22.10.2014 Views

TT-EIGENTENSORS FOR THE LICHNEROWICZ LAPLACIAN ON ...

TT-EIGENTENSORS FOR THE LICHNEROWICZ LAPLACIAN ON ...

TT-EIGENTENSORS FOR THE LICHNEROWICZ LAPLACIAN ON ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>TT</strong>-<strong>EIGENTENSORS</strong> <strong>FOR</strong> <strong>THE</strong> <strong>LICHNEROWICZ</strong><br />

<strong>LAPLACIAN</strong> <strong>ON</strong> SOME ASYMPTOTICALLY HYPERBOLIC<br />

MANIFOLDS WITH WARPED PRODUCTS METRICS<br />

ERWANN DELAY<br />

Abstract. Let (M =]0, ∞[×N, g) be an asymptotically hyperbolic<br />

manifold equipped with a warped product metrics. We show that there<br />

exist no <strong>TT</strong> L 2 -eigentensors with eigenvalue in the essential spectrum of<br />

the Lichnerowicz Laplacian ∆ L. If (M, g) is the real hyperbolic space,<br />

there is no symmetric L 2 -eigentensors of ∆ L.<br />

Keywords : Asymptotically hyperbolic manifold, warped product, Lichnerowicz<br />

Laplacian, symmetric 2-tensor, <strong>TT</strong>-tensor, essential spectrum,<br />

asymptotic behavior.<br />

2000 MSC : 35P15, 58J50, 47A53.<br />

1. Introduction<br />

The study of Laplacians acting on symmetric 2-tensors like the Lichnerowicz<br />

Laplacian ∆ L is very important to the understanding of geometric<br />

problems involving deformations of metrics and their Ricci tensor. These<br />

problems appear in a Riemannian context [2] (see also [11], [4], [5] ), in<br />

general Relativity and string theory (see [15], [20], [14], [13] for instance).<br />

Usually, the study of (positivity of) ∆ L on <strong>TT</strong>-tensors plays an important<br />

role [4] [15].<br />

A natural geometric problem is to find a metric with prescribed Ricci<br />

curvature [11], and the infinitesimal version of that problem is to invert the<br />

Lichnerowicz Laplacian on symmetric 2-tensors. In [6], [7] it was shown<br />

that the Ricci curvature can be arbitrarily prescribed in the neighborhood<br />

of the hyperbolic metric on the real hyperbolic space when the dimension is<br />

strictly larger than 9 (see also [9], [10] for some generalizations). In [21], [8]<br />

it is proved that the essential spectrum of the Lichnerowicz Laplacian acting<br />

on trace free symmetric 2-tensors on asymptotically hyperbolic manifolds of<br />

dimension n + 1 is the ray<br />

[ [<br />

n(n − 8)<br />

, +∞ ,<br />

4<br />

with no eigenvalues below if the manifold is the hyperbolic space. The goal<br />

of this article is to study more precisely this essential spectrum for some<br />

particular warped product manifolds. We prove<br />

Date: March 13, 2006.<br />

1


2 E. DELAY<br />

Theorem. For n ≥ 2, let us consider (N, ĝ) be an n-dimensionnal<br />

compact Einstein manifold. Let M =]0, +∞[×N equipped with an asymptotically<br />

hyperbolic metric g = dr 2 +f 2 (r)ĝ (as in section 2). Then there are<br />

no L 2 <strong>TT</strong>-eigentensors of the Lichnerowicz Laplacian ∆ L with eigenvalue<br />

embedded in the essential spectrum. For the real hyperbolic space, there are<br />

no L 2 eigentensors of ∆ L .<br />

As this result concern only embedded eigenvalues in an essential spectrum<br />

and it is well know that the essential spectrum is characterized at infinity,<br />

the manifold can be replaced by any manifold which is of the form given<br />

here only outside a compact set.<br />

This theorem is a consequence of the corollary 3.6 and the theorem 6.2.<br />

Note that this result can easily be adapted to Laplacians on symmetric<br />

covariant 2-tensors of the form ∇ ∗ ∇+ curvatures terms. The complete proof<br />

is in fact given for the rough Laplacian.<br />

Our proof, as the one used by Donnelly [12] for forms on H n+1 , consist<br />

to use a Hodge type decomposition theorem for symmetric 2-tensors on a<br />

compact Einstein manifold N (see also [19] for a similar decomposition on<br />

AdS where N = S n ). This decomposition gives a separation of variable<br />

technique to the equations studied here. The problem is then reduced to<br />

asymptotic integration for ordinary differential equations of the form<br />

y ′′ (r) + (α 2 + O(e −εr ))y(r) = O(e −εr ), α ≥ 0, ε > 0.<br />

The asymptotic behaviour of the solutions to these equations and the L 2<br />

condition will give the result.<br />

As discussed with R. Mazzeo, it is certainly possible to generalize the<br />

result to certain conformally compact manifolds by carefully constructing a<br />

parametrix for the Lichnerowicz Laplacian as he did in [23] for the Hodge<br />

Laplacian. The proof given here has the merit to be purely geometric and<br />

easily tractable.<br />

When the dimension n + 1 is less than or equal to 9, 0 is in the essential<br />

spectrum of ∆ L . The result given here proves that ∆ L stays injective (in L 2 )<br />

when n + 1 ≥ 3, so its image is not closed (but dense) in L 2 . In particular<br />

∆ L is not surjective. This suggests it might be possible to find some (at<br />

least infinitesimally) prohibited directions to prescribe the Ricci curvature<br />

near this space.<br />

The action of the Lichnerowicz Laplacian in the conformal direction corresponds<br />

to the action of the Laplacian on functions. Since we know that<br />

spectrum (see [23] for instance), we are only interested here about the trace<br />

free direction.<br />

This article is organized as follows : first in section 2, we will introduce<br />

all the objects we need along the paper. In section 3 we will see that on the<br />

real hyperbolic space any L 2 eigentensor must be a <strong>TT</strong>-tensor. In section 4<br />

we compute the components of the Laplacian of a <strong>TT</strong>-tensor for a warped<br />

product metric. In section 5, we detail the Hodge type decomposition that<br />

we will need in the proof of the main theorem. In section 6 we give the<br />

proof of the main theorem. The appendix, section 7, recall some results of<br />

asymptotic integration for ordinary differential equations.


