainedin 0

[f0g.

(10)

u(t;z)v(t;z) satisesonditions1 and 2 ofDenition3.1. Moreover,

`

t w(t;z)

=

`

X

p=0

`

p

p

t

u(t;z)

` p

t

v(t;z)

C

2

K

`

(1+2`)

`

X

p=0

(1+`)

(1+2`)

(1+2p)

(1+p)

1+2(` p)

1+(` p)

C

2

K

`

(`+1) (1+2`)

C

0

K 0

`

(1+2`) for adequateC 0

;K 0

>0:

Thisprovesondition3ofDenition3.1forw(t;z),thatis,stabilityofO(D

)fftgg

1;

undermultipliation.

Stability under

t ,

1

t

or 1

z

is straightforward. Stability under

z

is obtained

usingtheIntegral Cauhy Formula on adisD

r 0

withr 0

<r. 2

Wemaynotiethatthe1-sumu(t;z)ofa1-summableseriesu(t;b z) 2O(D

)fftgg

1;

may be analytiwith respet to z on a disD

r

smaller than theommon dis D

of analytiity of the oeÆients ub

j;

(z) of u(t;b z) = X

j0 t

j

j!

b u

j;

(z). With respet to

t, the1-sumu(t;z) isanalytiona setorsupposedlyopenand ontaininga losed

setor

;

biseted bywithopening;thereisnoontrolon theangularopening

exept that it must be larger than and no ontrol on the radius of this setor

exeptthat itmustbe positive. Thus,the 1-sumu(t;z) iswelldened asasetion

of the sheaf of analyti funtions in (t;z) on a germ of losed setor of opening

(i.e., alosed intervalI

;

of length on theirle S 1

ofdiretions issuingfrom 0,

f. [MalR92 ℄ 1.1 or[L-R94 ℄ I.2)times f0g C

z

. We denote O

I

; f0g

thespae of

suh setions.

Corollary 3.3 The operator of 1-summation

S: (

O(D

)fftgg

1;

! O

I

; f0g

b

u(t;z) 7! u(t;z)

is a homomorphism of dierential C-al gebras for the derivations

t and

z

and it

ommutes with 1

t

and 1

z .

(11)

Theorem 3.4

Let a diretion argt= issuing from 0 and a series b

f(t;z)2O(D

)[[t℄℄ begiven.

ReallD=1 a(z) 1

t

2

z

andassumethateithera(0)6=0ora(0)=0anda 0

(0)6=0.

Then, the unique solution bu(t;z) of Dbu= b

f in O(D

)[[t℄℄ is 1-summable in the

diretion if and only if bu

;0 (t);ub

;1

(t) and b

f(t;z) are1-summable in the diretion

.

Moreover, the 1-sum u(t;z), if any, satises equation (4) in whih b

f(t;z) is

replaed by the 1-sum f(t;z) of b

f(t;z) in diretion .

Proof. Werst plae ourselvesinthe asea(0)6=0.

Denote a(z)= X

n0 a

n z

n

.

As a preliminary remark we notie that, by identiation of equal powers of z in

Equation

(4) 1 a(z)

1

t

2

z

X

n0 b u

;n (t)

z n

n!

= X

n0 b

f

;n (t)

z n

n!

;

we get

8

>

<

>

: b u

;0 (t) a

0

1

t b u

;2 (t)=

b

f

;0 (t)

b u

;1 (t) a

1

1

t b u

;2 (t) a

0

1

t b u

;3 (t)=

b

f

;1 (t)

and soon:::

so that eah ub

;n

(t) is uniquely and linearly determined from ub

;0 (t); ub

;1

(t) and

b

f(t;z).

TheonditionisneessarybyProposition3.2. Indeed,ifubis1-summablethen

so arebu

;0

(t)=bu(t;0); bu

;1 (t)=

1

z b

u(t;z) bu

;0 (t)

z=0 and

b

f =Du.

Prove thattheonditionis suÆient. Assume thatub

;0 (t); ub

;1

(t) and b

f(t;z)

are 1-summableindiretion.

Set bu(t;z)=ub

;0

(t)+zub

;1 (t)+

2

z b

v(t;z) and wb= 1

t b v.

With these notations Equation(4) beomes

(6)

1 1

a(z)

t

2

z

b

w(t;z)=bg(t;z) where bg= 1

a(z) (bu

;0 +zbu

;1 b

f)

and itsuÆesto prove thatwbis1-summableindiretion whenbg is. Tothis

end, we proeedthrougha xed pointmethod asfollows.

Références

Documents relatifs

For several years, various works have been done on the divergent solutions of some classes of linear partial differential equations or integro-differential equations in two variables

that v admits an initial trace which belongs to M + (G). Proof of Theorem 6.2.. Step 1: Existence of a maximal solution. Existence is standard as well as uniqueness in the case where

By opposition to the power-like case [5], where the initial trace was constructed by duality arguments based upon H¨older inequality and delicate choice of test func- tions, our

Depending on the tail of J, we give a rather complete picture of the problem in optimal classes of data by: (i) estimating the initial trace of (possibly unbounded) solutions;

7.2 Representation formula for positive solutions. 65 Key words: Nonlinear parabolic equation; Initial trace; Representation formula; Bessel capacities; Borel measure; fine

In order to use Theorem 1.1 to construct sign-changing solutions of (1.1) with a singularity at x = 0, we consider separately the cases where U is a radially symmetric

Zambrini, Probabilistic deformation of contact geometry, diffusion pro- cesses and their quadratures, Seminar on Stochastic Analysis, Random Fields and applications V, 203-226,

It has only been in diamonds with polished surfaces that a nearly temperature independent boun- dary scattering with a phonon mean free path 2 to 3 times the Casimir value has