• Keine Ergebnisse gefunden

Towards an L1-theory for vector-valued elliptic boundary value problems

N/A
N/A
Protected

Academic year: 2022

Aktie "Towards an L1-theory for vector-valued elliptic boundary value problems"

Copied!
6
0
0

Wird geladen.... (Jetzt Volltext ansehen)

Volltext

(1)

TOWARDS AN L 1

-THEORY FOR VECTOR-VALUED ELLIPTIC

BOUNDARY VALUE PROBLEMS

ROBERTDENK, MATTHIAS HIEBER,ANDJAN PR



USS

1. Introduction

Vector-valuedellipticandparabolicboundaryvalueproblemssubjecttogeneralboundarycon-

ditions have been investigated recently in [DHP01] in the L

p

-context for 1 < p < 1. One

ofthemain goalsof this paperwasto deduce amaximal L

p

-regularityresultforthe solution

of the parabolic initial boundaryvalue problem. A classical reference in the elliptic context

are thecelebratedpapersof Agmon, Douglis and Nirenberg [ADN59]. Forfurther references

and informationon the scalarand vector-valuedcase we referto the [Ama01] and thelist of

referencesgivenin [DHP01].

Vector-valued elliptic and parabolicproblems on all of R n

we considered rst by Amann

[Ama01]onalargescaleoffunction spaces,includingL

1 (R

n

;E). HereE denotesanarbitrary

Banach space. He proved in particular that the L

1

-realization of such problems generates

an analytic C

0

-semigroup provided the top-order coeÆcients of the underlying operators are

uniformlybounded andHoldercontinuous.

Inthis note, weconsider vector-valued boundaryvalueproblems with constant coeÆcients

in the L

1

-setting for a half space. Following the approach described in [DHP01], we assume

the Lopatinskii-Shapiro to be true; we then obtaina representation of the solution u of the

elliptic problem by integraloperators which allows to deduce a-prioriestimates for u in the

L

1 (R

n+1

+

;E)-norm. HereEdenotesagainanarbitraryBanachspace. Theseestimatesimplyin

particularthattheL

1 (R

n+1

+

;E)-realizationofanellipticboundaryvalueproblemwithconstant

coeÆcientsin the half space R n+1

+

generatesan analytic C

0

-semigroupson L

1 (R

n+1

+

;E). For

dierentapproachesand resultswithvariablecoeÆcientsin thescalar-valuedcasewereferto

Amann[Ama83],Di Blasio[DiB91],Guidetti[Gui93]andTanabe[Tan97],Section5.4.

2. Elliptic Problems onL 1

(R n

;E)

Throughoutthissection,letE beaBanachspace. Followingthenotionof[Ama01]orSection

5of[DHP01],wecallahomogeneousB(E)-valuedpolynomialA()ofdegreem2N parameter-

elliptic ifthereis anangle2[0;)suchthat thespectrum(A())satises

(A())

forall2R n

; jj=1:

(1)

Wethencall

A

:=inff: (1)holdsg= sup

jj=1

jarg(A())j

theangle of ellipticity ofA. ForD= i(@

1

;:::;@

n

)wecall A(D)= P

jj=m a

D

parameter

elliptic,ifitssymbolA()isparameter-elliptic.

AssumenowthatA(D)isaparameter-ellipticoperatorwithangleof ellipticity

A

. Itwas

proved in Theorem 5.2 and Corollary5.3 of [DHP01] that for >

A

and k 2 N there are

1

Konstanzer Online-Publikations-System (KOPS) URL: http://www.ub.uni-konstanz.de/kops/volltexte/2008/5062/

URN: http://nbn-resolving.de/urn:nbn:de:bsz:352-opus-50628

(2)

constantsc

;k , C

;k

suchthatthesolution

of

u+A(D)u=Æ

0

satisestheestimate

jD

(x)jC

;k jj

n+k

m 1

p n

m;k (c

;k

jxj); x2R n

;jargj ;jj=k;

(2)

wherep n

m;k

isgivenby

p n

m;k (r)=

Z

1

0

s n 2

(1+s) m k 1

e r(1+s)

ds:

Wedene theL

1 (R

n

;E)-realizationA ofA(D)bymeansofA=A

0 ,where

[A

0

u](x)=A(D)u(x); x2R n

; u2D(A

0 )=W

m

1 (R

n

;E):

TherstassertionofthefollowingpropositionisaspecialcaseofTheorem5.10of[Ama01].

