• Aucun résultat trouvé

Optimal eigenvalue estimate for the Dirac-Witten operator on bounded domains with smooth boundary

N/A
N/A
Protected

Academic year: 2021

Partager "Optimal eigenvalue estimate for the Dirac-Witten operator on bounded domains with smooth boundary"

Copied!
19
0
0

Texte intégral

(1)

HAL Id: hal-00199521

https://hal.archives-ouvertes.fr/hal-00199521

Preprint submitted on 19 Dec 2007

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.

operator on bounded domains with smooth boundary

Daniel Maerten

To cite this version:

Daniel Maerten. Optimal eigenvalue estimate for the Dirac-Witten operator on bounded domains

with smooth boundary. 2007. �hal-00199521�

(2)

hal-00199521, version 1 - 19 Dec 2007

DANIELMAERTEN

Abstrat. Eigenvalue estimate for the DiraWitten operator is given on bounded

domains (with smooth boundary) of spaelike hypersurfaes satisfying the dominant

energy ondition,underfour natural boundaryonditions (MIT,APS, modiedAPS

andhiralonditions). Roughlyspeaking, anyeigenvalueoftheDiraWittenoperator

satises

|λ|2> n (n−1)R0 ,

where R0 is the innimumof (theoppositeof) the lorentziannormof the onstraints vetor. Equalityasesarealsoinvestigatedandleadtointerestinggeometrisituations.

1. Introdution

Theaimofspetralgeometryistoderivesomegeometripropertiesfromthestudyofthe

spetrumof a ertainellipti operator, whih is mostlya Laplae operator (Laplaian on

funtions, Hodge Laplaian on

p

forms or DiraLaplaian on spinors). In that ontext, a lassial issue is to give lower bounds for the eigenvalues of the Dira operator on a

ompat manifold

(M

n

, g)

,

(n > 2)

. In [5 ℄, T. Friedrih proved that any eigenvalue

λ

of

theDira operator satises

(1.1)

λ

2

> n

4(n − 1) inf

M

Scal

g

,

where

Scal

g denotes the salar urvature of

(M

n

, g)

. A few years later, O. Hijazi [9 , 10℄

improvedthis result byshowing that, for any

n > 3

,

(1.2)

λ

2

> n

4(n − 1) µ

1

,

where

µ

1 is the rst eigenvalue of the onformal Laplaian. Clearly, estimates (1.1 ) and

(1.2)areunderinterestonlyifthe salarurvatureispositive. Inareentwork,O. Hijazi

andX.Zhang [12℄establishedananalogous versionof(1.1 )for theDiraWittenoperator

underahiralboundaryondition, ofaompatspaelikehypersurfae

(M, g, k)

satisfying

thedominant energy ondition, namely

(1.3)

λ

2

> n 4(n − 1) inf

M

n

Scal

g

+ (tr

g

k)

2

− |k|

2g

− 2 |δ

g

k + dtr

g

k|

g

o .

On the other hand, S. Raulot [16℄ proved an eigenvalue estimate for the Dira operator

ondomainswithboundary. Morepreisely,if

isa ompatdomain ofan

n

dimensional Riemannianspinmanifold

(M, g)

,whoseboundary

∂Ω

haspositivemeanurvature

H

,he

showedundera naturalboundaryondition (alled"MIT"boundary ondition), thatany

Date:19thDeember2007.

(3)

eigenvalue

λ

of the spetrumof theDiraoperator on

(whih isan unbounded disrete

setofomplex numbers ofpositive imaginarypart)satises

(1.4)

|λ|

2

> n

4(n − 1) inf

Scal

g

+n Im(λ) inf

∂Ω

H .

Themostinteresting fatinInequality(1.4) isthatthebound dependsonsomeboundary

geometriquantity,whihisnottheaseforInequality(1.3). Inaddition,S.Raulotshowed

thatequalityin (1.4) leads to the existene ofimaginary Killing spinor on

,and also to

theonlusion that the boundary

∂Ω

is a totally umbilial and onstant mean urvature

hypersurfae.

The goal of this artile is to generalise Inequalities (1.3 ) and (1.4 ) in several dire-

tions. Indeed,weshallprove analogous versionsof(1.3) fortheDiraWittenoperator on

bounded domains, under four natural boundary onditions (see [11 ℄) and also generalise

(1.4)for the DiraWitten operator.

