• Aucun résultat trouvé

Slowly rotating Kerr–de Sitter black holes

3. The Kerr–de Sitter family of black hole spacetimes

3.2. Slowly rotating Kerr–de Sitter black holes

Given Schwarzschild–de Sitter parameters b0=(M,0,0)∈B and the smooth manifold (3.9) with time functiont∈C(M), we now proceed to define the Kerr–de Sitter fam-ily of metrics, depending on the parameters b∈B, as a smooth family gb of stationary Lorentzian metrics onMforb close tob0.

Given the angular momentuma=|a|of a black hole of massM(spinning around the axisa/|a|∈R3fora6=0), the Kerr–de Sitter metric with parametersb=(M,a) inBoyer–

Lindquist coordinates(t, r, φ, θ)∈R×Ib×S1φ×(0, π), withIb⊂Ran interval defined below, takes the form

gb=−%2b dr2

˜ µb +dθ2

b

bsin2θ

(1+λb)2%2b(a dt−(r2+a2)dφ)2 + µ˜b

(1+λb)2%2b(dt−a(sin2θ)dφ)2,

(3.12)

where

˜

µb(r) = (r2+a2) 1−13Λr2

−2Mr, %2b=r2+a2cos2θ, λb=13Λa2, b= 1+λbcos2θ.

(3.13) For a=0, we have ˜µb(r)=r2µb(r), with µb defined in (3.4). For a6=0, the spherical coordinates (φ, θ) are chosen such that the vector a/|a|∈S2 is defined by θ=0, and the vector field∂φ generates counterclockwise rotation around a/|a|, with R3 carrying the standard orientation. Thus, for a=(0,0, a), a>0, the coordinates (φ, θ) are the standard spherical coordinates ofS2,!R3. We note that, using these standard spherical coordinates, the expression forg(M,(0,0,−a))is given by (3.12) withareplaced by−a: this simply means that reflecting the angular momentum vector across the origin is equivalent to reversing the direction of rotation.

Lemma 3.3. Let rb0,<rb0,+denote the unique positive roots of µ˜b0. Then,for b in an open neighborhood b0∈UB⊂B, the largest two positive roots

rb,< rb,+

of µ˜b depend smoothly on b∈UB. In particular, for b near b0,we have

|rb,±−rb0,±|< εM, and so rb,±∈I, with I defined in (3.9).

Proof. This follows from the simplicity of the roots rb0,± of ˜µb0 and the implicit function theorem.

The interval Ib in which the radial variable r of the Boyer–Lindquist coordinate system takes values is then

Ib= (rb,, rb,+).

The coordinate singularity of (3.12) is removed by a change of variables

t=t−Fb(r), φ=φ−Φb(r), (3.14) whereFbb are smooth functions on (rb,, rb,+) such that

Fb0(r) =±

(1+λb)(r2+a2)

˜ µb

+cb,±

, Φ0b(r) =±

(1+λb)a

˜ µb

+˜cb,±

(3.15) nearr=rb,±, withcb,± and ˜cb,± smooth up torb,±. Here, forb=b0, we take

cb0,∈ C((rI,, rb0,+)) and cb0,+∈ C((rb0,, rI,+)),

withrI,± defined in (3.9), to be equal to any fixed choice for the Schwarzschild–de Sitter space (M, gb0), e.g. the one in Lemma3.1.

Lemma 3.4. Fix radii r1 and r2 with rb0,M<r1<r2<rb0,+−εM. Then, there exists a neighborhood b0∈UB⊂B such that the following holds:

(1) There exist smooth functions

UB×(rI,, r2)3(b, r)7−!cb,(r), UB×(r1, rI,+)3(b, r)7−!cb,+(r),

which are equal to the given cb0,± for b=b0,such that the two functions

±

(1+λb)(r2+a2)

˜ µb

+cb,±

agree on (r1, r2).

(2) There exist functions

UB×(rI,, r2)3(b, r)7−!c˜b,(r), UB×(r1, rI,+)3(b, r)7−!c˜b,+(r), with a−1b,± smooth, and c˜b0,±≡0,such that the two functions

±

(1+λb)a

˜

µb +˜cb,±

agree on (r1, r2).