<strong>TT</strong>-<strong>EIGENTENSORS</strong> <strong>ON</strong> SOME AH MANIFOLDS 3<br />

Acknowledgments. I am grateful to G. Carron, P. T. Chruściel, B. Colbois<br />

and R. Mazzeo for useful conversations, to Ph. Delanoë and F. Gautero for their<br />

comments on the original manuscript.<br />

2. Definitions, notations and conventions<br />

Let n ≥ 2 and let (N, ĝ) be an n dimensional riemannian manifold. The<br />

riemannian manifolds we are studying throughout the paper after the section<br />

3 are of the form M =]0, +∞[×N endowed with a warped product metric :<br />

g = dr 2 + f 2 (r)ĝ = dr 2 + f 2 (r)dθ 2 ,<br />

where f is a smooth positive function on ]0, ∞[. We will say that (M, g) is<br />

an asymptotically hyperbolic manifold if there exist ε > 0 such that, when<br />

r goes to infinity,<br />

( )<br />

1<br />

f(r), f ′ (r), f ′′ (r) = e r 2 + O(e−εr ) .<br />

The basic example studied here is when f = sinh and (N, ĝ) is the standard<br />

n-dimensional sphere endowed with its canonical metric, so (M, g) is the<br />

real hyperbolic space (minus a point). There are also interesting case when<br />

(N, ĝ) is Einstein with Ric(ĝ) = κ(n − 1)ĝ and f(r) = 1 2 (er − κe −r ) so<br />

Ric(g) = −ng.<br />

We denote by ∇ the Levi-Civita connexion of g and by Riem(g), Ric(g)<br />

respectively the Riemannian sectional and the Ricci curvature of g.<br />

We denote by T p the set of rank p covariant tensors. When p = 2, we<br />

denote by S 2 the subset of symmetric tensor which splits in G ⊕ ˚S 2 where G is<br />

the set of g-conformal tensors and ˚S 2 the set of trace-free tensor (relatively<br />

to g). We observe the summation convention, and we use g ij and its inverse<br />

g ij to lower or raise indices.<br />

The Laplacian is defined as<br />

△ = −tr∇ 2 = ∇ ∗ ∇,<br />

where ∇ ∗ is the L 2 formal adjoint of ∇. The Lichnerowicz Laplacian acting<br />

on symmetric covariant 2-tensors is<br />

where<br />

△ L = △ + 2(Ric − Riem),<br />

(Ric u) ij = 1 2 [Ric(g) iku k j + Ric(g) jk u k i ],<br />

and<br />

(Riem u) ij = Riem(g) ikjl u kl .<br />

For u a covariant 2-tensorfield on M we define the divergence of u by<br />

(divu) i = −∇ j u ji .<br />

For ω, a one form on M , we define its divergence<br />

d ∗ ω = −∇ i ω i ,<br />

the symmetric part of its covariant derivative :<br />

(Lω) ij = 1 2 (∇ iω j + ∇ j ω i ),


4 E. DELAY<br />

(note that L ∗ = div) and the trace free part of that last tensor :<br />

(˚Lω) ij = 1 2 (∇ iω j + ∇ j ω i ) + 1<br />

n + 1 d∗ ωg ij .<br />

The well known [2] weitzenböck formula for the Hodge-De Rham Laplacian<br />

on 1-forms reads<br />

∆ H ω i = ∇ ∗ ∇ω i + Ric(g) ik ω k .<br />

All quantities relative to ĝ will have a hat or will be indexed by ĝ (∇ bg , div bg ,<br />

d bg , ̂∆ L ,...)<br />

A <strong>TT</strong>-tensor (Transverse Traceless tensor) is by definition a symmetric<br />

divergence free and trace free covariant 2-tensor .<br />

L 2 denotes the usual Hilbert space of functions or tensors with the product<br />

(resp. norm)<br />

∫<br />

∫<br />

〈u, v〉 L 2 = 〈u, v〉dµ g (resp. |u| L 2 = ( |u| 2 dµ g ) 1 2 ),<br />

M<br />

M<br />

where 〈u, v〉 (resp. |u|) is the usual product (resp. norm) of functions or<br />

tensors relative to g, and the measure dµ g is the usual measure relative to<br />

g (we will omit the term dµ g ).<br />

3. Commutators of some natural operators<br />

We will study the commutator of the Laplacian with the divergence operator<br />

in order to apply the result for an eigentensor. We also study the<br />

commutator of the Laplacian with the Killing operator L needed in section<br />

5. For further references, we will be as general as possible, in particular the<br />

manifold M is not necessarily a product. We first begin with the lemma:<br />

Lemma 3.1. On a Riemannian manifold (M, g) with Levi-Civita connexion<br />

∇, for all symmetric 2-tensor field u, we have<br />

∇ i ∇ k ∇ k u ij = ∇ k ∇ k ∇ i u ij + R il ∇ l u ij + 2Rj<br />

l ki<br />

∇k u il + ∇ k R p k u pj + ∇ k R p i jk<br />

u ip<br />

= ∇ k ∇ k ∇ i u ij + R il ∇ l u ij − 2R j lki ∇ k u il + 1 2 ∇p Ru pj + (∇ p R ij − ∇ j R ip )u ip .<br />

Proof. For a covariant 3-tensor field T on M, we have<br />

∇ p ∇ q T kij − ∇ q ∇ p T kij = R l kqp T lij + R l iqpT klj + R l jqpT kil .<br />

Contracting this equality with g ip g qk , we obtain<br />

(3.1) ∇ i ∇ k T kij − ∇ k ∇ i T kij = R l jki<br />

Tkil .<br />

For a covariant 2-tensor field u on M, we have<br />

∇ m ∇ k u ij − ∇ k ∇ m u ij = R p ikm u pj + R p jkm u ip.<br />

Contracting this equality with g im , we obtain<br />

∇ i ∇ k u ij − ∇ k ∇ i u ij = R p k u pj + R p jki u ip .<br />

From this last equality, we obtain<br />

(3.2)<br />

∇ k ∇ i ∇ k u ij − ∇ k ∇ k ∇ i u ij = ∇ k R p k u pj + ∇ k R p i jk<br />

u ip + R p k ∇k u pj + R p i jk<br />

∇ k u ip .


