curvature in generalized Roberston-Walker spacetimes
Ximin Liu and Biaogui Yang
Abstract.In this paper we study stable spacelike hyersurfaces with con- stant scalar curvature in generalized Roberston-Walker spacetimeMn+1=
−I×φFn.
M.S.C. 2000: 53B30, 53C42, 53C50.
Key words: stability, spacelike hypersurface, constant scalar curvature, generalized Roberston-Walker spacetime.
1 Introduction
HyersurfacesMnwith constantr-mean curvature in Riemannian manifolds or Lorentz manifoldsMn+1(c) with constant sectional curvaturecare critical points of some area functional variations which keep constant a certain volume function. Stable hyersur- faces with constant mean curvature(CMC) (or constant r-mean curvature) in real space form are very interesting geometrical objects that were investigated by many geometricians. Barbosa and do Carmo [2] gave definition of stability of hyersurfaces with constant mean curvature in the Eucildean space Rn+1 and proved the round spheres are the only compact stable hyersurfaces with CMC inRn+1. Later, Barbosa, do Carmo and Eschenburg [3] extended ambient spaces to Riemannian manifolds and obtained the corresponding results. In [5] Barbosa and Oliker discussed stable spacelile hyersurfaces with CMC in Lorentz manifolds. At the same time, Alencar, do Carmo and Colares [1] investigated stable hyersurfaces with constant scalar curvature in Riemannian manifolds and obtained geodesic sphere is the only stable compact ori- entable hyersurface in Riemannain spaces. On the other hand, Barbosa and Colares [4] studied compact hyersurfaces without boundary immersed in space forms with constant r-mean curvature. Recently, Liu and Deng [9] also discussed stable space- like hyersurfaces with constant scalar curvature in de Siter spaceS1n+1. Barros, Brasil and Caminha [6] classified strongly stable spacelike hypersurfaces with constant mean curvature whose warping function satisfied a certain convexity condition.
Balkan Journal of Geometry and Its Applications, Vol.13, No.1, 2008, pp. 66-76.∗
c
°Balkan Society of Geometers, Geometry Balkan Press 2008.
In this paper we will study stable spacelike hypersurfaces with constant scalar curvature in generalized Roberston-Walker spacetimeMn+1 =−I×φFn.
2 Preliminaries
Consider Fn an n-dimensional manifold, let I be a 1-dimensional manifold (either a circle or an open interval of R). We denote by Mn+1 = −I×φ Fn the (n+ 1)- dimensional product manifoldI×F endowed with the Lorentzian metric
(2.1) g=h,i=−dt2+f2(t)h,iM,
where f > 0 is positive function on I, and h,iM stands for the Riemannian metric onFn. We refer to −I×φFn as a generalized Robertson-Walker (GRW) spacetime.
In particular, when the Riemannian factorFn has constant sectional curvature, then
−I×φFn is classically called a Robertson-Walker (RW) spacetime.
A vector fieldV on a Lorentz manifoldMn+1is said to be conformal if
(2.2) LVg= 2ψg,
for some smooth function ψ: Mn+1 → R, where L stands for the Lie derivative of Lorentz metric ofMn+1. The functionψis called the conformal factor ofV.V ∈T M is conformal if and only if
h∇XV, Yi+h∇YV, Xi= 2ψhX, Yi, (2.3)
for allX, Y ∈T(M).
Any Lorentz manifold Mn+1, possessing a globally defined, timelike conformal vector field is said to be a conformally stationary (CS) spacetime.