Proof. We can takecb,≡cb0, on (rI,, r2); then, for a cutoffχ∈C(R), withχ≡1 on [r1, r2] andχ≡0 on [rb0,+−εM, rI,+), we put

cb,+=−

2(1+λb)(r2+a2)

˜

µb +cb0,

χ+cb0,+(1−χ).

A completely analogous construction works for ˜cb,±: we can take ˜cb,≡0 and put

˜

cb,+=−2(1+λb)a

˜ µb

χ.

Clearly, the functionsa−1˜cb,± depend smoothly onb.

This lemma ensures that the definitions on Fb0 and Φ0b in the two regions in (3.15) coincide, hence makingFb and Φb well-defined up to an additive constant. Using

aFb0−(r2+a20b=±(acb,±−(r2+a2)˜cb,±)

and

Fb0−a(sin2θ)Φ0b

(1+λb)%2b

˜ µb

+cb,±−a(sin2θ)˜cb,±

forr>r1(+ sign) or r<r2 (−sign), one now computes gb=− bsin2θ

(1+λb)2%2b(a(dt±cb,±dr)−(r2+a2)(dφ±˜cb,±dr))2 + µ˜b

(1+λb)%2b(dt±cb,±dr−a(sin2θ)(dφ±˜cb,±dr))2

± 2 1+λb

(dt±cb,±dr−a(sin2θ)(dφ±˜cb,±dr))dr−%2b b

2,

(3.16)

which now extends smoothly to (and across)rb,±. Since one can compute the volume form to be(4)|dgb|=(1+λb)−2%2b(sinθ)dtdr dφdθ, the metric gb in the coordinates used in (3.16) is a non-degenerate Lorentzian metric, apart from the singularity of the spherical coordinates at θ=0, π, which we proceed to discuss: first, we compute the dual metric to be

%2bGb=−˜µb(∂r∓cb,±t∓˜cb,±φ)2

±2a(1+λb)(∂r∓cb,±t∓˜cb,±φ)∂φ

±2(1+λb)(r2+a2)(∂r∓cb,±t∓˜cb,±φ)∂t

−(1+λb)2

bsin2θ(a(sin2θ)∂t+∂φ)2bθ2.

(3.17)

Smooth coordinates on S2 near the polesθ=0, π are x=sinθcosφ and y=sinθsinφ, and for the change of variables ζ dφ+η dθ=λ dx+ν dy, one finds sin2θ=x2+y2 and ζ=νx−λy, that is,

φ=y∂x−x∂y,

and thus the smoothness of%2bGbnear the poles follows from writing−btimes the term coming from the last line of (3.17) as

(1+λb)2

sin2θ ∂φ2+2bθ2= (1+λb)2((sin−2θ)∂2φ+∂θ2)+(b2−(1+λb)2)∂2θ. Indeed, the first summand is smooth at the poles, since (sin−2θ)∂2φ

+∂θ2=G/ is the dual metric of the round metric onS2in spherical coordinates; and we can rewrite the second summand as

−(2+λb(1+cos2θ))λb(sin2θ)∂θ2 (3.18)

(4) For these calculations, a convenient frame ofT Mis

v1=r∓cb,±t∓˜cb,±φ, v2=a(sin2θ)∂t+∂φ, v3=t and v4=θ.

and observe that (sin2θ)∂θ2=(1−x2−y2)(x∂x+y∂y)2is smooth at (x, y)=0 as well. Since the volume form is given by

|dgb|= (1+λb)−2(r2+a2(1−x2−y2))(1−x2−y2)−1/2dtdr dx dy,

and thus smooth at the poles, we conclude that gb indeed extends smoothly and non-degenerately to the poles.

Using the map

(t, r, φ, θ)7−!(t, rsinθcosφ, rsinθsinφ, rcosθ)∈Rt×X⊂M,

withX⊂R3as in (3.10), we can thus push the metricgbforward to a smooth, stationary, non-degenerate Lorentzian metric, which we continue to denote bygb, onM, andgb0 is equal (pointwise!) to the extended Schwarzschild–de Sitter metric defined in§3.1. The Boyer–Lindquist coordinate patch of the Kerr–de Sitter black hole with parametersb∈B is the subset{rb,<r<rb,+}⊂M.

Since the choice of spherical coordinates does not depend smoothly onanear a=0, the smooth dependence ofgb, as a family of metrics on M, on b is not automatic; we thus prove the following result.