<strong>TT</strong>-<strong>EIGENTENSORS</strong> <strong>ON</strong> SOME AH MANIFOLDS 5<br />

Taking T kij = ∇ k u ij in equation 3.1 and combining with equation 3.2, we<br />

get the annonced equality. The equality in the lemma is due to Bianchi<br />

identities.<br />

□<br />

Remark. The formula in lemma 3.1 is much simpler when the curvature<br />

is harmonic (∇ k R kijl = 0) or equivalently when Ric(g) is a Codazzi tensor<br />

(∇ k R ij − ∇ i R kj = 0) and (or) the tensor u satisfy an equality of the form<br />

∇ k u ij − ∇ i u kj + f∇ j u ik = 0 (f is any real function).<br />

Corollary 3.2. On a Riemannian manifold (M, g) with Levi-Civita connexion<br />

∇, we have<br />

div ◦ ∆ L = ∆ H ◦ div − D Ric,<br />

where (D Ric u) j = (∇ p R ij − ∇ j R ip − ∇ i R jp )u ip = (∇ j R ik )u ik .<br />

Proof. From lemma 3.1, we have<br />

∇ i (∇ k ∇ k u ij − R ik u k j − R jk u k i + 2R jlik u kl ) = ∇ k ∇ k ∇ i u ij − R l j∇ i u il<br />

Also, we have the<br />

+(∇ p R ij − ∇ j R ip − ∇ i R jp )u ip<br />

= ∇ k ∇ k ∇ i u ij − R l j∇ i u il<br />

−(∇ j R ip )u ip .<br />

Lemma 3.3. On a Riemannian manifold with Levi-Civita connexion ∇, we<br />

have<br />

∆ L ◦ L = L ◦ ∆ H + D Ric,<br />

where (D Ric ξ) ij = (∇ p R ij − ∇ j R ip − ∇ i R jp )ξ p .<br />

Proof. A straightforward computation gives<br />

∇ k ∇ k ∇ i ξ j = ∇ i ∇ k ∇ k ξ j + R ik ∇ j ξ k − 2R ikjl ∇ k ξ l − (∇ k R ij − ∇ j R ik )ξ k .<br />

The result follows from the definitions of the Hodge and Lichnerowicz Laplacians.<br />

□<br />

In particular the Corollary 3.2 and Lemma 3.3 together give a theorem<br />

of Lichnerowicz ( [22] page 29)<br />

Theorem 3.4. On a Riemannian manifold (M, g) with parallel Ricci curvature,<br />

we have<br />

div ◦ ∆ L = ∆ H ◦ div ,<br />

and<br />

∆ L ◦ L = L ◦ ∆ H .<br />

Remark. The condition ∇ p R ij − ∇ j R ip − ∇ i R jp = 0 is equivalent to the<br />

parallel Ricci curvature condition.<br />

Corollary 3.5. On a (n+1)-dimensional Riemannian manifold (M, g) with<br />

Levi-Civita connexion ∇ and constant sectional curvature K, for all trace<br />

free symmetric 2-tensor fields u, we have<br />

∇ i ∇ k ∇ k u ij = ∇ k ∇ k ∇ i u ij + (n + 2)K∇ i u ij ,<br />


6 E. DELAY<br />

or equivalently<br />

div ◦ ∆ = [∆ − (n + 2)K] ◦ div.<br />

Corollary 3.6. On the (n+1)-dimensional real hyperbolic space with n ≥ 2,<br />

any L 2 symmetric trace free eigentensor field is divergence free, and so is a<br />

<strong>TT</strong>-tensor.<br />

Proof. If u is an L 2 symmetric trace free eigentensor field on the real hyperbolic<br />

space, then div u is an L 2 eigenform, but from [12] (see also [23] for a<br />

generalization) this form must be trivial.<br />

□<br />

4. Components of the divergence and the Laplacian of a<br />

2-tensor<br />

We will compute the components of the divergence and Laplacian for a<br />

symmetric 2-tensor and adapt the result for the case of trace free tensor. In<br />

this section the riemannian manifold (N, ĝ) is not assumed to be Einstein.<br />

On M =]0, +∞[×N, we consider the metric<br />

g = dr 2 + f 2 (r)ĝ.<br />

The computation is given in local coordinates adapted to the character of<br />

M. We will use a local coordinate (x 1 , ..., x n ) = (x A ) on the manifold N<br />

and the coordinate (x 0 , x 1 , ..., x n ) = (r, x 1 , ..., x n ) = (x i ) on M. The indices<br />

with big letters are relative to the coordinate on N, for example<br />

g 00 = 1, g 0A = 0, g AB = f 2 ĝ AB .<br />

The prime in the calculation will denote derivative relative to r.<br />

The Christoffel’s symbol of g are<br />

Γ 0 00 = Γ C 00 = Γ 0 A0 = 0,<br />

Γ 0 AB = −ff ′ ĝ AB , Γ C A0 = f −1 f ′ δ A C , Γ C AB = ̂Γ C AB.<br />

If h is a covariant symmetric 2-tensor, we decompose<br />

(4.1) h = udr 2 + ξ ⊗ dr + dr ⊗ ξ + ĥ,<br />

where ĥ is the orthogonal projection of h on {r} × N, and ξ is a one form<br />

on {r} × N. In particular we have<br />

The covariant derivatives of h are<br />

∇ 0 h 00 = ∂ 0 h 00 = u ′ ,<br />

h 00 = u, h 0A = ξ A , h AB = ĥAB.<br />

∇ 0 h A0 = ∂ 0 h A0 − f −1 f ′ h A0 = ξ ′ A − f −1 f ′ ξ ′ A,<br />

∇ 0 h AB = ∂ 0 h AB − 2f −1 f ′ h AB = ĥ′ AB − 2f −1 f ′ ĥ AB ,<br />

∇ C h 00 = ∂ C h 00 − 2f −1 f ′ h C0 = d C u − 2f −1 f ′ ξ C ,<br />

∇ C h A0 = ̂∇ C h A0 + ff ′ ĝ AC h 00 − f −1 f ′ h AC = ̂∇ C ξ A + ff ′ ĝ AC u − f −1 f ′ ĥ AC ,<br />

∇ C h AB = ̂∇ C h AB + ff ′ (ĝ AC h 0B + ĝ CB h 0A ) = ̂∇ C ĥ AB + ff ′ (ĝ AC ξ B + ĝ CB ξ A ).<br />

In particular, we can compute the components of minus the divergence of h:<br />

∇ i h i0 = u ′ + nf −1 f ′ u − f −3 f ′ h A A + f −2 ̂∇A ξ A ,<br />

∇ i h iA = ξ ′ A + nf −1 f ′ ξ A + f −2 ̂∇B ĥ BA .


<strong>TT</strong>-<strong>EIGENTENSORS</strong> <strong>ON</strong> SOME AH MANIFOLDS 7<br />

If Tr g h = h 00 + f −2 h A A = u + f −2 Tr bg ĥ = 0, this yield<br />

We thus obtain the<br />

∇ i h i0 = u ′ + (n + 1)f −1 f ′ u + f −2 ̂∇A ξ A ,<br />

∇ i h iA = ξ ′ A + nf −1 f ′ ξ A + f −2 ̂∇B ĥ BA .<br />

Lemma 4.1. If h is a symmetric covariant 2-tensor decomposed as in 4.1<br />

then the components of its divergence are<br />

(divh) 0 = −(u ′ + nf −1 f ′ u − f −3 f ′ Tr bg ĥ − f −2 d ∗ bg ξ),<br />

(divh) A = −(ξ ′ + nf −1 f ′ ξ − f −2 div bg ĥ) A .<br />

If moreover h is trace free, the first component becomes<br />

(divh) 0 = −(u ′ + (n + 1)f −1 f ′ u − f −2 d ∗ bg ξ).<br />

In a similar way, we compute the components of the Laplacian of h. After<br />

a straightforward calculation we obtain the following Lemma<br />

Lemma 4.2. Let h be a trace free covariant symmetric 2-tensor decomposed<br />

as is 4.1. The components of minus the Laplacian of h are<br />

∇ k ∇ k h 00 = u ′′ + nf −1 f ′ u ′ − 2(n + 1)f −2 (f ′ ) 2 u + f −2 ̂∇A ̂∇A u − 4f −3 f ′ ̂∇A ξ A ,<br />