Let x : Mn → Mn+1 denote an orientable spacelike hyersurface in the time- oriented Lorentz manifold Mn+1 and N be a globally defined unit normal vector field onMn. ∇ and∇ denote the Levi-Civita connection of Mn and ambient space Mn+1respectively. R and Ric denote the curvature tensor and Ricci curvature tensor onMn+1 respectively, which are defined by
(2.4) R(X, Y)Z =∇X∇YZ− ∇Y∇XZ− ∇[X,Y]Z, and
(2.5) R(W, Z, X, Y) =h∇X∇YZ, Wi − h∇Y∇XY, Wi − h∇[X,Y]Z, Wi, then
(2.6) Ric(X, Y) =
n+1X
k=1
R(ek, X, ek, Y),
whereX, Y, Z, W ∈T M, and{ek}nk=1 is a basis ofTpM,en+1=N. In particular we have
Ric(N, N) = Xn
k=1
R(ek, N, ek, N).
(2.7)
The shape operator A associated toN ofMn, defined by
(2.8) A =−∇N (i.e Aek =−∇ekN)
is a self-adjoint linear operator in each tangent space TpM. Its eigenvalues are the principal curvatures of immersion and are represented byλ1, λ2, · · · , λn.The elemen- tary symmetic functionsSr associated to A can be defined, using the characteristic polynomial of A, by
det(tI−A) = Xn k=0
(−1)kSktn−k,
where S0 = 1. If p∈ M, and {ek} is a basis of TpM formed by eigenvector of Ap, with corresponding eigenvaluesλk, one immediately sees that
Sr=σr(λ1,· · · , λn),
whereσr is ther-th elementary symmetric polynomial. In particular
(2.9) kAk2=X
k
λ2k=S12−2S2, and
(2.10) X
k
λ3k=S13−3S1S2+ 3S3.
Ther-th classical Newton transformation Pr onM is defined as following P0= I,
Pr=SrI−APr−1, 1≤r≤n.
Associated to each Newton transformation Pr of immersion x : Mn → Mn+1, we have a second order differential operator defined by
(2.11) Lr(f) = trace(Pr◦Hessf).
WhenMn+1has constant sectional curvature, then
(2.12) Lr(f) = div(Pr∇f),
where div stands for the divergence of a vector field on M, it was proved by H.
Rosenberg in [12].
Remark 1.1.According (2.11) or (2.12), whenr= 0, L0f = div(P0∇f) =4f is Laplace operator onMn, and ifr= 1, then
L1f = div[P1◦hessf] = div[(S1I−AP0)◦hessf]
= X
i,j
(S1δij−hij)fij
(2.13)
become Cheng-Yau’s operator2onMn, wherehij andfij denote the component of A and hessf respectively.
3 The variational problem in Lorentz manifolds
Letx:Mn →Mn+1 denotes an orientable spacelike hyersurface in the time-oriented Lorentz manifoldMn+1andN be a globally defined unit normal vector field onMn. A variation ofxis a smooth mapX :Mn×(−ε, ε)→Mn+1 satisfying the following conditions:
(1) For t ∈ (−ε, ε), the map Xt : Mn → Mn+1 given by Xt(p) = X(t, p) is a spaelike immersion such thatX0=x.
(2)Xt|∂M =x|∂M, for allt∈(−ε, ε).
The variational field vector associated the variationXis vector fieldX∗(∂t∂) = ∂X∂t. Letf =h∂X∂t, Ni, we have
(3.1) ∂X
∂t = (∂X
∂t )>−f N,
where>denotes tangential components. The balance of volume of the variationX is the functionV : (−ε, ε)→R given by
(3.2) V(t) =
Z
M×[0,t]
X∗(dM), wheredM denotes the volume element ofM.
The area functionalA: (−ε, ε)→Ris given by
(3.3) A(t) =
Z
M
S1dMt,
wheredMt denotes the volume element of the metric induced inM byXt. Then we have the following classical result.
