Horizon Cross-Section Topology

The robust higher-dimensional restriction on a black-hole cross-section is positive Yamabe type, not spherical topology. In four spacetime dimensions, orientability and the classification of closed surfaces turn that curvature condition into a two-sphere conclusion. In higher dimensions the same condition permits many nonspherical manifolds, including the topology of a black ring.

Required background. Exterior Algebra and Differential Forms supplies integration and orientation for the Gauss–Bonnet step; Hypersurface Normals and Tangential Projectors supplies the two nested hypersurface geometries and their normals.

Helpful background. Pseudo-Riemannian Metrics and Signature fixes causal normalization; Affine and Variational Geodesics supplies variation language; Causality and Global Structure distinguishes event and apparent horizons; Timelike Congruences and Spatial Projection supplies null expansion.

Marginally outer trapped cross-sections

Let $V^n$ be a smooth spacelike hypersurface in a smooth time-oriented spacetime $(M^{n+1},g)$, with $n\geq3$. Let $\Sigma^{n-1}\subset V$ be smooth, closed, connected, two-sided, and separating in the region under consideration. Write $U$ for the future unit normal to $V$, $N$ for the chosen outward unit normal to $\Sigma$ in $V$, and

$$K=U+N$$

for the outward future null normal. The outward null second fundamental form and expansion are

$$\chi(X,Y)=g(\nabla_XK,Y), \qquad \theta_+=\operatorname{tr}_\Sigma\chi.$$

A marginally outer trapped surface (MOTS) has $\theta_+=0$. It is stably outermost if the principal eigenvalue $\lambda_1$ of the outward MOTS stability operator is nonnegative, and strictly stably outermost if $\lambda_1>0$. An outer apparent horizon—an outermost MOTS with no homologous outer trapped surface outside it—has the stability needed for the theorem under the standard smoothness assumptions.

Before stating the result, fix every global qualification.

• The spacetime dimension is $n+1\geq4$; $\dim\Sigma=n-1$.
• Time orientation is assumed. Global spacetime orientability is not needed for the positive-Yamabe statement, but two-sidedness is needed to choose $N$. The four-dimensional spherical corollary additionally assumes $\Sigma$ orientable; this follows, for example, when $V$ is oriented and $\Sigma$ is two-sided.
• The theorem is local near $\Sigma$ and does not assume global hyperbolicity, stationarity, or asymptotic flatness. Such hypotheses enter only when a cross-section of an event horizon is identified with a stable outermost MOTS.
• Write Einstein’s equation as $G_{ab}=8\pi G\,T^{\rm eff}_{ab}$. With the site’s convention $G_{ab}+\Lambda g_{ab}=8\pi G T_{ab}$, this means $T^{\rm eff}_{ab}=T_{ab}-\Lambda g_{ab}/(8\pi G)$. The effective tensor is assumed to satisfy the dominant energy condition. No separate ANEC or null-completeness assumption appears.
• $\Sigma$ lies in the manifold interior. If $V$ has a remote boundary, restrict the differential calculation to a boundaryless tubular neighborhood of $\Sigma$; outermost and homology are still defined relative to the specified outer region. A surface meeting a boundary or corner requires a theorem with boundary terms and is not covered.

Positive Yamabe theorem

A closed manifold is of positive Yamabe type if it admits at least one Riemannian metric with positive scalar curvature. This is an existence property of its differentiable topology; it does not say that the metric induced on the physical horizon already has positive scalar curvature pointwise.

Stable-MOTS topology theorem. Under the assumptions above, a stably outermost MOTS $\Sigma^{n-1}$ is of positive Yamabe type unless all of the following exceptional conditions hold: its induced metric is Ricci flat, its outward null second fundamental form vanishes, and $T^{\rm eff}(U,K)=0$ along $\Sigma$. For a strictly stably outermost MOTS, the positive-Yamabe conclusion holds without this exception. The outer-apparent-horizon formulation and its stability extension are proved in Galloway–Schoen 2006, Theorem 2.1, proof, and concluding remark 1, pp. 3–6.

The exceptional alternative is rigid, not a generic second family. In the dimensions relevant to four- and five-dimensional spacetimes, Ricci flatness of the cross-section implies flatness. Strict stability eliminates it directly. A different refinement applies to time-symmetric initial data: if the surface is locally outer area minimizing and is not of positive Yamabe type, an outer neighborhood splits as a product. Excluding further homologous MOTSs outside the surface then rules out that product. Local outer area minimization by itself, without these additional hypotheses, is not the general rigidity theorem Galloway–Schoen 2006, concluding remarks 1–3, p. 6.

Proof architecture from stability to scalar curvature

The outward variation of $\Sigma$ by a normal speed $\phi$ changes its expansion through an elliptic operator

$$\left.\frac{\partial\theta_+}{\partial t}\right|_{t=0}=L\phi,$$

where

$$L\phi =-\Delta\phi+2\langle X,\nabla\phi\rangle +\bigl(Q+\operatorname{div}X-\lvert X\rvert^2\bigr)\phi$$

and, in the present normalization,

$$Q=\frac12R_\Sigma-8\pi G\,T^{\rm eff}(U,K)-\frac12\lvert\chi\rvert^2.$$

The vector $X$ records the tangential part of $\nabla_NU$. Although $L$ is generally not self-adjoint, its principal eigenvalue is real and has a strictly positive eigenfunction.