∇ k ∇ k h A0 = ξ ′′ A + (n − 2)f −1 f ′ ξ ′ A − [(2n + 1)f −2 (f ′ ) 2 + f −1 f ′′ ]ξ A<br />

+f −2 ̂∇D ̂∇D ξ A − 2f −3 f ′ ̂∇D h AD + 2f −1 f ′ d A u,<br />

∇ k ∇ k h AB = ĥ′′ AB + (n − 4)f −1 f ′ ĥ ′ AB − [(2n − 4)f −2 (f ′ ) 2 + 2f −1 f ′′ ]ĥAB<br />

+f −2 ̂∇D ̂∇D ĥ AB + 2f −1 f ′ ( ̂∇ A ξ B + ̂∇ B ξ A ) + (f ′ ) 2 ĝ AB u.<br />

In particular, if h is divergence free,<br />

∇ k ∇ k h 00 = u ′′ + (n + 4)f −1 f ′ u ′ + 2(n + 1)f −2 (f ′ ) 2 u + f −2 ̂∇A ̂∇A u,<br />

∇ k ∇ k h A0 = ξ ′′ A + (n)f −1 f ′ ξ ′ A − [f −2 (f ′ ) 2 + f −1 f ′′ ]ξ A<br />

+f −2 ̂∇D ̂∇D ξ A + 2f −1 f ′ d A u,<br />

∇ k ∇ k h AB = ĥ′′ AB + (n − 4)f −1 f ′ ĥ ′ AB − [(2n − 4)f −2 (f ′ ) 2 + 2f −1 f ′′ ]ĥAB<br />

+f −2 ̂∇D ̂∇D ĥ AB + 2f −1 f ′ ( ̂∇ A ξ B + ̂∇ B ξ A ) + (f ′ ) 2 ĝ AB u.<br />

In order to study the action of the zero order terms of the Lichnerowicz<br />

Laplacian, we compute the curvatures of g. A straightforward computation<br />

shows that the non trivial components of the Riemann Christoffel curvature<br />

are<br />

R A 0B0 = −f −1 f ′′ δ A B,<br />

R 0 A0B = −ff ′′ ĝ AB ,<br />

R A BCD = ̂R A BCD − (f ′ ) 2 (ĝ BD δ A C − ĝ BC δ A D).<br />

Thus the non trivial components of the Ricci curvature are<br />

R 00 = −nf −1 f ′′ , R AB = ̂R AB − [ff ′′ + (n − 1)(f ′ ) 2 ]ĝ AB .


8 E. DELAY<br />

Then, the zero order terms that appear in the Lichnerowicz Laplacian are<br />

2(Riem h) 00 = −2f −3 f ′′ Tr bg ĥ = 2f −1 f ′′ u,<br />

2(Ric h) 00 = −2nf −1 f ′′ u,<br />

2(Riem h) 0A = 2f −1 f ′′ ξ A ,<br />

2(Ric h) 0A = f −2 ̂RC A ξ C − ((n + 1)f −1 f ′′ + (n − 1)f −2 (f ′ ) 2 )ξ A ,<br />

2(Riem h) AB = 2f −2 ̂RACBD ĥ AB + 2f −2 (f ′ ) 2 (ĥAB − Tr bg ĥ ĝ AB ) − 2ff ′′ uĝ AB ,<br />

= 2f −2 ̂RACBD ĥ AB + 2f −2 (f ′ ) 2 ĥ AB ,<br />

2(Ric h) AB = 2f −2 (Ric bg ĥ) AB − 2(f −1 f ′′ + (n − 1)f −2 (f ′ ) 2 )ĥAB.<br />

Lemma 4.3. Let h be a trace free covariant symmetric 2-tensor decomposed<br />

as is 4.1. The components of the Lichnerowicz Laplacian of h are<br />

∆ L h 00 = −u ′′ − nf −1 f ′ u ′ − f −2 ̂∇A ̂∇A u + 4f −3 f ′ ̂∇A ξ A ,<br />

∆ L h A0 = −ξ ′′ A − (n − 2)f −1 f ′ ξ ′ A + [(3n + 4)f −2 (f ′ ) 2 + nf −1 f ′′ ]ξ A<br />

−f −2 ( ̂∇ D ̂∇D ξ A − ̂R C Aξ C ) + 2f −3 f ′ ̂∇D h AD − 2f −1 f ′ d A u,<br />

∆ L h AB = −ĥ′′ AB − (n − 4)f −1 f ′ ĥ ′ AB − 4f −2 (f ′ ) 2 ĥ AB<br />

+f −2 ̂∆L ĥ AB − 2f −1 f ′ ( ̂∇ A ξ B + ̂∇ B ξ A ) − (f ′ ) 2 ĝ AB u.<br />

In particular, if h is divergence free, we obtain<br />

∆ L h 00 = −u ′′ − (n + 4)f −1 f ′ u ′ − 2(n + 1)(f −2 (f ′ ) 2 + f −1 f ′′ )u − f −2 ̂∇A ̂∇A u,<br />

∆ L h A0 = −ξ ′′ A − nf −1 f ′ ξ ′ A − [(n + 2)f −2 (f ′ ) 2 + (n − 2)f −1 f ′′ ]ξ A<br />

−f −2 ( ̂∇ D ̂∇D ξ A − ̂R C Aξ C ) − 2f −1 f ′ d A u,<br />

∆ L h AB = −ĥ′′ AB − (n − 4)f −1 f ′ ĥ ′ AB − 4f −2 (f ′ ) 2 ĥ AB<br />

+f −2 ̂∆L ĥ AB − 2f −1 f ′ ( ̂∇ A ξ B + ̂∇ B ξ A ) − (f ′ ) 2 ĝ AB u.<br />

5. Hodge type decomposition for trace free symmetric<br />

2-tensors<br />

We recall here how to construct an L 2 orthonormal eigenbasis for trace<br />

free symmetric covariants 2-tensors on a compact Einstein manifold N with<br />

Ric(ĝ) = κ(n − 1)ĝ.<br />

We recall the classical L 2 -orthogonal Hodge-de Rahm decomposition for one<br />

forms on N :<br />

(5.1) C ∞ (N, T 1 ) = Im d bg ⊕ Ker d ∗ bg .<br />

We also recall the L 2 -orthogonal decomposition for trace free symmetric<br />

2-tensors on N [1] :<br />

(5.2) C ∞ (N, ˚S 2 ) = Im ˚L bg ⊕ Ker div bg ,<br />

where, we recall that for a one form ξ, ˚L bg ξ = ˚L bg ξ + 1 n d∗ bg<br />

ξĝ is the ĝ-trace free<br />

part of L bg ξ.<br />

Let (Ψ i ) i∈N be an orthonormal basis of L 2 (N) of eigenfunctions, with<br />

∇ ∗ bg ∇ bgΨ i = λ i Ψ i , 0 = λ 0 < λ 1 ≤ λ 2 ≤ ...


