Δ+ = 3. The 𝜅= 1 harmonics give
Δ+ = 2,3,4. (5.71)
For𝜅= 2,3,4, . . ., we find
Δ+ =𝜅−1, 𝜅, 𝜅+ 1, 𝜅+ 2, 𝜅+ 3. (5.72) Hence if there is a CFT description that obeys the usual AdS/CFT rules then the𝑚= 0 gravitational perturbations give rise to five infinite families of operators with integer dimensions, just as for NHEK.5 This result hints that some symmetry is protecting the dimensions of operators dual to 𝑚 = 0 gravitational perturbations. Note that the operator of lowest dimension is marginal (in 1D): Δ+= 1.
5.4 Instabilities from near-horizon geometries
Does an instability of the near-horizon geometry imply the existence of an instability of the full spacetime? We conjectured in the introduction that this was the case for a particular class of perturbation modes and explained how extreme Kerr is consistent with the conjecture. In Section 5.3 we have shown that our conjecture predicts an instability for certain Myers-Perry black holes, and this prediction is confirmed by studies of perturbations of the full black hole geometry.
In this section, we will present some ideas that explain why our conjecture appears to work. In the case of a scalar field, we shall sketch a proof of the conjecture. We shall present some evidence suggesting that the method of proof in the scalar field case might also generalize to gravitational perturbations.6
5.4.1 Scalar field instabilities
Consider an uncharged, scalar field Ψ of mass 𝑀 in the extreme planar Reissner- Nordstr¨om-AdS black hole background in arbitrary dimension 𝑑 ≥ 4. This has a near- horizon geometry of the form𝐴𝑑𝑆2×ℝ𝑑−2. So, in the language described above, we have ℋ=ℝ𝑑−2 here.
5A massless scalar field would give operators with Δ+ =𝜅+ 1 for 𝜅= 0,1,2. . .. For 𝑁 >1, if we ignore the instability and calculate Δ+formally for gravitational perturbations (for stable modes) then the results are generically irrational.
6The material in this section provides motivation for much of the rest of work already described in this chapter. However, the results of this section were largely derived by my supervisor Harvey Reall, and appear in our paper [6].
As before, we can Fourier analyze on ℝ𝑑−2 to reduce the scalar field equation of motion to that of a massive scalar in𝐴𝑑𝑆2. The BF bound (5.31) associated to the𝐴𝑑𝑆2
is more restrictive than that associated to the asymptotic 𝐴𝑑𝑆𝑑 geometry. Numerical work [167, 177] suggests that if the scalar field violates the 𝐴𝑑𝑆2 BF bound then the scalar field is unstable in the full black hole geometry (even when the 𝐴𝑑𝑆𝑑 BF bound is respected). Moreover, it has been proved [175] that if the 𝐴𝑑𝑆2 BF bound is satisfied then the scalar field is stable in full black hole geometry, i.e., stability of the near-horizon geometry implies stability of the full black hole. Here we will prove that instability of the near-horizon geometry implies instability of the full black hole, in agreement with our conjecture.
Consider an extreme static black hole with geometry
𝑑𝑠2 =−𝑓(𝑟)𝑑𝑡2+𝑓(𝑟)−1𝑑𝑟2+𝑟2𝑑Σ2𝑘. (5.73) where𝑑Σ2𝑘 is the metric on a unit sphere if 𝑘= 1, a unit hyperboloid if 𝑘 =−1 and flat if 𝑘 = 0. This metric encompasses the Schwarzschild(-AdS) and Reissner-Nordstr¨om(- AdS) black holes with various horizon topologies.
As the black hole is extreme, we can assume that it has a degenerate horizon at 𝑟=𝑟+, and hence that
𝑓(𝑟) = (𝑟−𝑟+)2
𝐿2 +𝒪(𝑟−𝑟+)3. (5.74)
The near horizon geometry is then𝐴𝑑𝑆2×Σ𝑘 where the 𝐴𝑑𝑆2 has radius 𝐿.