The artile is organised as follows: In Setion 2, we give our geometri onventions and

preliminary results; Setion 3 is devoted to the statements of the main results and their

proves.

2. Geometri Bakground

2.1. Notations. We onsider

(N

n+1

, γ)

aLorentzian manifoldofsignature

(−, +, · · · , +)

whih ontains

(M

n

, g, k)

,a spin(indimension 3this only meansorientable) Riemannian hypersurfae(notneessarilyompat)whoseinduedmetriis

g

andseondfundamental form (extrinsi urvature) is

k

. Let

be a ompat domain in

(M

n

, g, k)

satisfying the

dominant energy ondition, whih readsasthe following inequality along

Ω Scal

g

+ (tr

g

k)

2

− |k|

2g

> 2 |δ

g

k + dtr

g

k|

g

.

We willworkwiththe omplexspinor bundleof

N

restrited to thehypersurfae domain

,thatisto say

Σ := Σ(N )

|Ω whih isgivenbythehoieofa unitnormal

e

0 of

M

in

N

,

along

. More preisely, ifone denotes by

P

Spin(n,1)

(N )

thebundle of Spin

(n, 1)

frames

on

N

,and by

ρ

n,1 the standardrepresentation ofSpin

(n, 1)

then

Σ(N ) := P

Spin(n,1)

(N ) ×

ρn,1

C

[(n+1)/2]

.

Now the hoie of unit normal

e

0 of

M

in

N

, along

, indues a natural inlusion

Spin

(n) ⊂

Spin

(n, 1)

andsowe an dene

Σ := P

Spin(n,1)

(N )

|Ω

×

n,1)|

Spin(n)

C

[(n+1)/2]

.

Σ

naturally arries two sesquilinear inner produts: the rst one denoted by

(∗, ∗)

is

Spin

(n, 1)

-invariant (it is not neessary positive); the seond one whih is denoted by

h∗, ∗i := (e

0

· ∗, ∗)

is Spin

(n)

-invariant and Hermitian denite positive (

·

is the Cliord

ation with respet to the metri

γ

). The Hermitian or anti-Hermitian harater of the Cliord multipliation by vetors diers if we onsider

(∗, ∗)

or

h∗, ∗i

and is desribed

in [13 ℄, for instane.

Σ

is also endowed with two dierent onnetions

∇, ∇

whih are

respetively the Levi-Civitaonnetions of

γ

and

g

. Letus take a spinor eld

ψ ∈ Γ(Σ)

anda vetor eld

X ∈ Γ(T Ω)

,thenour onventions are

X

ψ = ∇

X

ψ − 1

2 k(X) · e

0

· ψ

hk(X), Y i

γ

= h∇

X

Y, e

0

i

γ

.

(4)

Inthese formulae

·

denotes theCliord ation withrespet to themetri

γ

. Theindued

metri, Levi-Civita onnetion and seond fundamental form of the boundary

∂Ω

, are

respetivelydenotedby

ℓ, ∇, θ e

. Ouronventionsare,forany

X, Y ∈ Γ(T ∂Ω)

and

ψ ∈ Σ

|∂Ω,

X

Y = ∇ e

X

Y + θ(X, Y )ν

X

ψ = ∇ e

X

ψ + 1

2 θ(X) · ν · ψ,

where

ν

isthe unitnormalto

∂Ω

pointinginside

,and

·

still denotes theCliordation

withrespetto the metri

γ

. Finallywe dene the following geometri quantity

R

0

:= 1

4 inf

n

Scal

g

+ (tr

g

k)

2

− |k|

2g

− 2 |δ

g

k + dtr

g

k|

g

o .