Proposition 3.5. Let the neighborhood UB⊂B of b0 be as in Lemma 3.3. Then, the smooth Lorentzian metric gb on M depends smoothly on b∈B.

Here, the smoothness of the familygb is equivalent to the statement that the map B×Rt×X3(b, t, p)7−!gb(t, p),

with gb(t, p) on the right the matrix of gb at the given point in the global coordinate system (t, p)∈M, is a smooth mapUB×Rt×X!R4×4.

Proof. Given (M,a)∈B witha=|a|6=0, let us denote the spherical coordinate sys-tem onXwith north poleθ=0 ata/|a|by (φb,∗, θb), so the push-forwards of the functions in (3.13) toM are simply obtained by replacing θ byθb. Then, if (r, φb,∗, θb) are the polar coordinates of a pointp∈X, we have

r=|p|, acosθb=

a, p

|p|

, a2sin2θb=|a|2−a2cos2θb,

where| · |andh ·,· idenote the Euclidean norm and inner product onX⊂R3, respectively.

Since 0∈X, this shows that/ r,acosθb anda2sin2θb, and hence the push-forwards of ˜µb,

%bbandbare smooth (inb) families of smooth functions onM, as arecb,±anda−1˜cb,±

(which only depend onr), on their respective domains of definition, by Lemma3.4.

We can now prove the smooth dependence of the dual metric Gb onb: in light of the expression (3.17) and the discussion around (3.18), all we need to show is that the vector fields∂t, ∂r, a∂φb,∗ and a(sinθb)∂θb depend smoothly on a, in particular near a where they are defined to be identically zero. (Note that the 2-tensor in (3.18) is a smooth multiple of a2(sin2θb)∂θ2

b.) Indeed, this proves that Gb is a smooth family of smooth sections of S2T M, and we already checked the non-degeneracy of Gb at the poles, where the spherical coordinates are singular.

For ∂t and for the radial vector field ∂r=|p|−1p∂p, which do not depend onb, the smoothness is clear. Further, we have

a∂φb,∗=∇a×p at p∈X,

i.e. differentiation in the direction of the vectora×p. Indeed, ifa=a~e3:=(0,0, a), both sides equala(x∂y−y∂x) onR3x,y,z, and ifa∈R3is any given vector andR∈SO(3) is a ro-tation withRa=a~e3,a=|a|, then (R(a∂φb,∗))|p=(a∂φa~e3,∗)|R(p), which we just observed to be equal to∇a~e3×R(p)=Ra×p.

In a similar vein, one sees that

a(sinθb)∂θb=|p|−1p×(p×a) atp∈X,

and the latter expression is clearly smooth ina, finishing the proof of the proposition.

Remark 3.6. If one were interested in analyzing the non-linear stability of Kerr–

de Sitter spacetimes for general parameters—i.e. dropping the assumption of small an-gular momentum—one notes that the construction described in this section can be per-formed in the neighborhood of any Kerr–de Sitter spacetime which is non-degenerate in the sense that the two largest roots of ˜µb are simple and positive.

Given the smooth family of Kerr–de Sitter metricsgbonMdefined in the previous section, we can define its linearization around anygb,b∈UB.

Definition 3.7. Forb∈UBandb0∈TbB, we define the elementgb0(b0) of the linearized Kerr–de Sitter family, linearized aroundgb, by

g0b(b0) = d dsgb+sb0

s=0

.

Notice here that UB⊂B⊂R4 is an open subset of R4, and thus we can identify TbB=R4. The linearization of the Kerr–de Sitter family aroundgb is the 4-dimensional vector spaceg0b(TbB)≡{g0b(b0):b0∈TbB}.

Remark 3.8. Define db to be the number of parameters needed to describe a lin-earized Kerr–de Sitter metric modulo Lie derivatives (i.e. the number of ‘physical degrees of freedom’); that is,

Γb= g0b(TbB) (ranδg

b)∩g0b(TbB), db= dim Γb. Then one can show that

db=

4, ifa=0, 2, ifa6=0.

See also the related discussion at the end of [37, §6.2.2]. The reason for db=4 for Schwarzschild–de Sitter parametersbis that slowly rotating Kerr–de Sitter metrics with rotation axes which are far apart are related by a rotation by a large angle. This is one of the reasons to use the redundant (for non-zero angular momenta) parametrization (3.1) of the Kerr–de Sitter family.