In the full spacetime, the equation of motion of a scalar field Ψ of mass 𝑀 can be written
−∂2Ψ
∂𝑡2 =ℬΨ, (5.75)
where
ℬΦ≡𝑓 [
− 1 𝑟𝑑−2∂𝑟(
𝑟𝑑−2𝑓∂𝑟Ψ) + 1
𝑟2∇ˆ2Ψ +𝑀2Ψ ]
, (5.76)
with ˆ∇the connection on Σ𝑘. Now define the following inner product between functions defined on a surface of constant𝑡 outside the horizon:
(Ψ1,Ψ2) =
∫ ∞
𝑟+
𝑑𝑟 𝑑Σ𝑘𝑟𝑑−2𝑓−1Ψ1Ψ2. (5.77) We impose boundary conditions that the functions of interest must decay sufficiently fast for this integral to converge at 𝑟 = ∞, and they must vanish at least as fast as (𝑟−𝑟+) as 𝑟→𝑟+ in order that the integral converges at 𝑟=𝑟+. Now, if our functions decay fast enough at infinity, thenℬ is self-adjoint with respect to this inner product.7
7Note that this is different to the self-adjointness of operators discussed in Section 5.2.2; we are integrating over the exterior region of the full spacetime, not just over the manifoldℋ.
5.4. INSTABILITIES FROM NEAR-HORIZON GEOMETRIES 127 We can estimate the lowest eigenvalue𝜆0 ofℬusing the Rayleigh-Ritz method, noting that
𝜆0 ≤ (Ψ,ℬΨ)
(Ψ,Ψ) , (5.78)
for any function Ψ satisfying the boundary conditions.
Suppose that 𝜆0 is negative, with Ψ0 the associated eigenfunction. Then (5.75) has solutions
Ψ(𝑡, 𝑟, 𝑥) =𝑒±√−𝜆0𝑡Ψ0. (5.79) From the form ofℬ, it is easy to show that near𝑟 =𝑟+, the eigenfunction behaves as
Ψ0 ∼exp (
−
√−𝜆0𝐿2 𝑟−𝑟+
)
. (5.80)
Transforming to ingoing Eddington-Finkelstein coordinates (𝑑𝑣 = 𝑑𝑡 + 𝑑𝑟/𝑓 so 𝑡 ∼ 𝑣+𝐿2/(𝑟−𝑟+) near 𝑟=𝑟+) reveals that the solution𝑒+√−𝜆0𝑡Ψ0 is regular at the future horizon. This grows exponentially with time, and hence represents an instability of the scalar field in the black hole background.
The idea now is to show that violation of the 𝐴𝑑𝑆2 BF bound (5.31) implies the existence of a trial function Ψ with (Ψ,ℬΨ) <0. This implies that 𝜆0 must be negative, hence the scalar field is unstable and the conjecture is proved.
To see how this works, consider the case of a 4D extreme Reissner-Nordstr¨om-AdS black hole, for which
𝑓(𝑟) = ( 1− 𝑟+
𝑟 )2(
𝑘+ 3𝑟+2 + 2𝑟𝑟++𝑟2 ℓ2
)
, (5.81)
whereℓ is the 𝐴𝑑𝑆4 radius. This has a near-horizon geometry with 1
𝐿2 = 6 ℓ2 + 𝑘
𝑟2+. (5.82)
Consider the following trial function (motivated by a similar example in Ref. [175]) Ψ(𝑟) = (𝑟−𝑟+)ℓ9/2
𝑟4(𝑟−𝑟++𝜖ℓ)3/2, (5.83) with𝜖 >0. This satisfies the boundary conditions required for self-adjointness of ℬ. As 𝜖→0, this gives
(Ψ,ℬΨ) ≡
∫ ∞
𝑟+
𝑑𝑟 𝑑Σ𝑘𝑟2(
𝑓(∂𝑟Ψ)2+𝜇2Ψ2)
=𝑉𝑘 (
𝑀2+ 1 4𝐿2
) ℓ9 𝑟6+log(
𝜖−1)
+. . . (5.84) where the ellipsis denotes terms subleading in𝜖, and 𝑉𝑘 is the volume of Σ𝑘. The 𝐴𝑑𝑆2 BF bound states that the quantity in brackets on the RHS should be non-negative.8
8More precisely: this is the BF bound for modes which are homogeneous on Σ𝑘.