Theproofgivenbelowisbasedonestimatesonthefundamental solution. Viaestimate(2)we

alsoobtainsomeinformationonthedomainofA, whichwill beimportantin thefollowing.

Proposition 2.1. Let n;m2N, E be aBanach space, a

2B(E) andsupposethat A(D) is

parameterelliptic with angle ofellipticity

A

<

2

. Then A generatesan analytic semigroup

onL

1 (R

n

;E)of angle

2

A

andwehave

W m

1 (R

n

;E)D(A)W m 1

1 (R

n

;E):

Proof. Obviously,A hasdensedomain. If f 2L

1 (R

n

;E), choose asequence f

n 2C

1

0 (R

n

;E)

such thatf

n

!f inL

1 (R

n

;E). For2

, >

A

,wehaveu

n :=

f

n 2W

m

1 (R

n

;E)

aswell asu

n

+A(D)u

n

=f

n

. Since u

n

!u=

f in L

1 (R

n

;E)asn !1, wesee that

u 2 D(A) and u+Au = f. This shows that +A is invertible for each 2

with

(+A) 1

f =

f. Thusby(2) andYoung'sinequalityweobtain

W m 1

1 (R

n

;E)D(A)D(A

0 )=W

m

1 (R

n

;E):

Furthermore,estimate(2)yields

(A) and

j(+A) 1

j

B (L

1 (R

n

;E))

M

;

foreach>

A .

3. EllipticProblems ina HalfSpace

Inthis sectionweconsider boundaryvalueproblemsoftheform

u+A(D)u = f; in R n+1

+

B

j

(D)u = g

j

; on @R

n+1

+

;j=1;:::;m:

onahalfspace. HereA(D)aswellasB

j

(D)forj2f1;:::;mgaredierentialoperatorswith

operator-valuedcoeÆcients. WealsoassumethatA(D)andB

j

(D)consistonlyoftheprincipal

parts,i.e.

A(D) = X

jj=2m a

D

B

j

(D) = X

jj=m b

j D

;

(3)

wherem

j

2f0;:::;2m 1g, a

2B(E)andb

j

2B(E)forj 2f1;:::;mg. Inthefollowing

weassumethatA(D)isparameter-ellipticwithangleofellipticity

A

2[0;),i.e. thereexists

2[0;)suchthat

(A())

; 2R n+1

; jj=1 (3)

and

A

isdenedastheinmumofallsatisfying(3). HereA()isthesymbolofA(D)dened

by

A()= X

jj=2m a

; 2R n+1

;a

2B(E):

(4)

WesupposethatthefollowingLopatinskii-ShapiroConditionholdstrue:

Lopatinskii-ShapiroCondition:

Foreach 0

2R n

and2

withj 0

j+jj6=0,theproblem

v(y)+A(

0

;D

y

)v(y) = 0; y>0;

B

j (

0

;D

y

)v(0) = g

j

; j=1;:::;m

admitsauniquesolutionu2C

0 (R

+

;E)foreach(g

1

;:::;g

m )

T

2E m

.

Inthefollowingweareinterestedin theL

1

-theoryoftheaboveproblem; morespecically,

for>

A

weconsiderthefollowingproblem:

Given 2

, f 2 L

1 (R

n +1

+

;E) and g

j 2 W

2m m

j

1

(R n+1

+

;E) for j 2 f1;:::;mg, nd

u2W 2m

1 (R

n+1

+

;E)whichsatises

u+A(D)u = f; in R n+1

+ (5)

B

j

(D)u = g

j

; on @R

n+1

+

;j=1;:::;m:

(6)

To this end, forA(D) and B

j

(D)dened asabove, wedene an operator A in L

1 (R

n+1

+

;E)

associated to the boundary value problem (5) and (6) with g

j

= 0 for all j 2 f1;:::;mg

bymeans of A =A

min

, where A

min : D(A

min )! L

1 (R

n+1

+

;E) m

j=1 W

2m mj

1

(R n+1

+

;E) is

denedby

D(A

min

) := W 2m

1 (R

n+1

+

;E)

A

min u :=

0

B

B

B

@ A(D)

B

1 (D)

.

.

.