Lemma 3.1.LetMn+1be a time-oriented Lorentz manifold andx:Mn→Mn+1 a spacelike hyersurface. IfX:Mn×(−ε, ε)→Mn+1 is a variation ofx, then
(i)
(3.4) dV(t)
dt |t=0= Z
M
f dM; (ii)
(3.5) ∂(dMt)
∂t = (S1+ div(∂X
∂t )>)dMt. Proof.For (i) see [3, 9], and for (ii) see [4, 11].2
Barros, Brasil and Caminha [6] proved the following proposition:
Proposition 3.2.Let x:Mn →Mn+1 be a spacelike hypersurface of the time- oriented Lorentz manifoldMn+1, andNbe a globally defined unit normal vector field onMn. IfX:Mn×(−ε, ε)→Mn+1 is a variation ofx, then
(3.6) dS1
dt =4f−(Ric(N, N) +kAk2)f+h(∂X
∂t )>,∇S1i.
Supposeλis a constant, andJ : (−ε, ε)→R is given by
(3.7) J(t) =A(t) +λV(t),
J is called the Jacobi functional associated to the variation X. Then we have the following proposition:
Proposition 3.3.Let x:Mn →Mn+1 be a spacelike hypersurface in the time- oriented Lorentz manifoldMn+1, andNbe a globally defined unit normal vector field onMn. IfX:Mn×(−ε, ε)→Mn+1 is a variation ofx, then
(3.8) dJ(t) dt =
Z
M
[div(S1(∂X
∂t )>) + ∆f−(Ric(N, N) +kAk2−S12−λ)f]dMt. In particular, whenMn is closed andMn+1 has constant sectional curvature c, then
(3.9) dJ(t)
dt = Z
M
(2S2−cn+λ)f dMt.
Proof. We can get this result from Lemma 3.1 and Proposition 3.2. In fact, dJ(t)
dt = Z
M
dS1
dt dMt+ Z
M
S1(S1f+ div(∂X
∂t )>)dMt+ Z
M
λf dMt
= Z
M
[h(∂X
∂t )>,∇S1i+S1div(∂X
∂t )>+4f
−(Ric(N, N) +kAk2)f+S12f+λf]dMt
= Z
M
[div(S1(∂X
∂t )>) +4f−(Ric(N, N) +kAk2f−S21f−λ)f]dMt. WhenMn is closed and Mn+1 has constant sectional curvaturec, then we have
Z
M
div(S1(∂X
∂t )>)dMt= 0, Z
M
4f dMt= 0, and Ric(N, N) =nc, then using (2.9), we have (3.9).
Proposition 3.4. Let x : Mn → Mn+1 is a spacelike hypersurface in Lorentz space form Mn+1(c) with constant sectional curvaturec, and X : Mn×(−ε, ε)→ Mn+1 is a variation ofx, then
(3.10) dS2
dt = L1(f)−(S1S2−3S3)f−f(n−1)cS1+h(∂X
∂t )>,∇S2i.
In particular, ifS2 is a constant, then one has
(3.11) dS2
dt = L1(f)−(S1S2−3S3)f−f(n−1)cS1.
Proof. According to the proof of proposition 3.2 in [6], we can get dhkk
dt =fkk−cf−h2kkf+h∇hkk,(∂X
∂t )>i.
(3.12)
Using (2.9), we can get dS2
dt =S1dS1
dt −X
k
hkkdhkk
dt . (3.13)
Substituting (3.6) and (3.12) into (3.13), using (2.9) and (2.10), then we have dS2
dt = S1[4f−(Ric(N, N) +kAk2)f+h(∂X
∂t )>,∇S1i]
−X
k
hkk[fkk−cf−h2kkf+h∇hkk,(∂X
∂t )>i]
= S14f−S1(nc+S12−2S2)f+S1h(∂X
∂t )>,∇S1i+X
k
(S1fkk−hkkfkk)
−S1
X
k
fkk+cS1f +fX
k
λ3k−1
2h∇(S21−2S2),(∂X
∂t )>i
= L1(f)−(S1S2−3S3)f−f(n−1)cS1+h(∂X
∂t )>,∇S2i.
IfS2 is constant, then the last term in the above is equal to zero, so we have (3.11).