From the above expression we see that ℬ admits a negative eigenvalue if this bound is violated. Hence there is an instability of the scalar field when the 𝐴𝑑𝑆2 BF bound is violated. The argument generalizes easily to 𝑑 >4.9
A similar example is the cohomogeneity-1 Myers-Perry-AdS black hole discussed in Section 5.3. For large black holes, the𝐴𝑑𝑆2 BF bound is more restrictive than that of 𝐴𝑑𝑆𝑑. Hence a scalar field can violate the 𝐴𝑑𝑆2 BF bound but respect the 𝐴𝑑𝑆𝑑 BF bound. Ref. [175] studied the case of a scalar field invariant under∂/∂𝜓(i.e. those modes corresponding to𝑚 = 0 in Section 5.3.2) and presented numerical evidence that such a scalar field is indeed unstable if its mass lies between the two BF bounds. Furthermore, it was proved that the scalar field (with𝑚= 0) is stable if it respects both bounds.
This example also can be understood using the argument above. Even though the black hole is rotating, the fact that the scalar field is invariant under ∂/∂𝜓 implies that its equation of motion takes the form (5.75). The only difference is the form ofℬ:
ℬΨ = 𝑉(𝑟) ℎ(𝑟)2
[
− 1 𝑟𝑑−2
∂
∂𝑟 (
𝑟𝑑−2𝑉(𝑟)∂Ψ
∂𝑟 )
− 1
𝑟2∇ˆ2Ψ +𝜇2Ψ ]
, (5.85)
where ˆ∇ is the connection of the metric on ℂℙ𝑁. ℬ is self-adjoint with respect to the scalar product
(Ψ1,Ψ2) = 2𝜋
∫ ∞
𝑟+
𝑑𝑟 𝑑ˆΣ𝑁𝑟𝑑−2ℎ(𝑟)2
𝑉(𝑟)Ψ1Ψ2, (5.86) where𝑑ˆΣ𝑁 is the volume element onℂℙ𝑁. Consider, for simplicity, the five-dimensional case (where𝑁 = 1). We can use the trial function (5.83) with the modification𝑟4 7→𝑟5 (to improve the convergence at𝑟 =∞). The result is the same: (Ψ,ℬΨ) is proportional to log(𝜖−1) with a coefficient of proportionality that is negative if, and only if, the 𝐴𝑑𝑆2
BF bound is violated. Hence we have proved that the scalar field is unstable in the extreme black hole geometry if it violates the 𝐴𝑑𝑆2 BF bound, in agreement with our conjecture.
Now recall from the introduction to this chapter that for the extreme Kerr black hole, we know that instability of the near-horizon geometry does not always imply instability of the full black hole. Even for a scalar field, there exist modes that violate the effective BF bound in the near-horizon geometry [87]. So how does the above argument fail for Kerr? The key step above was to impose a symmetry condition on the scalar field that makes its equation of motion take the form (5.75), in which first time derivatives are
9Ref. [175] proved thatstabilityof the near-horizon geometry implies stability of the full black hole for𝑘=−1,0. Combining this with our result, we learn that, for these cases, the full black hole is stable if, and only if, its near-horizon geometry is stable. For𝑘 = 1, stability of the near-horizon geometry is not sufficient to guarantee stability of the full black hole because the𝐴𝑑𝑆2 BF bound can be less restrictive than the𝐴𝑑𝑆𝑑 bound.