B

m (D)

1

C

C

C

A u:

Moreover,wesetg:=(g

1

;:::;g

m )

T

. For2

,f 2L

1 (R

n+1

+

;E)andg

j 2W

2m mj

1

(R n+1

+

;E)

forj=1;:::;m,ourboundaryvalueproblem canrewrittenas

Jv+Av=

f

g

; (7)

T

(4)

Theorem3.1. Let A(D) be aparameter-elliptic operator oforder2mandangle of ellipticity

A

. Let >

A

. For j 2 f1;:::;mg let B

j

(D) be boundary operators of order m

j

< 2m.

AssumethattheLopatinskii-Shapiroconditionholds. LetEbeaBanachspace,f 2L

1 (R

n+1

+

;E)

and g

j 2W

2m mj

1

(R n+1

+

;E) for j =1;:::;m. Let 2

and let A be dened as above.

Then thereexistsauniquefunction u2W 2m 1

1 (R

n+1

+

;E)\D(A)satisfying

Ju+Au=

f

g

: (8)

Moreover, uisgiven by

u=P(+A

R n+1)

1

E

0 f+

m

X

j=1 R

j

f+

m

X

j=1 S

j

g

j

; (9)

where R j

and S

j

are kernel operators as dened in Propositions 6.8 and 6.9 of [DHP01].

Furthermore, there existsaconstantC >0suchthat for 0jj2m 1and 2

we

have

j 1

jj

2m

D

uj

L

1 (R

n+1

+

;E)

C[jfj

L

1 (R

n+1

+

;E) +

m

X

j=1

j( +jj 1

m

) 2m m

j

2

g

j j

L

1 (R

n+1

+

;E)

+ m

X

j=1

j( +jj 1

m

) 2m m

j 1

2

D

y g

j j

L1(R n+1

+

;E) ]:

Proof. ThefollowingproofisamodicationoftheargumentsgivenintheproofofTheorem6.10

of[DHP01]totheL

1

-situation. Indeed,notethatunderthegivenassumptionstheproblem(8)

hasauniquesolutionu2W 2m 1

1 (R

n+1

+

;E)\D(A)ifandonlyifudened asin(9)belongsto

W 2m 1

1 (R

n+1

+

;E)\D(A).

Givenf 2L

1 (R

n+1

+

;E),itfollowsfromProposition2.1(anditsproofin case>=2)that

thersttermontherighthandsideof(9)belongstotherequiredregularityclass.

Inordertotreatthesecond term,noticethatbyProposition6.6of[DHP01]thekernelk R;j

ofR j

satisesfor0jj2m 1and2

anestimateoftheform

jD

K R;j

(;y;y 0

)j

L 1

(R n

;B (E)) Cjj

2m+jj+1

2m

p n+1

2m+n;jj (cjj

1

2m

(y+y 0

)); y;y 0

>0:

(10)

The abovekernelestimates allowus to deriveL 1

-estimates for thesecond termon the right

hand sideof (9) viathefollowingsimplelemma on L 1

-continuity ofintegraloperatorsacting

inhalf spaces.

Lemma3.2. LetT beanintegral operator inL

1 (R

n+1

+

;E)of the form

(Tf)(x;y)= 1

Z

0 Z

R n

k(x x 0

;y;y 0

)f(x 0

;y 0

)dx 0

dy 0

; x2R n

;y>0;

wherek:R n

R

+ R

+

!B(E)isameasurablefunction. If

sup

y 0

>0 1

Z

0

jk(;y;y 0

)j

1

dy=:M <1;

thenT 2L(L

1 (R

n+1

+

;E))andjTj

L(L

1 (R

n+1

;E))

M.

(5)

Theproofof Lemma3.2 isconsists onlyofan applicationof Young'sandHolder'sinequality.

Combiningestimate(10)withLemma3.2 itfollowsthat

j 1

jj

2m

D

R j

fj

L1(R n+1

+

;E) Cjfj

L1(R n+1

+

;E) :

Similarly,byProposition6.8of[DHP01] thekernelk S;j

ofS

j

satisesfor0jj2m 1

and2

anestimateoftheform

jD

K S;j

(;y

0

)j

L 1

(R n

;B (E)) Cjj

2m+jj+1

2m

p n+1

2m+n;jj (cjj

1

2m

y 0

); y 0

>0:

Again,togetherwithLemma 3.2,thisestimateimpliesthat

j 1

jj

2m

D

S j

g

j j

L1(R n+1

+

;E)

C (

m

X

j=1

j( +jj 1

m

) 2m m

j

2

g

j j

L1(R n+1

+

;E)

+ m

X

j=1

j( +jj 1

m

) 2m m

j 1

2

D

y g

j j

L

1 (R

n+1

+

;E) ):

Thisprovestheassertion.