2.2. Dira-Witten operators and BohnerLihnerowiz formulæ. From now on,

(e

k

)

nk=0 denotesanorthonormalbasisatthepoint,withrespetto themetri

γ

,andwhere

ν = e

1. We an thendene theDiraWittenoperator of

D ϕ = X

n k=1

e

k

· ∇

ek

ϕ ,

where

·

istheCliordationwithrespettothemetri

γ

. Notiethat

D

anbeonsidered asa deformation of

D

the usual Dira operator of

sine we reover

D = D

as soon as

k ≡ 0

. The Dira-Witten operator

D

islearly formally selfadjoint in

L

2 withrespet to

h∗, ∗i

inthelassofompatlysupportedspinor elds,andwehave thelassialBohner-

Lihnerowiz-Weitzenbök formula (f.[1, 7 ,15℄for instane)

(2.1)

D

D = DD = ∇

∇ + R ,

where

R := 1 4

Sal

γ

+ 4

Riγ

(e

0

, e

0

) + 2e

0

·

Riγ

(e

0

)

= 1 4

n

Scal

g

+ (tr

g

k)

2

− |k|

2g

+ 2(δ

g

k + dtr

g

k) · e

0

o .

As usual, we derive from (2.1 ) an integration formula. The idea is to onsider a ertain

spinoreld

ϕ

andto dene the 1form

ω

ϕ

∈ Γ(T

Ω)

bytherelation

ω

ϕ

(X) = h∇

X

ϕ + X · D ϕ, ϕi .

Then, omputingthe

g

divergene of

ω

ϕ,we obtain

div

g

ω

ϕ

= | D ϕ|

2

− |∇ϕ|

2

− h R ϕ, ϕi ,

whih gives,applying Stokes' theorem

(2.2)

Z

| D ϕ|

2

= Z

|∇ϕ|

2

+ Z

h R ϕ, ϕi + Z

∂Ω

h∇

ν

ϕ + ν · D ϕ, ϕi .

We now have to introdue

P

the twistor operator with respet to the onnetion

,

whih isdened bytherelation

P

X

ϕ := ∇

X

ϕ + 1

n X · D ϕ ,

for every

X ∈ Γ(T Ω)

and every spinor eld

ϕ ∈ Γ(Σ)

. We derive a seond integration

(5)

2.1. Proposition. Forany spinor eld

ϕ ∈ Γ(Σ)

, we have

(2.3)

Z

|P ϕ|

2

=

n − 1 n

Z

| D ϕ|

2

− Z

h R ϕ, ϕi − Z

∂Ω

ω

ϕ

(ν) .

Proof: We rst provethe BohnerLihnerowiz formulafor

P

,namely

|P ϕ|

2

= X

n k=1

ek

ϕ + 1

n e

k

· D ϕ, ∇

ek

ϕ + 1

n e

k

· D ϕ

= |∇ϕ|

2

+ 1

n − 2 n

| D ϕ|

2

= |∇ϕ|

2

− 1

n | D ϕ|

2

.

Weintegrate this formulaon

sothat

Z

|P ϕ|

2

= Z

|∇ϕ|

2

− 1 n

Z

| D ϕ|

2

=

n − 1 n

Z

| D ϕ|

2

− Z

h R ϕ, ϕi − Z

∂Ω

ω

ϕ

(ν) ,

where theseondline isobtained thanksto (2.2).

Itislear thatthe valueof the1form

ω

ϕ dependsupontheboundary onditionthatwill

be used. We will need the intermediateresult:

2.2. Lemma. For any spinor eld

ϕ ∈ Γ(Σ)

, we have along the boundary

∂Ω

(2.4)

ω

ϕ

(ν) =

ν · e

j

· ∇ e

ej

ϕ + 1 2

(tr

g

k)ν · e

0

· −k(ν) · e

0

· +(tr

θ) ϕ, ϕ

.

Proof: Just ompute, using the relations of ompatibility between thedierent onne-

tions

ν

ϕ + ν · D ϕ = ν · X

n j=2

e

j

· ∇

ej

ϕ

= ν · e

j

·

∇ e

ej

− 1

2 k(e

j

) · e

0

· + 1

2 θ(e

j

) · ν·

ϕ

= ν · e

j

· ∇ e

ej

ϕ + 1 2

(tr

g

k)ν · e

0

· −k(ν ) · e

0

· +(tr

θ) ϕ .