For all i ∈ N\{0}<br />

and<br />

<strong>TT</strong>-<strong>EIGENTENSORS</strong> <strong>ON</strong> SOME AH MANIFOLDS 9<br />

̂∆ H dΨ i = λ i dΨ i ⇐⇒ ∇ ∗ bg ∇ bgdΨ i = [λ i − (n − 1)κ]dΨ i ,<br />

∫<br />

N<br />

∫<br />

〈dΨ i , dΨ j 〉 bg dθ =<br />

N<br />

Ψ j ∇ ∗ ∇Ψ i dθ = λ i<br />

∫<br />

N<br />

Ψ i Ψ j dθ.<br />

From the Hodge decomposition theorem (see in particular equation 5.1),<br />

the L 2 (N, T 1 ) orthonormal family { 1 √ λi<br />

dΨ i } i∈N\{0} can be completed with<br />

an L 2 (N, T 1 ) orthonormal family {ν j } j∈N with<br />

∇ ∗ bg ∇ bgν j = µ j ν j , d ∗ bg ν j = 0<br />

to obtain an orthonormal basis {η k } k∈N of L 2 (N, T 1 ) composed of eigenforms<br />

:<br />

∇ ∗ bg ∇ bgη k = α k η k .<br />

Thank’s to some easy integrations by parts, we get: 1<br />

∫<br />

k ,<br />

N〈˚Lη ˚Lη q 〉 bg dθ = 1 ∫<br />

[〈∆ bg η k , η q 〉 bg − (n − 1)κ〈η k , η q 〉 bg + n − 2<br />

2 N<br />

n<br />

d∗ bg η kd ∗ bg η q]dθ<br />

⎧ ∫<br />

α k −(n−1)κ<br />

2<br />

〈η k , η q 〉 bg dθ if d<br />

⎪⎨<br />

∗ bg η k or d ∗ bg η q = 0,<br />

N<br />

∫<br />

= n−1<br />

2 √ √ [ 2 λ i λj n λ iλ j − κ(λ i + λ j )] Ψ i Ψ j dθ<br />

N<br />

⎪⎩ if η k = √ 1<br />

λi<br />

dΨ i and η q = √ 1 dΨ j ,<br />

λj<br />

1<br />

so the family { ˚Lη<br />

‖˚Lη k ‖ k } k∈N is orthonormal. If the sectional curvature is<br />

L 2<br />

constant, we also have<br />

∇ ∗ bg ∇ bg ˚L bg η k = [α k − κ(n + 2)]˚L bg η k .<br />

In this case from the decomposition of trace free symmetric 2-tensors (see<br />

equation 5.2), the L 2 (N, ˚S 1<br />

2 ) orthonormal family { ˚Lη<br />

‖˚Lη k ‖ k } k∈N can be<br />

L 2<br />

completed with an L 2 (N, ˚S 2 ) orthonormal family {σ l } l∈N such that<br />

∇ ∗ bg ∇ bgσ l = β l σ l , div bg σ l = 0,<br />

to obtain an orthonormal basis {Υ m } m∈N of L 2 (N, ˚S 2 ) composed of eigentensors,<br />

with<br />

∇ ∗ bg ∇ bgΥ m = γ m Υ m .<br />

In the more general case where N is only Einstein, then we have<br />

̂∆ L ˚Lbg η k = [α k + κ(n − 1)]˚L bg η k ,<br />

1<br />

So the orthonormal family { ˚Lη<br />

‖˚Lη k ‖ k } k∈N can be completed with an<br />

L 2<br />

L 2 (N, ˚S 2 ) orthonormal family {π l } l∈N such that<br />

̂∆ L π l = β l π l , div bg π l = 0,<br />

1 This is at this step we really need (N, bg) to be Einstein because the Laplacian ∆ bg −<br />

Ric(bg) that appear in the calculation is not the Hodge Laplacian due to the fact we work<br />

on symmetric 2-tensors and not on (skew symmetric) 2-forms


10 E. DELAY<br />

to obtain an orthonormal basis {Φ m } m∈N of L 2 (N, ˚S 2 ) of eigentensors, with<br />

̂∆ L Φ m = ε m Φ m .<br />

Remark. When (N, ĝ) is the standard unit sphere of R n+1 or the real projective<br />

space, the spectrum of the Lichnerowicz Laplacian is given explicitly<br />

in [3] (see also [17] and [18]).<br />

6. Conclusion<br />

In this section we work near the infinity of asymptotically hyperbolic<br />

warped product manifold ]0, +∞[×N equipped with the metric<br />

g = dr 2 + f 2 (r)ĝ,<br />

where ĝ is for the moment assumed to be of constant sectional curvature.<br />

Theorem 6.1. For any λ ≥ n2<br />

4<br />

+ 2, there is no non-trivial trace free, divergence<br />

free, L 2 covariant symmetric 2-tensor on (]0, +∞[×N, g) satisfying<br />

∇ ∗ ∇h = λh.<br />

Proof. For h a symmetric 2-tensor field on ]0, +∞[×N, we decompose<br />

h = udr 2 + ξ ⊗ dr + dr ⊗ ξ + ĥ,<br />

where ĥ is the orthogonal projection of h on {r} × N and ξ is a one form<br />

on {r} × N. We assume first that h is a trace free L 2 eigentensor for an<br />

eigenvalue λ > n2<br />

4 + 2 so ∇ ∗ ∇h = λh.<br />

We assume also that h is divergence free (see lemma 4.1):<br />

(6.1) u ′ + (n + 1)f −1 f ′ u − f −2 ̂d∗ ξ = 0,<br />

(6.2) ξ ′ + nf −1 f ′ ξ − f −2 div bg ĥ = 0.<br />

From Lemma 4.2, we have for the components of h:<br />

(6.3) u ′′ + (n + 4)f −1 f ′ u ′ + [2(n + 1)f −2 (f ′ ) 2 + λ]u − f −2 ∇ ∗ bg ∇ bgu = 0,<br />