IfM has constant normalized scalar curvatureR, and we choose λ= 2S2−nc=n(n−1)(c−R)−nc,
(3.14)
thenλis a constant too, so we have
Proposition 3.5.Letx:Mn →Mn+1(c) is a spacelike hypersurface in the time- oriented Lorentz manifoldMn+1(c), andX:Mn×(−ε, ε)→Mn+1 is a variation of x, andS2 is constant, then
(3.15) d2J(0) dt2 (f) = 2
Z
M
[L1(f)−(S1S2−3S3)f−f(n−1)cS1]f dM.
Proof. Sinceλ= 2S2−nc=n(n−1)(c−R)−nc, using (3.9) and (3.11), we can get d2J(0)
dt2 (f) = 2 Z
M
dS2(0)
dt f dM = 2 Z
M
[L1(f)−(S1S2−3S3)f −f(n−1)cS1]f dM.
Definition 3.6.Supposex:Mn→Mn+1(c) has constant scalar curvature. The immersionxis stable if
d2J(0) dt2 (f) = 2
Z
M
[L1(f)−(S1S2−3S3)f−f(n−1)cS1]f dM ≤0, (3.16)
for all volume-presering variations ofx. IfMn is noncompact,xis stable if for every conpact submanifoldsM0⊂Mn with boundary, the restrictionx|M0 is stable.
For conformally stationary spacetimes, we have the following proposition.
Proposition 3.7.LetMn+1 be a conformally stationary Lorentz manifold, with conformal vectorV having conformal factorψ:Mn+1→R. Supposex:Mn→Mn+1 is a spacelike hypersurface inMn+1 =I×φFn with constant sectional curvature c, andNa future-pointing, unit normal vector field globally defined onMn,f =hV, Ni, then
(3.17) L1(f) = (S1S2−3S3)f+f(n−1)cS1−(n−1)S1N(ψ)−2S2ψ− hV>,∇S2i.
In particular, ifR is constant, thenS2 is a constant too, so
(3.18) 2f = L1(f) = (S1S2−3S3)f +f(n−1)cS1−(n−1)S1N(ψ)−2S2ψ.
Proof. We can choose{ek}as a moving frame on neighborhoodU ⊂M ofp, geodesic at p, and diagonalizing the shape operator A ofM at p, with Aek =λkek, for 1 ≤ k≤n. ExtendN andek (1≤k≤n) to a neighborhood of pin M, such that
hN, eki= 0 and (∇Nek)(p) = 0.
Let
V =X
l
αlel−f N, so we have
ek(f) =h∇ekN, Vi+hN,∇ekVi=−hAek, Vi+hN,∇ekVi.
Then
ekek(f) = −ekhAek, Vi+ekhN,∇ekVi
= −h∇ek(Aek), Vi −2hAek,∇ekVi+hN,∇ek∇ekVi.
(3.19)
For the first term in (3.19), we have
h∇ek(Aek), Vi = h∇ek(Aek),X
l
αlel−f Ni
= X
j
ek(hkj)hej,X
l
αleli+X
j
hkjh∇ekej,−f Ni
= X
l
αlek(hkl)−X
l
h2klf.
(3.20)
For the second term in (3.19), we have hAek,∇ekVi=X
j
hkjhej,∇ekVi=λkhek,∇ekVi=λkψ, (3.21)
where in the last equality we use the fact thatV is conformal vector having conformal factorψ, and we have
hN,∇ekVi+hek,∇NVi= 2ψhek, Ni= 0, then we can get
h∇ekN,∇ekVi+hN,∇ek∇ekVi+h∇ekek,∇NVi+hek,∇ek∇NVi= 0.
(3.22) Since
h∇ekN,∇ekVi=−hAek,∇ekVi=−λkψ, and
h∇NV,∇ekeki = −h∇NV,h∇ekek, NiNi
= −h∇NV, hkkNi=−hkkψhN, Ni=λkψ, then
hN,∇ek∇ekVi=−hek,∇ek∇NVi.