3. Main Results

3.1. "MIT" boundary ondition. Remind that our aim isto nd an estimate for the

spetrum of the DiraWitten operator under a natural boundary ondition (that had

been used in order to obtain some blak hole version of the positive mass theorem for

asymptotiallyhyperboli manifolds [4 , 13℄). It onsistsonnding a lower bound for

|λ|

2

where

λ

is anynonzero omplex (a priori)number involvinginthefollowing elliptirst

orderboundaryproblem

(

MIT

)

D ϕ = λϕ

on

F (ϕ) = ϕ

on

∂Ω ,

(6)

where

ϕ

is nonzero (eigen)spinor eld, and where

F ∈ End Σ

|∂Ω is dened by the

relation

F (ψ) = iν · ψ

. The boundary ondition

F(ψ) = ψ

was originally introdued by

physiistsoftheMIT,andthisthereasonwhyitisoftenalled"MIT"boundaryondition.

Their idea is based on the fat that the Dira (and also the DiraWitten) operator on

manifoldwithboundaryis not formallyselfadjoint anymore, sine we have thefollowing

integrationbypartsformula

(3.1)

Z

h D ϕ, ψi = Z

hϕ, D ψi − Z

∂Ω

hν · ϕ, ψi ,

foranyspinorelds

ϕ, ψ

. Thisdefetofselfadjointnesshasaonsequeneonthespetrum ofthe problem(MIT).

3.1. Lemma. The spetrum of (MIT) is a disrete set of omplex numbers with positive

imaginaryparts.

Proof: Let

λ

be anyeigenvalue of

(MIT)

with

ϕ

a orresponding eigenspinor eld. Just take

ψ = iϕ

in(3.1) andonsequently get

2 Im(λ) Z

|ϕ|

2

= Z

∂Ω

|ϕ|

2

> 0 ,

and so

Im(λ) > 0

. Assume now that

Im(λ) = 0

, then

ϕ

should vanish on

∂Ω

and so on

thewhole

, by the ontinuation priniple. Thisontradits the fatthat an eigenspinor

isbydenition nonidentiallyzero.

For later usewe introdue some geometri quantities.

3.2. Denition. We set

H

0MIT

:= inf

∂Ω

tr

θ −

k

∂Ω

(ν)

g

, k

∂Ω

(ν) :=

X

n j=2

k(ν, e

j

)e

j

.

The rst main result of this noteis a generalisation of the lower bound of [16 ℄ for the

DiraWittenoperator (we reover(1.4) when we set

k ≡ 0

on

).

3.3.Theorem. Let

beaompatdomainofaspaelikespinhypersurfae

(M, g, k)

whih

satises the dominant energy ondition along

(so that

R

0

> 0

). The boundary

∂Ω

is

assumed to verify

H

0MIT

> 0

. Then under the (MIT) boundary ondition, the spetrum

of the DiraWittenoperator on

is an unbounded disrete set of omplex numbers with

positive imaginary part, suh that any eigenvalue satises

(3.2)

|λ|

2

> n

(n − 1)

R

0

+ H

0MIT

Im(λ) .

Proof: The fatthat, underthe (MIT) boundary ondition, thespetrum oftheDira

Witten operator on

is an unbounded disrete set of omplex numbers with positive

imaginarypart, hasbeen proved intheprevious lemma.

As far as Inequality (3.2 ) is onerned, we onsider a 1parameter family of modied

spinorial LeviCivita onnetion. Indeed,for any

α ∈ R

,we dene theation of

α on

Σ

bytherelation

αX

ϕ := ∇

X

ϕ + iαX · ϕ ,

(7)

for every

X ∈ Γ(T Ω)

. Wean then dene theDiraWitten operator with respet to the

Killingonnetion

α

D

α

ϕ = X

n k=1

e

k

· ∇

αe

k

ϕ ,

where

·

isthe Cliordationwithrespettothe metri

γ

. An easyomputation givesthe

relation

D

α

= D − inα

,sothat

D

α isnot formallyselfadjointin

L

2 withrespetto

h∗, ∗i

inthe lassofompatlysupported spinor eldsif

α 6= 0

.

Weonsider aertain spinor eld

ϕ

and dene the 1form

ω

αϕ

∈ Γ(T

Ω)

bytherelation

ω

ϕα

(X) = h∇

αX

ϕ + X · D

α

ϕ, ϕi .