Outermost stability forces $\lambda_1\geq0$: if it were negative, the positive principal eigenfunction would deform $\Sigma$ outward to a surface with $\theta_+<0$, contradicting outermostness. Completing a square in the stability equation then yields, for every smooth test function $\psi$,

$$\int_\Sigma \left(\lvert\nabla\psi\rvert^2+Q\psi^2\right)d\mu_\Sigma \geq0.$$

The dominant energy condition gives $T^{\rm eff}(U,K)\geq0$ because $U$ and $K$ are future causal. Applying the preceding inequality to the first eigenfunction of the associated self-adjoint operator and conformally rescaling the induced metric produces a metric of nonnegative scalar curvature. If any of the nonnegative contributions is nonzero, a further conformal adjustment gives strictly positive scalar curvature. If they all vanish, one reaches precisely the Ricci-flat, shear-free, zero-flux exceptional case.

Notice the division of labor. Stability provides an elliptic inequality; the energy condition fixes the sign of the flux term; conformal geometry turns the inequality into a Yamabe statement. Neither topological censorship nor null geodesic completeness is used.

Why four dimensions give a sphere

In four spacetime dimensions, $\Sigma$ is a closed connected two-manifold. Assume it is orientable and the exceptional case is excluded. Positive Yamabe type supplies a metric with $R_\Sigma>0$. Since $R_\Sigma=2K_\Sigma$, Gauss–Bonnet gives

$$\int_\Sigma R_\Sigma\,d\mu =4\pi\chi(\Sigma)>0.$$

For a closed orientable surface of genus $g$,

$$\chi(\Sigma)=2-2g.$$

Thus $g=0$ and $\Sigma$ is diffeomorphic to $S^2$. This is the route from Hawking’s four-dimensional horizon-topology theorem to the sphere conclusion: energy and stability give positive Euler characteristic, while orientability and surface classification do the final topological work Hawking 1972, §§3–4, pp. 155–160.

If one assumes only non-strict stability and falls into the exceptional flat case, a torus has $\chi=0$ and is not excluded by Gauss–Bonnet alone. One must invoke the theorem’s rigidity refinements or an additional outermost condition; silently replacing “nonnegative” by “positive” loses this case.

Why higher dimensions do not

When $\dim\Sigma\geq3$, positive scalar curvature does not classify the manifold as a sphere. In five spacetime dimensions the three-manifold $S^2\times S^1$ admits a product metric of positive scalar curvature: the $S^2$ factor contributes positive curvature and the circle contributes zero. The rotating black-ring solution realizes exactly this horizon topology Emparan–Reall 2002, pp. 101101-1–4.

Spherical space forms, lens spaces, and certain connected sums can also be of positive Yamabe type. Conversely, index theory, minimal hypersurfaces, and enlargeability obstruct positive scalar curvature on many manifolds. The theorem supplies a powerful filter, not a complete higher-dimensional classification.

Negative cosmological constant illustrates another limit. If $\Lambda<0$ is moved to $T^{\rm eff}$, that effective vacuum term does not satisfy the dominant energy condition used above. Asymptotically locally AdS black holes can therefore have flat or higher-genus horizon sections without contradicting the theorem. One must not check the matter tensor alone while ignoring the cosmological term in the geometric inequality.

Event, apparent, and slicing distinctions

An event horizon is the boundary of a causal past of infinity; a MOTS is defined on a chosen spacelike slice. In a smooth stationary asymptotically flat black-hole exterior, suitable horizon sections satisfy the outermost marginal condition, so the theorem yields the familiar result. For a dynamical horizon, individual leaves can be stable MOTSs and obey the positive-Yamabe restriction.

An arbitrary cut of a nonsmooth event horizon need not be a stable MOTS. Crease sets, nontransverse slices, and surfaces with boundary lie outside the hypotheses. Nor does this theorem say that every apparent horizon is the event horizon. Topology, causal definition, and slicing dependence must be checked separately.

Exercises

Assume the strict stable-MOTS theorem. (a) Show that an orientable horizon cross-section in a four-dimensional spacetime must be $S^2$. (b) Explain why the same reasoning does not exclude $S^2\times S^1$ in a five-dimensional spacetime.

Solution

(a) Strict stability gives positive Yamabe type. For the two-dimensional $\Sigma$, choose a metric with $R_\Sigma>0$. Then Gauss–Bonnet gives $4\pi\chi(\Sigma)=\int_\Sigma R_\Sigma d\mu>0$, so $\chi>0$. A closed connected orientable surface has $\chi=2-2g$, hence $g=0$ and $\Sigma\simeq S^2$.

(b) A five-dimensional spacetime has a three-dimensional cross-section. There is no two-dimensional Gauss–Bonnet classification step. Give $S^2\times S^1$ the product of a round two-sphere and a circle. Its scalar curvature is the positive scalar curvature of the sphere plus zero from the circle, so it is of positive Yamabe type. It satisfies the theorem’s conclusion while remaining nonspherical.

References

• Emparan, Roberto, and Harvey S. Reall, “A Rotating Black Ring Solution in Five Dimensions”, Physical Review Letters 88, 101101, 2002. Open PDF.
• Galloway, Gregory J., and Richard Schoen, “A Generalization of Hawking’s Black Hole Topology Theorem to Higher Dimensions”, Communications in Mathematical Physics 266, pp. 571–576, 2006. Open PDF.
• Hawking, Stephen W., “Black Holes in General Relativity”, Communications in Mathematical Physics 25, pp. 152–166, 1972.