5.4. INSTABILITIES FROM NEAR-HORIZON GEOMETRIES 129 absent. For Kerr, eliminating first time derivatives implies that the scalar field must be axisymmetric, and axisymmetric modes do respect our conjecture.
More generally, if we consider an extreme black hole with metric (5.1) then the necessary and sufficient condition for the equation of motion of a massive scalar field to reduce to (5.75) is
𝑁𝐼(𝑥) ∂
∂𝜙𝐼Ψ = 0. (5.87)
if we Fourier analyze Ψ ∝ 𝑒𝑖𝑚𝐼𝜙𝐼 for integers 𝑚𝐼 then this equation reduces the ax- isymmetry condition (5.2) in the conjecture that we made in the introduction. If this condition is satisfied then the argument we have sketched above should apply. This explains why our conjecture should work for scalar fields.
5.4.2 Gravitational perturbations
We have sketched an argument that explains why a scalar field instability in the near- horizon geometry of an extreme black hole implies an instability of the full black hole, provided the scalar field satisfies the symmetry condition (5.2). We are really interested in linearized gravitational perturbations. If we attempt to repeat the same argument, we would need to convert the equations governing gravitational perturbations to something of the form
−∂2Ψ𝛼
∂𝑡2 =𝒜𝛼𝛽Ψ𝛽 (5.88)
with Ψ𝛼 a vector encoding the perturbation, and 𝒜𝛼𝛽 a matrix operator self-adjoint with respect to a suitable inner product. Can this be done? For axisymmetric (i.e.
𝑚 = 0) metric perturbations of the Kerr black hole, in a certain gauge, it can indeed:
a variational formula analogous to (5.78) is given in Chandrasekhar [178, §114]. Hence the extreme Kerr black hole should obey our conjecture and, as we discussed in the introduction, it does.
What about higher dimensions? Can we bring the equations governing gravitational perturbations of, for example, a Myers-Perry black hole to the form (5.88), provided the perturbation satisfies the symmetry condition (5.2)? Evidence that this is indeed possible comes from recent work [109] on instabilities of cohomogeneity-1 MP black holes.
This work considered metric perturbations satisfying (5.2). In the cases for which an instability was found, the time dependence was 𝑒−𝑖𝜔𝑡 where 𝜔 has positive imaginary part. In general, one would expect 𝜔 to be complex but it turned out that unstable modes had purely imaginary𝜔. This would be explained if perturbations were governed by an equation of the form (5.88) (with𝒜𝛼𝛽 self-adjoint), which predicts that𝜔2 should be real.
Perturbations of Myers-Perry black holes with a single non-vanishing angular mo- mentum have also been considered [107]. Again, perturbations satisfying (5.2) were considered ((5.2) reduces to 𝑚1 = 0 in this case). The critical mode associated to the onset of instability was identified. This mode has zero frequency, which suggests that unstable modes should have purely imaginary frequency (if unstable modes had 𝜔 with a non-zero real part then there is no reason why the mode at the threshold of instability should have𝜔 = 0 rather than𝜔 some non-zero real number). Again, this suggests that a formula of the form (5.88) exists for this situation.
In these two examples, it appears that the condition (5.2) is indeed sufficient to obtain an equation of the form (5.88) governing gravitational perturbations (in a certain gauge). This is encouraging evidence that it will indeed be possible to demonstrate that an instability of the near-horizon geometry of an extreme black hole will imply instability of the full black hole provided this symmetry condition is respected.
Chapter 6
Hidden symmetries of black rings
6.1 Introduction
The final chapter of this thesis starts out along a slightly different track. Rather than deriving general results about Einstein spacetimes in higher dimensions as we have done in the previous chapters, we will discuss the properties of a particular spacetime in five dimensions, namely the doubly-spinning black ring.