(3.23)
On the other hand, noting that
[N, ek](p) =−∇Nek(p)− ∇ekN(p) =−λkek(p), so we have
hR(N, ek)V, eki = h∇N∇ekV − ∇ek∇NV − ∇[N,ek]V, ekip
= Nh∇ekV, eki+hN,∇ek∇ekVi − h∇λkekV, eki
= hN,∇ek∇ekVi+N(ψ)−λkψ.
For the third term in (3.19), we have
hN,∇ek∇ekVi=hR(N, ek)V, eki −N(ψ) +λkψ.
(3.24) Also we have
hR(N, ek)V, eki = hR(N, ek)(X
l
αlel−f N), eki
= X
l
αlR(N, ek)el, eki −fhR(N, ek)N, eki, and
hR(N, ek)el, eki= R(ek, el, N, ek) =ek(hkl)−el(hkk), so (3.24) become
hN,∇ek∇ekVi=X
l
[αlek(hkl)−el(hkk)] +fR(N, ek, N, ek)−N(ψ) +λkψ.
(3.25)
Substituting (3.20), (3.21) and (3.25) into (3.19) we can get
fkk = ekek(f) =−X
l
αlek(hkl) +X
l
h2klf−2λkψ+ X
l
(αlek(hkl)−αlel(hkk)) +fR(N, ek, N, ek)−N(ψ) +λkψ
= (X
l
h2kl+ R(N, ek, N, ek))f− hV>,∇hkki −N(ψ)−λkψ.
(3.26) So we have
4f = (kAk2+ Ric(N, N))f− hV>,∇S1i −nN(ψ)−S1ψ, (3.27)
and X
k
λkfkk = −2X
k
λ2kψ−X
k,l
λkek(hkl) +X
k
λkh2klf+X
k
λkαl(ek(hkl)
−el(hkk)) +X
k
f λkR(N, ek, N, ek)−X
k
λkN(ψ) +X
k
λ2kψ
= −X
k
λ2kψ+X
k
λ3kf −X
k,l
αlλkel(λk) +f cS1−S1N(ψ).
(3.28)
Note that (2.9) and (2.10) X
k
λ2k =S12−2S2, X
k
λ3k =S31−3S1S2+ 3S3, substituting (3.27) and (3.28) into (2.13), so we can get
L1(f) = X
i,j
(S1δij−hij)fij=X
i
S1fii−X
i
hiifii
= (S1S2−3S3)f+f(n−1)cS1−(n−1)S1N(ψ)−2S2ψ− hV>,∇S2i.
4 Stable hyersurfaces with constant scalar curva- ture in GRW
In the following, we will consider the generalized Roberston-Walker spacesMn+1 =
−I×φFn, let
πI :Mn+1→I
denote the canonical projection onto theI. Then the vector field V = (φ◦πI)∂
∂t
is conformal, timelike and closed (in the sense that its metrically equavalent 1-form is closed), with conformal factorψ=φ0. Now we have the follow corollary
Corollary 4.1.IfMnis a closed spacelike hypersurface having constant normal- ized scalar curvature R in generalized Roberston-Walker spaces Mn+1 =−I×φFn with constant sectional curvature c. Let N be a future-pointing unit normal vector field globally defined onMn. If V = (φ◦πI)∂t∂ andf =hV, Ni, then
(4.1) L1(f) = (S1S2−3S3)f+f(n−1)cS1+ (n−1)S1φ00hN, ∂
∂ti −2S2ψ.
Proof. Since we have
(4.2) ∇φ0=−h∇φ0, ∂
∂ti∂
∂t =−φ00∂
∂t, then
(4.3) N(φ0) =hN,∇φ0i=−φ00h∂
∂t, Ni.
Substituting (4.3) into (3.18), we can get (4.1).