We notie that

ω

αϕ

(X) = ω

ϕ

(X) − α(n − 1) hiX · ϕ, ϕi

. Then, we only have to ompute

the

g

divergene ofthe1form

ξ(X) := hiX · ϕ, ϕi

. Tothis end,we assumethatourloal

basesatises

ej

e

m

= 0

atthepoint wherethe omputation ismade(thisisequivalent to

ej

e

m

= −k(e

j

, e

m

)e

0) sothat

div

g

ξ = − X

n j=1

e

j

· ξ(e

j

)

= −

* i ∇

ej

e

j

| {z }

=0

·ϕ, ϕ +

ie

j

· ∇

ej

ϕ, ϕ

ie

j

· ϕ, ∇

ej

ϕ

= − hi D ϕ, ϕi − 1 2

*

i e

j

· k(e

j

)

| {z }

=−trgk

·e

0

· ϕ, ϕ +

− hϕ, i D ϕi + 1 2

*

i k(e

j

) · e

j

| {z }

=−trgk

·e

0

· ϕ, ϕ +

= − hi D ϕ, ϕi − hϕ, i D ϕi .

UsingStokes' theoremandintegration formula(2.2 ),we get

Z

∂Ω

ω

αϕ

(ν ) = Z

| D ϕ|

2

− Z

|∇ϕ|

2

− Z

h R ϕ, ϕ, i + α(n − 1) Z

hi D ϕ, ϕi + hϕ, i D ϕi .

Easyomputations leadto the relations

|∇

α

ϕ|

2

= |∇ϕ|

2

+ nα

2

|ϕ|

2

+ α hi D ϕ, ϕi + hϕ, i D ϕi

| D

α

ϕ|

2

= | D ϕ|

2

+ n

2

α

2

|ϕ|

2

+ nα hi D ϕ, ϕi + hϕ, i D ϕi .

Plugging thisinto our integration byparts formula,we obtain

(3.3)

Z

∂Ω

ω

αϕ

(ν ) = Z

| D

α

ϕ|

2

− Z

|∇

α

ϕ|

2

− Z

h R

α

ϕ, ϕ, i ,

where

R

α

:= R + n(n − 1)α

2. Thetwistor operator withrespetto the onnetion

α is

dened by

P

Xα

ϕ := ∇

αX

ϕ + 1

n X · D

α

ϕ ,

forevery

X ∈ Γ(T Ω)

and every spinor eld

ϕ ∈ Γ(Σ)

. We straightly have

|P

α

ϕ|

2

= |∇

α

ϕ|

2

− 1

n | D

α

ϕ|

2

> 0 ,

and,afterintegration on

we get,using (3.3 )

(3.4)

0 6 Z

|P

α

ϕ|

2

=

n − 1 n

Z

| D

α

ϕ|

2

− Z

h R

α

ϕ, ϕ, i − Z

∂Ω

ω

αϕ

(ν) .

(8)

We now need to give some elementary properties of the boundary spinorial endomor-

phism

F

(theproof isleftto thereader).

3.4. Proposition. The endomorphism

F

is symmetri, isometri with respet to

h∗, ∗i

,

ommutes to theation of

ν·

and antiommutes to eah

e

k

·, (k 6= 1)

.

The important fat is that the boundary ondition of (MIT) allows us to express the

boundary integrand of(3.4) intermsof theextrinsiurvaturetensors

k

and

θ

.

3.5. Lemma. If

F (ϕ) = ϕ

then along the boundary

∂Ω

we have

(3.5)

ω

αϕ

(ν ) = 1 2

D e

0

· n

tr

θ − 2(n − 1)α

e

0

+ k

∂Ω

(ν ) o

· ϕ, ϕ E .

Proof of Lemma 3.5: We useEquation(2.4 )

ω

αϕ

(ν ) =

ν · e

j

· ∇ e

ej

ϕ + αiν · e

j

· e

j

· ϕ + 1 2

(tr

g

k)ν · e

0

· −k(ν ) · e

0

· +(tr

θ) ϕ, ϕ

=

ν · e

j

· ∇ e

ej

ϕ − (n − 1)αiν · ϕ + 1 2

(tr

k)ν · e

0

· −k

∂Ω

(ν) · e

0

· +(tr

θ) ϕ, ϕ

.

