Spacetime Topology and Topological Censorship

Spacetime topology enters general relativity through several mathematically different questions. Global hyperbolicity asks whether the entire spacetime is a product evolution of one Cauchy hypersurface. Topological censorship asks whether causal observers communicating with infinity can probe nontrivial loops in the exterior. Horizon-topology theorems constrain the intrinsic geometry of selected marginally trapped cross-sections. Lorentzian-cobordism results ask whether two spatial topologies can be joined without losing chronology, causal compactness, metric nondegeneracy, or completeness. This chapter separates those questions before showing how causal theory, covering spaces, focusing, and elliptic stability connect them.

Required background. Exterior Algebra and Differential Forms supplies orientation, pullbacks, and integration; Hypersurface Normals and Tangential Projectors supplies induced metrics, boundary normals, and codimension-two geometry.

Helpful background. Pseudo-Riemannian Metrics and Signature supplies causal cones and conformal rescaling; Affine and Variational Geodesics supplies null generators and focal points; Causality and Global Structure supplies global hyperbolicity and causal boundaries; Timelike Congruences and Spatial Projection prepares the focusing arguments.

Enter this chapter

Unless a page states otherwise, spacetimes are smooth, connected, time-oriented Lorentzian manifolds with the global signature $(-,+,\ldots,+)$. A physical manifold boundary is absent. When a conformal completion is used, its null or timelike conformal boundary is explicitly part of the completed causal space. When a theorem permits a timelike physical boundary, the page invokes a result formulated for manifolds with timelike boundary; it does not silently apply a boundaryless theorem. Dimension, orientability, global hyperbolicity, asymptotics, and energy or convergence assumptions are restated at the result that needs them.

The most important distinction is between existing topology and operationally accessible topology. A globally hyperbolic spacetime may have a toroidal Cauchy hypersurface and retain that topology forever. Nothing in global hyperbolicity makes the slice simply connected. Topological censorship adds an asymptotic region and a focusing condition, then constrains curves that leave and return to that asymptotic region. It can make the domain of outer communications simply connected under sufficiently strong asymptotically flat hypotheses, while leaving complicated topology behind an event horizon.

A second distinction is between the topology of an exterior and the topology of a horizon cut. A covering-space argument controls the fundamental group of a domain. A stable-marginally-trapped-surface argument controls whether a closed cross-section admits a metric of positive scalar curvature. These conclusions are related in examples but neither implies the other. In particular, positive Yamabe type means “admits some positive-scalar-curvature metric”; it does not mean “is a sphere.”

The chapter sequence

The six pages form one argument with several carefully marked changes of hypotheses.

1. Topology of Cauchy Surfaces proves that a smooth globally hyperbolic spacetime without boundary splits as $\mathbb R\times\Sigma$. All smooth spacelike Cauchy hypersurfaces are diffeomorphic, so spatial topology is constant within that Cauchy evolution. A separate boundary-capable theorem gives $\mathbb R\times\overline\Sigma$ when the spacetime has timelike boundary.
2. Covering Spaces and Causal Topology explains the method behind censorship. A fixed-endpoint homotopy class becomes the endpoint sheet of a lifted path. If a causal curve detects a new loop, its lift communicates between distinct copies of infinity; global causality and focusing are then used to rule out that communication.
3. The Topological Censorship Theorem states a later asymptotically flat theorem with the completed-region and outward, decreasing-$\Omega$ normal-confinement hypotheses needed by its proof. It also separates the 1995 Friedman–Schleich–Witt erratum about passive censorship from Krasnikov’s later observation of a gap in the original active-censorship proof.
4. Horizon Cross-Section Topology derives a positive-Yamabe restriction from MOTS stability and the effective dominant energy condition. It then explains why an orientable cross-section is $S^2$ in four spacetime dimensions but can be nonspherical in higher dimensions.
5. Cobordism and Topology Change distinguishes a topological cobordism from a causally acceptable Lorentzian evolution. Under causal compactness, chronology forces a product. A stronger dynamical version drops chronology and instead assumes the null generic condition plus a directed past-half-integral null-convergence condition; causal compactness supplies completeness of the relevant imprisoned horizon generator rather than assuming global null completeness.
6. Scope and Limits of Topological Censorship compares asymptotically flat, timelike-boundary, traversable-throat, higher-dimensional, and topology-changing settings. It identifies the failed hypothesis rather than labeling every nonspherical or multiply connected geometry a counterexample.