(6.4) ξ ′′ +nf −1 f ′ ξ ′ −[f −2 (f ′ ) 2 +f −1 f ′′ −λ]ξ −f −2 ∇ ∗ bg ∇ bgξ +2f −1 f ′ d bg u = 0,<br />

(6.5)<br />

ĥ ′′ + (n − 4)f −1 f ′ ĥ ′ − [(2n − 4)f −2 (f ′ ) 2 + 2f −1 f ′′ − λ]ĥ<br />

−f −2 ∇ ∗ bg ∇ bgĥ + 4f −1 f ′ L bg ξ + (f ′ ) 2 uĝ = 0.<br />

The ĝ-trace free part of equation 6.5 gives<br />

(6.6)<br />

˚h′′ + (n − 4)f −1 f ′˚h′ − [(2n − 4)f −2 (f ′ ) 2 + 2f −1 f ′′ − λ]˚h<br />

−f −2 ∇ ∗ bg ∇ bg˚h + 4[f −1 f ′ L bg ξ + 1 n ̂d ∗ ξĝ] = 0,<br />

where ˚h is the ĝ-trace free part of ĥ.<br />

With the change of variables u = f − n+4<br />

2 v, ξ = f − n 2 ζ and ˚h = f − n−4<br />

2 b,<br />

equations 6.1 and 6.2 become respectively<br />

(6.7) (f n 2 −1 v) ′ = f n 2 −1 d ∗ bg ζ<br />

(6.8) (f n 2 ζ) ′ = f n 2 (div bg b + 1 n f −2 d bg v),


<strong>TT</strong>-<strong>EIGENTENSORS</strong> <strong>ON</strong> SOME AH MANIFOLDS 11<br />

while equation 6.3, 6.4 and 6.6 become respectivily<br />

(6.9) v ′′ + [λ −<br />

(6.10)<br />

n(n − 2)<br />

4<br />

f −2 (f ′ ) 2 − n + 4 f −1 f ′′ ]v − f −2 ∇ ∗ bg<br />

2<br />

∇ bgv = 0,<br />

ζ ′′ +[λ− n2 − 2n + 4<br />

f −2 (f ′ ) 2 − n + 2 f −1 f ′′ ]ζ−f −2 ∇ ∗ bg<br />

4<br />

2<br />

∇ bgζ = −2f −2 (f −1 f ′ d bg v),<br />

(6.11)<br />

b ′′ + [λ − n2 −2n+8<br />

4<br />

f −2 (f ′ ) 2 − n 2 f −1 f ′′ ]b − f −2 ∇ ∗ bg ∇ bgb<br />

= −4f −2 [f −1 f ′ L bg ζ + 1 ̂d n ∗ ζĝ].<br />

We now decompose each component of h according to the L 2 basis described<br />

in section 5. This yield<br />

v(r, θ) = ∑ i∈N<br />

x i (r)Ψ i (θ),<br />

where x i (r) = ∫ N u(r, θ)Ψ i(θ)dθ is smooth and the same is true for<br />

ζ(r, θ) = ∑ k∈N<br />

y k (r)η k (θ),<br />

b(r, θ) = ∑ m∈N<br />

z m (r)Υ m (θ).<br />

From equation 6.9, each component x i of v satisfies<br />

(6.12) x ′′ n(n − 2)<br />

+ [λ − f −2 (f ′ ) 2 − n + 4 f −1 f ′′ − f −2 λ i ]x = 0.<br />

4<br />

2<br />

This equation can be rewritten<br />

(6.13) x ′′ + [α 2 + O(e −εr )]x = 0,<br />

√<br />

where α = λ − n2<br />

4<br />

− 2. An asymptotic integration (see theorem 7.1 of<br />

appendix) gives<br />

(6.14) x = (A 1 + O(e −εr ))e αir + (A 2 + O(e −εr ))e −αir ,<br />

and<br />

(6.15) x ≡ 0 ⇐⇒ A 1 = A 2 = 0.<br />

We have<br />

d bg v = ∑ i∈N<br />

x i d bg Ψ,<br />

with each of the x i ’s having the same asymptotic as x in 6.14. From equation<br />

6.10, each components y k of ζ satisfies<br />

(6.16)<br />

y ′′ +[λ− n2 − 2n + 4<br />

4<br />

f −2 (f ′ ) 2 − n + 2 f −1 f ′′ −f −2 µ k ]y =<br />

2<br />

which can be written<br />

⎧<br />

⎨ O(e −εr )<br />

(6.17) y ′′ + [α 2 + O(e −εr )]y = or<br />

⎩<br />

0<br />

⎧<br />

⎨<br />

⎩<br />

.<br />

−2f −2 (f −1 f ′ x)<br />

or<br />

0<br />

,


12 E. DELAY<br />

From asymptotic integration (see corollary 7.2 of appendix), we thus have<br />

(6.18) y = (B 1 + O(e −εr ))e αir + (B 2 + O(e −εr ))e −αir ,<br />

and if x = 0,<br />

(6.19) y ≡ 0 ⇐⇒ B 1 = B 2 = 0.<br />

In a similar way since<br />

each components z m of b satisfies<br />

(6.20)<br />

L ζ ĝ + 1 n d∗ bg ζĝ = ∑ k∈N<br />

y k (L ηk ĝ + 1 n d∗ bg η kĝ),<br />

z ′′ + [λ − n2 − 2n + 8<br />

f −2 (f ′ ) 2 − n 4<br />

2 f −1 f ′′ − f −2 α m ]z =<br />

which can also be written<br />

⎧<br />

⎨ O(e −εr )<br />

(6.21) z ′′ + [α 2 + O(e −εr )]z = or<br />

⎩<br />

0<br />

⎧<br />

⎨<br />

.<br />

⎩<br />

−4f −2 f −1 f ′ y<br />

or<br />

0<br />

From asymptotic integration (see corollary 7.2 in appendix), once again<br />

(6.22) z = (C 1 + O(e −εr ))e αir + (C 2 + O(e −εr ))e −αir ,<br />

and if y = 0,<br />

(6.23) z ≡ 0 ⇐⇒ C 1 = C 2 = 0.<br />

Now, the squared norm of h is<br />

|h| 2 = u 2 + 2f −2 |ξ| 2 bg + f −4 |ĥ|2 bg<br />

= u 2 (1 + 1 n ) + 2f −2 |ξ| 2 bg + f −4 |˚h| 2 bg<br />

= f −n [f −4 v 2 (1 + 1 n ) + 2f −2 |ζ| 2 bg + |b|2 bg ],<br />

and the mesure relative to g is dµ = f n drdθ.<br />

But as h ∈ L 2 , the only possibility is C 1 = C 2 = 0 for all components<br />

z m of b. If we include this fact in equation 6.8, we obtain that all the<br />

components y k of ζ are of order O(e −εr ) thus B 1 = B 2 = 0 for all of them.<br />