Taking Proposition3.4into aount andalso the assumption

F (ϕ) = ϕ

,itomes out

ω

ϕα

(ν) = 1

2 D

(tr

θ − 2α(n − 1)) − k

∂Ω

(ν) · e

0

· ϕ, ϕ E

= 1 2

D

e

0

· (tr

θ − 2α(n − 1))e

0

+ k

∂Ω

(ν)

· ϕ, ϕ E

= 1

2 he

0

· K

αMIT

· ϕ, ϕi ,

where wehave dened the vetor eld

K

αMIT

∈ Γ(T N

|∂Ω

)

as

K

αMIT

= tr

θ − 2α(n − 1)

e

0

+ k

∂Ω

(ν) .

Then, it is well known that the boundary integrand

ω

αϕ

(ν )

is nonnegative ifand only if

K

αMIT isausal and futureoriented,whih readsas

(3.6)

tr

θ − 2α(n − 1) >

k

∂Ω

(ν)

g

.

Asaonsequene,ifone assumes that(3.6) holds andthat

D ϕ = λϕ

,then(3.4) implies

0 6

n − 1 n

|λ − inα|

2

Z

|ϕ|

2

− Z

h R

α

ϕ, ϕ, i 0 6

n − 1

n |λ|

2

+ n

2

α

2

− 2nα Im(λ) Z

|ϕ|

2

− Z

h R

α

ϕ, ϕ, i 0 6

n − 1

n |λ|

2

− 2nα Im(λ) Z

|ϕ|

2

− Z

h R ϕ, ϕ, i ,

whihispossibleonlyif

|λ|

2

>

n

(n−1)

R

0

+ 2nα Im(λ)

. Nowjusttake

α = α

0

=

2(n−1)1

H

0MIT

sothat

K

αMIT0 is learly ausal and future oriented and we obtain the desired inequality,

namely

|λ|

2

> n (n − 1)

R

0

+ H

0MIT

Im(λ) .

(9)

The equality ase in (3.2 ) is not easy to treat in general, but we an however dedue

someinformation under anatural additional assumption.

3.6. Theorem. Under the assumptions of Theorem 3.3, equality in (3.2) always leads to

the existene of an imaginary

Killing spinor on

of Killing number

−iα

0

= −λ/n

.

If we assume furthermore that

k

∂Ω

(ν ) = 0

along the boundary

∂Ω

, then

∂Ω

is a totally

umbilial andonstant mean urvature hypersurfae of

.

Proof: Suppose now that equality holds in (3.2). Thereby, there exists a nonzero

λ

eigenspinor eld

ϕ

suh that

(3.7)

P

α0

ϕ ≡ 0 ,

e

0

· K

αMIT0

· ϕ, ϕ

≡ 0 .

The seond equation of (3.7 ) simply says that the vetor eld

K

αMIT0 is lightlike, whih

meansinother words

tr

θ −

k

∂Ω

(ν)

≡ 2(n − 1)α

0

.

Therst equationof(3.7 ) an bereformulated as

X

ϕ + λ

n

X · ϕ = 0 , ∀X ∈ Γ(T Ω),

thatis,

ϕ

isaKillingspinor. ItiswellknownthattheKillingnumber −λn

hastobeeither

real or purely imaginary. But remind that

Im(λ) > 0

, and so

λ ∈ i R

+. For later use, we set

λ = iµ

with

µ

a positive real number. We onsider now any vetor eld

X ∈ Γ(T ∂Ω)

tangent to the boundary,and we ompute

−i µ n

X · ϕ = ∇

X

ϕ

= ∇

X

(iν · ϕ)

= i ∇

X

ν · ϕ + ν · ∇

X

ϕ

= i

X

ν · −k(X, ν )e

0

· −i µ n

ν · X·

ϕ

= i

−θ(X) · −k(X, ν)e

0

· + µ n

ϕ .

Thisnallyleads to

(3.8)

∀X ∈ Γ(T ∂Ω) ,

−θ(X) · −k(X, ν)e

0

· + 2µ

n

ϕ = 0 .

Theondition(3.8)notably impliesthatforanyvetoreld

X ∈ Γ(T ∂Ω)

,thevetoreld

n

−θ(X) − k(X, ν)e

0

+

2µ n

X o

islightlike, andonsequently we have

(3.9)

∀X ∈ Γ(T ∂Ω) , k(X, ν)

2

=

θ(X) − 2 µ n

X

2

g

.