This doubly-spinning black ring is an asymptotically flat solution to the vacuum field equations, discovered by Pomeransky and Sen’kov [70] using solution generating techniques for higher-dimensional Weyl solutions [75]. It is a generalisation of the original Emparan-Reall black ring [68] with rotation around the𝑆2as well as the𝑆1. The solution is rather more complicated than [68], but reduces to the balanced version of that solution in a particular limit.
Various authors have studied properties of this solution in the past. Kunduri et al.
[88] studied the extremal limit and associated near-horizon geometry, while Elvang &
Rodriguez [74] studied its phase structure, asymptotics and horizon. There is a more general version of the solution, corresponding to an ‘unbalanced’ ring with conical sin- gularities, which is explicitly presented in [179]. Much of the current literature on this spacetime is reviewed in [16], we give some brief details of its properties in Section 6.2.
More recently, and after the work of this chapter was complete, Chruscielet al.[180]
produced an extensive rigorous study of various properties; and in particular constructed an analytic extension of the spacetime through its event horizon, as well as explicitly exhibiting the regularity of the metric on both the ergosurface and the axes of rotation.
In later work, Chrusciel & Szybka [181] proved stable causality of the domain of outer communications.
As the black ring is rotating, there is an ergosurface. The topology of this ergosurface changes as the black ring parameters vary. For a ring with sufficiently small rotation
131
about the𝑆2, the topology is 𝑆1×𝑆2, as in the singly spinning case. However, for more rapid𝑆2 rotation the ergosurface has topology𝑆3∪𝑆3, consisting of a small sphere around the centre of the ring, and a large sphere enclosing the entire ring. There is a critical case, where a topology change occurs: the surface ‘pinches’ on an𝑆1 (see Section 6.2.6).
After the work of this section was mostly complete, a similar analysis [182] appeared, as part of a paper discussing the properties of ergoregions in various higher-dimensional solutions.
At face value, the doubly-spinning black ring metric seems to be extremely com- plicated. However, we will see that it admits some expected symmetry that makes it possible to study certain properties analytically. Consider the Hamilton-Jacobi (HJ) equation, describing geodesic motion in the spacetime. We will see in Section 6.3 that, for both the singly-spinning and doubly-spinning rings, this equation admits separable solutions in the case of null, zero energy geodesics. These null, zero energy geodesics can only exist inside the ergoregion, and correspond to massless particles coming out of the white hole horizon in the past, and falling into the black hole horizon in finite parame- ter time in the future. On the other hand, the Klein-Gordon equation is not separable in ring-like coordinates, even if we restrict to looking for massless, time-independent solutions. We will briefly discuss the reasons for this in Section 6.5.4.
In Section 6.4, we will see that the existence of these geodesics allows us to construct new coordinate systems for the black ring that are valid across the event horizon. In the singly spinning case, it is possible to construct a new set of coordinates (𝑣, 𝑥, 𝑦,𝜙,˜ 𝜓˜) such that𝑣, ˜𝜙, ˜𝜓are constant along one of these geodesics. These coordinate systems are regular at the future black hole horizon, and a particular subset of them cover the entire horizon. The coordinate change given in [69] is included in this family of coordinate systems, and hence this allows us to understand its geometric significance.
In the doubly-spinning case, the best approach is to use coordinates (𝑣, 𝑥, 𝑦,𝜙,˜ 𝜓˜) where only ˜𝜙 and ˜𝜓 are constant along the geodesics, and a change of coordinate 𝑣 is made that simply makes the metric regular at the horizon (rather than demanding that it is constant along the geodesics). Using this approach, we are able to present explicitly a form for the doubly-spinning metric that is valid across the horizon.
We then see in Section 6.5 that the null, zero energy separability of the Hamilton- Jacobi equation is related to the existence of a conformal Killing tensor in a 4-dimensional spacetime obtained by a spacelike Kaluza-Klein reduction of the black ring spacetime in the ergoregion, reducing along the asymptotically timelike Killing vector field. Further- more, a pair of conformal Killing-Yano tensors exist for the 4-dimensional spacetime if, and only if, the associated ring is singly-spinning.