Now we can state and prove our main result:
Theorem 4.2.IfMna is closed hypersurface, having constant normalized scalar curvatureRin generalized Roberston-Walker spacesMn+1=−I×φFnwith constant sectional curvaturec. If the warping function φis not constant and satisfies Hφ00 ≥ max{(R−c)φ0,0}, and Mn is stable, then
(I)R=c onM, or
(II)M is spacelike sliceMt0=t0×Fn, for some t0∈I, satisfying Hφ00= (R−c)φ0.
Proof.Using Proposition 3.5 and Corollary 4.1, we can get J00(0)(f) = 2
Z
M
[(n−1)S1φ00hN, ∂
∂ti −2S2φ0]f dM
= 2 Z
M
[(n−1)S1φ00hN, ∂
∂ti −n(n−1)(c−R)φ0]φhN, ∂
∂tidM.
Let hN,∂t∂i = −coshθ, where θ denotes the hyperbolic angle between the timelike verctor fieldsN and ∂t∂. SinceMn is stable, so
0 ≥2 Z
M
[(n−1)S1φ00hN, ∂
∂ti −n(n−1)(c−R)φ0]φhN, ∂
∂tidM
≥2 Z
M
(n−1)S1φφ00coshθ(coshθ−1)dM ≥0, and hence
Hφ00(coshθ−1) = 0 and Hφ00= (R−c)φ0 (4.4)
holds onMn. IfR6=candφ06= 0 thenHφ006= 0, and coshθ= 1, soM is an umbilical
leaf satisfyingHφ00= (R−c)φ0. 2
Acknowledgements. The first author is supported by RFDP((20050141011), NCET(06-0276) and MXDUT073011.
References
[1] H. Alencar, M. do Carmo and A. G. Colares, Stable hyersurface with constant scalar curvature, Math. Z. 213(1993), 117-131.
[2] J. L. Barbosa and M. do Carmo, Stability of hyersurface with constant mean curvature, Math. Z. 185(1984), 339-353.
[3] J. L. Barbosa, M. do Carmo and J. Eeschenburg, Stability of hyersurface with constant mean curvature in Riemannian manifolds, Math. Z. 197(1988), 123-138.
[4] J. L. Barbosa and A. G. Colares, Stability of hyersurfaces with constant r-mean curvature, Ann. Global Anal. Geom., 15(1997), 277-297.
[5] J. L. Barbosa and V. Oliker,Stable hyersurface with constant mean curvature in Lorentz space, Geom. Global Analysis, Tohoku Universtiy, Sendai(1993), 161-164.
[6] A. Barros, A. Brasil and A. Caminha, Stability of spacelike hypersurfaces in foliated spacetimes, preprint.
[7] A. Brasil and A. G. Colares, Stability of spacelike hypersurface with constant r-mean curvature in de Sitter space, Proceedings of the XII Fall Workshop on Geometry and Physics, 139–145, Publ. R. Soc. Mat. Esp., 7, R. Soc. Mat. Esp., Madrid, 2004.
[8] A. Caminha,On the spacelke hyersurfaces of constant sectional curvature Lorentz manifolds, J. Geom. Phy. 56(2006), 1144-1174.
[9] X. Liu and J. Deng, Stable space-like hyersurfaces in the De Sitter space, Arch.
Math. 40(2004), 111-117.
[10] B. O’Nelill, Semi-Riemannian Geometry with Appications to Relativity, Aca- demic Press, New York, 1983.
[11] R. C. Reilly,Variational properties of functions of constant mean curvature hy- ersurface in space forms, J. Diff. Geom. 8(1973), 465-477.
[12] H. Rosenberg, Hpyersurfaces of constant curvature in space forms, Bull. Sc.
Math., 2e S´eri´e 117(1993), 211-239.
Authors’ address:
Ximin Liu, Biaogui Wang
Department of Applied Mathematics Dalian University of Technology Dalian 116024, P.R. China.
E-mail: [email protected], [email protected]