(10)

Thestrikingrelation(3.9 )somehowmeasuresthefailureof

∂Ω

tobetotallyumbilial. We

nowprove that

µ = nα

0. Indeed, using(3.8 ), we ompute alongtheboundary

n X

n

j=2

he

j

· ϕ, e

j

· ϕi = 2(n − 1)µ n |ϕ|

2

= X

n j=2

he

j

· ϕ, {θ(e

j

) · +k(e

j

, ν)e

0

} · ϕi

= − X

n j=2

hϕ, e

j

· {θ(e

j

) · +k(e

j

, ν)e

0

} · ϕi

= D ϕ, n

tr

θ − k

∂Ω

(ν) · e

0

· o ϕ E

=

ϕ, e

0

· K

0MIT

· ϕ ,

whih anbewritteninshortas

ϕ, e

0

· K

0MIT

ϕ

=

2µ(n−1)n

|ϕ|

2. Besides,using theseond

equationof (3.7 ),we knowthat

0 = Z

∂Ω

e

0

· K

αMIT0

· ϕ, ϕ

= Z

∂Ω

ϕ, e

0

· K

0MIT

· ϕ

− 2(n − 1)α

0

Z

∂Ω

|ϕ|

2

= 2(n − 1) µ

n − α

0

Z

∂Ω

|ϕ|

2

,

whih entails

µ = nα

0 sine the eigenspinor

ϕ

annot vanish identially along

∂Ω

. Sup-

posefurthermore, that

k

∂Ω

(ν) = 0

on theboundary, then itis lear from(3.9) that

∂Ω

is

atotally umbilial and onstant mean urvature hypersurfae.

3.2. APS BoundaryCondition. TheAtiyahPatodiSinger(APS)boundaryondition

wasoftenusedinorderto prove positive masstheorem forasymptotially at blakholes

(that is to say for asymptotially at manifolds with boundary [8, 14℄). More preisely,

we denote by

Π

± the

L

2-orthogonal projetions on the spaes of eigenspinors of positive (respetivelynegative)eigenvaluesoftheDiraoperatoroftheboundary

D e := P

j>2

e

j

·

∂Ω

∇ e

ej,

where

·

∂ΩdenotestheCliordationwiththerespettoboundarymetri

. Inthissetion,

ouraimistondalowerboundfor

|λ|

2where

λ

isanynonzeroomplex(apriori)number

involvinginthe following elliptirst orderboundaryproblem

(

APS

)

D ϕ = λϕ

on

Ω Π

+

(ϕ) = 0

on

∂Ω ,

where

D

still denotes the DiraWitten operator. We rst prove that the spetrum of (APS)isreal.

3.7. Lemma. The spetrum of (APS) isa disrete set of real numbers.

Proof: Let

λ ∈ C

be anyeigenvalue of

(APS)

with

ϕ

a orresponding eigenspinor eld.

Now,

·

N

, ·

, ·

∂Ω the Cliordationsofrespetively

γ, g, ℓ

satisfythefollowing relations

(3.10)

X ·

∂Ω

ψ = X ·

ν ·

ψ, X ·

ψ = iX ·

N

e

0

·

N

ψ .

Références

Documents relatifs

Separable solutions play a key role for describing the boundary behaviour and the singularities of solutions of a large variety of quasilinear equations.. The aim of this article is

Kröger, Upper bounds for the Neumann eigenvalues on a bounded domain in Eulidean spae,

eigenvalue enclosures, Maxwell equation, spectral pollution, finite element

In this paper, we are interested in the minimization of the second eigen- value of the Laplacian with Dirichlet boundary conditions amongst convex plane domains with given area..

Note that our result requires that the manifold is spin (which is a stronger assumption compared with Esco- bar’s theorem) however we don’t assume that the whole manifold is

Druet [1] has proved that for small volumes (i.e. τ > 0 small), the solutions of the isoperimetric problem are close (in a sense to be made precise) to geodesic spheres of

We extend the Friedrich inequality for the eigenvalues of the Dirac operator on Spin c manifolds with boundary under different boundary conditions.. The limiting case is then

Namely, he proved that if M is as in Lichnerowicz theorem, except that it has now a boundary such that its mean curvature with respect to the outward normal vector field is