Now we insert this last fact in equation 6.7 and we obtain that the x i ’s are<br />

also of order O(e −εr ). Finally from 6.15, for all i ∈ N, x i = 0, from 6.19, for<br />

all k ∈ N, y k = 0, and from 6.23 all the z m ’s are equals to zero. We thus<br />

conclude that h = 0 and this complete the proof when λ > n2<br />

4 + 2.<br />

Assume now that λ = n2<br />

4<br />

+ 2 and repeat the preceding arguments step<br />

by step. Here we have α = 0 and from Theorem 7.3 and Corollary 7.4 the<br />

equations 6.14, 6.18, 6.22 are substituted by<br />

(6.24) x = A 1 + (A 2 + O(r 2 e −εr )r,<br />

(6.25) y = B 1 + (B 2 + O(r 2 e −εr )r,<br />

,


<strong>TT</strong>-<strong>EIGENTENSORS</strong> <strong>ON</strong> SOME AH MANIFOLDS 13<br />

(6.26) z = C 1 + (C 2 + O(r 2 e −εr )r.<br />

The end of the proof proceeds as before.<br />

The adaptation of the proof above for the Lichnerowicz Laplacian in place<br />

of the rough Laplacian is achieved with some minor modifications. This<br />

proceed using Lemma 4.3 and the end of section 5 only assuming here that<br />

N is compact and Einstein. We finally obtain:<br />

Theorem 6.2. For all λ ≥ n(n−8)<br />

4<br />

, there is no non-trivial trace free, divergence<br />

free, L 2 covariant symmetric 2-tensor on (]0, +∞[×N, g) satisfying<br />

∆ L h = λh.<br />

7. Appendix : asymptotic integration for ordinary differential<br />

equations<br />

In this section, we recall some results of asymptotic integration which are<br />

certainly well known since a long time. They can be found partly in [16]. We<br />

detail some parts of the proof given there to obtain a more precise version<br />

in the particular case which is of interest here.<br />

Theorem 7.1. Let a be a continuous function on ]0, ∞[ such that for r<br />

going to infinity, a(r) = O(e −εr ) for some ε > 0. If α is a positive real then<br />

any solution to the equation<br />

(7.1) y ′′ + (α 2 + a)y = 0,<br />

satisfies<br />

(7.2) y = [A 1 + O(e −εr )]e iαr + [A 2 + O(e −εr )]e −iαr ,<br />

(7.3) y ′ = [A 1 iα + O(e −εr )]e iαr − [A 2 iα + O(e −εr )]e −iαr ,<br />

for some constants A 1 and A 2 , and<br />

y ≡ 0 iff A 1 = A 2 = 0.<br />

Moreover, for any given A 1 and A 2 there exists at least one solution to 7.1<br />

with the asymptotics 7.2 7.3.<br />

Proof. From [16], Theorem XI.8.1 page 370, the theorem given here is true<br />

with the O(e −εr ) there replaced with o(1). So we only have to prove this<br />

behavior at infinity. Let us denote by u 0 (r) = e iαr and v 0 (r) = e −iαr . We<br />

define the matrics<br />

A =<br />

(<br />

0 1<br />

−α 2 0<br />

) (<br />

0 0<br />

, B =<br />

−a 0<br />

Since Z ′ = AZ, the usual change of variable<br />

( ) ( )<br />

y ξ1<br />

y ′ = Z = Zξ,<br />

ξ 2<br />

gives<br />

In particular<br />

) (<br />

u0 v<br />

, Z =<br />

0<br />

u ′ 0 v 0<br />

′<br />

ξ ′ (r) = F (r, ξ(r)), where F (r, ξ) = Z −1 BZξ.<br />

‖F (r, ξ)‖ ≤ C|a|‖ξ‖,<br />

)<br />

.<br />


14 E. DELAY<br />

where C is a constant, ( so)<br />

from [16], Theorems X.1.1 and X.1.2 page 274,<br />

A1<br />

lim r→∞ ξ = ξ ∞ =: exists. Moreover (see [16] equation (1.11) page<br />

A 2<br />

275),<br />

∫ ∞<br />

‖ξ(r) − ξ ∞ ‖ ≤ C 1 |a(s)|ds,<br />

where C 1 is a constant. As a(r) = O(e −εr ), the integral on the right-hand<br />

side of this last equation is of order O(e −εr ). This concludes the proof. □<br />

Corollary 7.2. Let a, b be two continuous functions on ]0, ∞[ such that for<br />

r going to infinity, a(r), b(r) = O(e −εr ) for some ε > 0. If α is a positive<br />

real, then any solution to the equation<br />

(7.4) y ′′ + (α 2 + a)y = b,<br />

satisfies<br />

(7.5) y = [A 1 + O(e −εr )]e iαr + [A 2 + O(e −εr )]e −iαr ,<br />

(7.6) y ′ = [A 1 iα + O(e −εr )]e iαr − [A 2 iα + O(e −εr )]e −iαr ,<br />

for some constants A 1 and A 2 .<br />

Proof. From Theorem 7.1, there exist two solutions u 1 , v 1 to the equation<br />

7.1 such that<br />

u 1 = [1+O(e −εr )]e iαr +O(e −εr )e −iαr , u ′ 1 = [iα+O(e −εr )]e iαr +O(e −εr )e −iαr ,<br />

v 1 = O(e −εr )e iαr +[1+O(e −εr )]e −iαr , v 1 ′ = O(e −εr )e iαr −[iα+O(e −εr )]e −iαr .<br />

Let us consider the matrics<br />

(<br />

) ( )<br />

0 1<br />

u1 v<br />

A =<br />

−α 2 , Z<br />

− a 0 1 =<br />

1<br />

u ′ 1 v 1<br />

′ .<br />

Once again Z 1 ′ = AZ 1, so the usual change of variable<br />

( ) ( )<br />

y η1<br />

y ′ = Z 1 = Z<br />

η 1 η,<br />

2<br />

gives<br />

( )<br />

η ′ (r) = G(r), where G(r) = Z1<br />

−1 0<br />

.<br />

b<br />

In particular from the asymptotic behavior of a and b we have<br />

‖G(r)‖ ≤ Ce −εr ,<br />

where C is a constant. This shows that ∫ ( )<br />

∞ η ′ converges absolutely so that<br />

A1<br />

lim r→∞ η = η ∞ = exists. Moreover<br />

A 2<br />

‖η(r) − η ∞ ‖ = ‖<br />

∫ ∞<br />

We finally obtain η =<br />

r<br />

η ′ (s)ds‖ ≤<br />

(<br />

A1<br />

∫ ∞<br />

r<br />

r<br />

‖η ′ (s)‖ds ≤ C<br />

∫ ∞<br />

r<br />

e −εs ds = C ε e−εr .<br />

A 2<br />

)<br />

+ O(e −εr ), which concludes the proof. □<br />

Similarly to theorem 7.1 and to corollary 7.2:


<strong>TT</strong>-<strong>EIGENTENSORS</strong> <strong>ON</strong> SOME AH MANIFOLDS 15<br />

Theorem 7.3. Let a be a continuous function on ]0, ∞[ such that for r<br />

going to infinity, a(r) = O(e −εr ) for some ε > 0. Then any solution to the<br />

equation<br />

(7.7) y ′′ + ay = 0,<br />

satisfies<br />

(7.8) y = A 1 + [A 2 + O(r 2 e −εr )]r,<br />

(7.9) y ′ = A 2 + O(r 2 e −εr ),<br />

for some constants A 1 and A 2 , and<br />

y ≡ 0 iff A 1 = A 2 = 0.<br />

Moreover, for any given A 1 and A 2 there exists at least one solution to 7.7<br />

with the asymptotics 7.8 7.9.<br />

Proof. The proof proceeds as for theorem 7.1 with α = 0, u 0 (r) = 1 and<br />

v 0 (r) = r. Here we will only have<br />

for r ≥ r 0 . Then<br />

‖F (r, ξ)‖ ≤ C|a|r 2 ‖ξ‖,<br />

∫ ∞<br />

‖ξ(r) − ξ ∞ ‖ ≤ C 1 |a(s)|s 2 ds,<br />

where C 1 is a constant. As a(r) = O(e −εr ), the integral on the right-hand<br />

side of this last equation is of order r 2 e −εr and this complete the proof. □<br />

Corollary 7.4. Let a, b be two continuous functions on ]0, ∞[ such that for<br />

r going to infinity, a(r), b(r) = O(e −εr ) for some ε > 0. Then any solution<br />

to the equation<br />

(7.10) y ′′ + ay = b,<br />

satisfies<br />

(7.11) y = A 1 + [A 2 + O(r 2 e −εr )]r,<br />

r<br />

(7.12) y ′ = A 2 + O(r 2 e −εr ),<br />

for some constants A 1 and A 2 .<br />

Proof. The proof proceeds as in Corollary 7.2, using Theorem 7.3.<br />

□<br />

References<br />

1. M. Berger and D. Ebin, Some decompositions of the space of symmetric tensors on a<br />

Riemannian manifold, Jour. Diff. Geom. 3 (1969), 379–392.<br />

2. A.L. Besse, Einstein manifolds, Ergebnisse d. Math. 3. folge, vol. 10, Springer, Berlin,<br />

1987.<br />

3. M. Boucetta, Spectre des laplaciens de Lichnerowicz sur les sphères et les projectifs<br />

réels, Publ. Matematiques 43 (1999), no. 2, 451–483.<br />

4. Huai-Dong Cao, Richard S. Hamilton, and Tom Ilmanen, Gaussian densities and<br />

stability for some ricci solitons, math.DG/0404165.<br />

5. X. Dai, X. Wang, and G. Wei, On the stability of riemannian manifold with parallel<br />

spinors, Invent. Math. 161 (2005), no. 1, 151–176.<br />

6. E. Delay, Prescription de la courbure de ricci au voisinage de la métrique hyperbolique,<br />

C. R. Acad. Sci. Paris 326 (1998), no. I, 335–338.


16 E. DELAY<br />

7. , Etude locale d’opérateurs de courbure sur l’espace hyperbolique, J. Math. Pures<br />

Appli. 78 (1999), 389–430.<br />

8. , Essential spectrum of the lichnerowicz laplacian on two tensor on asymptotically<br />

hyperbolic manifolds, J. of Geom. and Physics 43 (2002), 33–44.<br />

9. , Study of some curvature operators in the neighbourhood of an asymptotically<br />

hyperbolic einstein manifold, Advances in Math. 168 (2002), 213–224.<br />

10. E. Delay and M. Herzlich, Ricci curvature in the neighbourhood of rank-one symmetric<br />

spaces, J. Geometric Analysis 11 (2001), no. 4, 573–588.<br />

11. D. DeTurck, Metrics with prescribed Ricci curvature, Seminar in Differential Geometry<br />

(S.-T. Yau, ed.), Ann. Math. Studies, vol. 102, Princeton, 1982.<br />

12. H. Donnelly, The differential form specrum of hyperbolic space, Manuscripta Math 33<br />

(1981), 365–385.<br />

13. G. Miele D.V. Fursaev, Cones, spins and heat kernel, Nuclear Phys. B 3 (1997),<br />

697–723.<br />

14. Gibbons, G. W. Page, and D. N. Pope, Einstein metrics on S 3 , R 3 and R 4 bundles,<br />

Comm. Math. Phys. 3 (1990), 529–553.<br />

15. G.W. Gibbons, Sean E. Hartnoll, and C.N. Pope, Bohm and einstein-sasaki metrics.<br />

black holes and cosmological event horizons, hep–th/0208031.<br />

16. P. Hartman, Ordinary differential equation, second ed., Classics in mathematics, 2002.<br />

17. A. Higuchi, Symmetric tensor spherical harmonics on the N-sphere and their application<br />

to the de sitter group SO(N,1), J. Math. Phys. 28 (1987), 1553–1566.<br />

18. , Erratum : Symmetric tensor spherical harmonics on the N-sphere and their<br />

application to the de sitter group SO(N,1)[J. Math. Phys. 28, 1553 (1987)], J. Math.<br />

Phys. 43 (2002), 6385.<br />

19. J. Isenberg and M. Weaver, On the area of the symmetry orbits in T 2 symmetric<br />

spacetimes, Class. Quantum Grav. 20 (2003), 3783–3796, gr-qc/0304019.<br />

20. A. Kehagias and J.G. Russo, Hyperbolic spaces in string and M-theory, J. High Energy<br />

Phys. 7 (2000), 19 pp.<br />

21. J.M. Lee, Fredholm operators and Einstein metrics on conformally compact manifolds,<br />

Memoirs AMS, in press, math.DG/0105046.<br />

22. A. Lichnerowicz, Propagateurs et commutateurs en relativité générale, Pub. Math. de<br />

l’IHES 10 (1961), 5–56.<br />

23. R. Mazzeo, Unique continuation at infinity and embedded eigenvalues for asymptotically<br />

hyperbolic manifolds, Am. J. Math. 113 (1991), 25–45.<br />

Erwann Delay, Laboratoire d’analyse non linéaire et géométrie, Faculté<br />

des Sciences, 33 rue Louis Pasteur, 84000 Avignon, France<br />

E-mail address: Erwann.Delay@univ-avignon.fr<br />

URL: http://www.math.univ-avignon.fr/Delay

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!