Foraparameter-ellipticoperator A(D)of order2mand angle ofellipticity

A

, wedenethe

L

1 (R

n+1

+

;E)-realizationA 0

B

oftheboundaryvalueproblem(8)withg=0as

A 0

B

u := A(D)u (11)

D(A 0

B

) := fu2W 2m

1 (R

n+1

+

;E);B

j

(D)u=0 forall j=1;:::;mg (12)

andset

A

B :=A

0

B :

ItfollowsfromTheorem3.1that (+A

B

)isinvertibleforall2

with>

A

andthat

(+A

B )

1

=P(+A

R n+1)

1

E

0 +

m

X

j=1 R

j

:

Theorem3.1alsoimpliesthat P

(A)andthatj(+A

B )

1

jMfor2

with

>

A

. Wethushavethefollowingcorollary.

Corollary3.3. Suppose that

A

<

2

. Then A

B

generates an analytic C

0

-semigroup on

L

1 (R

n+1

+

;E).

References

[ADN59] Agmon,S.,Douglis,A.,Nirenberg,L.: Estimatesneartheboundaryforsolutionsofellipticpartial

dierentialequationssatisfyinggeneralboundaryconditions,I.Comm.PureAppl.Math.22(1959),

623{727.II.Comm.PureAppl.Math.17(1964),35-92.

[Ama83] Amann, H.: Dualsemigroups and second orderlinear elliptic boundary valueproblems. Israel J.

Math.45(1983),225-254.

[Ama01] Amann,H.:Ellipticoperatorswithinnite-dimensionalstatespaces.J.Evol.Equ.1(2001),143-188.

[DHP01] Denk,R.,Hieber, M.,Pruss,J.: R-boundedness,Fouriermultipliersand problemsofelliptic and

(6)

[DiB91] DiBlasio, G.: Analytic semigroupsgenerated byelliptic operatorsinL 1

and parabolicequations.

OsakaJ.Math.28(1991),367-384.

[Gui93] Guidetti,D.:OnellipticsystemsinL 1

.OsakaJ.Math.30(1993),397-429.

[Tan97] Tanabe, H.: FunctionalAnalytic Methodsfor PartialDierential Equations.Marcel Dekker, New

York,1997.

Universi

atRegensburg,Naturwissenschaftliche

Fakult

at I -Mathematik,D-93040Regensburg,

Germany

E-mailaddress: robert.denk@mathematik.u ni- rege nsb urg. de

Technische

Universit

atDarmstadt,FachbereichMathematik,Schlossgartenstr. 7,D-64289Darm-

stadt,Germany

E-mailaddress: hieber@mathematik.tu-dar mst adt. de

Martin-Luther-Universit

atHalle-Wittenberg,FachbereichMathematikundInformatik,Institut

f

urAnalysis,Theodor-Lieser-Strae5,D-06120Halle,Germany

E-mailaddress: anokd@volterra.mathemati k.u ni-h all e.de

Referenzen

ÄHNLICHE DOKUMENTE

So we can see that the non-standard ellip- ticity conditions (ii) and (iii) in Definition 3.1 lead (under the additional assumption that the boundary value problem in 3.1 (ii) is

The main questions concerning general boundary value problems of the form (1.2) consist in finding the appropriate Sobolev spaces (i.e. parameter- dependent norms) for which

A.: On boundary value problems for linear parabolic systems of differential equations of general form (Russian). A.: Regular degeneration and boundary layer for linear

This analog in some sense was suggested by the deep connection of mixed order problems with large parameter to the Luysternik–Vishik theory of boundary layers, developed for

Boundary value problems, higher order Sobolev spaces, parameter- ellipticity, resolvent

The two main examples of boundary value problems of the form (1-1)–(1-2) are the Stefan problem and the Cahn-Hilliard equation... Recently it was treated in detail by

This question was answered by Arendt and Bu in [AB02] for the one-dimensional case n = 1, where a discrete operator-valued Fourier multiplier result for UMD spaces and applications

In this paper we prove parameter-dependent a priori estimates for mixed-order boundary value problems of rather general structure.. In partic- ular, the diagonal operators are