The order matters. The covering construction is easier to understand after the product theorem, and the censorship theorem is easier to interpret after the covering construction. Horizon topology and cobordism then show two genuinely different ways in which causal and curvature assumptions constrain topology.

Four theorem packages

No single phrase such as “reasonable spacetime” supplies the hypotheses below. Each package has its own boundary and energy assumptions.

Result	Geometric and causal setting	Curvature or energy input	Topological conclusion
Smooth Cauchy splitting	smooth connected time-oriented globally hyperbolic spacetime without boundary	none	$M\simeq\mathbb R\times\Sigma$; all smooth spacelike Cauchy hypersurfaces are diffeomorphic
Asymptotically flat censorship	dimension $n\geq3$; null conformal completion; global hyperbolicity of the physical and specified completed future regions; simply connected asymptotic neighborhood; compact cuts and confinement of outward normals with $d\Omega/ds\lt0$	pointwise null convergence, or the theorem’s directed future-half-geodesic averaged condition	the domain of outer communications is simply connected
Timelike-boundary censorship	globally hyperbolic completed exterior with one timelike boundary component and a compact spacelike boundary cut	nonnegative Ricci integral along the relevant future-complete null half-geodesics	inclusion of the boundary induces a surjection on fundamental groups
Stable-MOTS topology	spacetime dimension at least four; closed connected two-sided stably outermost interior MOTS; orientability only for the four-dimensional surface-classification corollary	effective dominant energy condition; no ANEC or null-completeness hypothesis	positive Yamabe type except for the stated rigid case; in four dimensions, an orientable strict case is $S^2$

The first theorem is kinematical: it uses global hyperbolicity but no Einstein equation. The second and third are causal-focusing theorems: they require asymptotic control, an appropriate global causal compactness statement, and null convergence. The fourth is elliptic: stability of a MOTS produces an inequality whose energy term has the right sign. It neither assumes conformal infinity nor proves simple connectivity of the exterior.

The smooth Cauchy splitting is given by Bernal–Sánchez 2005, Theorem 1.1 and proof, pp. 44–49. For a spacetime with timelike boundary, the corresponding smooth product and boundary product are established directly in Aké Hau–Flores–Sánchez 2021, Theorem 1.1 and §5.2, pp. 47–49 and 77–79. These are separate statements because doubling a Lorentzian manifold across an arbitrary timelike boundary need not preserve the smooth causal structure required by a theorem.

The covering-space contradiction

The central censorship proof can be remembered as a five-link chain:

$$\text{nontrivial exterior loop} \Longrightarrow \text{different lifted infinities} \Longrightarrow \text{achronal null generator} \Longrightarrow \text{focal point} \Longrightarrow \text{contradiction}.$$

The first implication is topological. A connected cover of a manifold inherits the pullback Lorentz metric, and path lifting converts relative homotopy into a statement about which sheet contains the endpoint. The second implication is global and causal: compactness properties obtained from global hyperbolicity must produce a generator on the boundary of a causal future, while the asymptotic hypotheses prevent the construction from escaping through spatial infinity. The third implication selects the initially converging null-normal family. The fourth is analytic: Raychaudhuri evolution plus null convergence, or the precisely directed averaged condition in the theorem, forces a focal point. The final implication uses causal variation: after a focal point, that generator can no longer remain on an achronal boundary.

This division of labor explains why ANEC alone is not a censorship theorem. It can power the focusing step but cannot create the required generator, choose the correct asymptotic end, or guarantee that the relevant causal diamonds are compact. The later asymptotically flat result records these global ingredients explicitly in Chruściel–Galloway 2019, Theorem 3.1, Remarks 3.2–3.4, and Proposition 3.6, pp. 4–9.

The historical qualification is substantive. Friedman, Schleich, and Witt’s 1995 erratum withdrew a secondary statement about passive visibility of $K(\pi,1)$ prime factors but said their active theorem was unaffected Friedman–Schleich–Witt 1995, erratum, p. 1872. Krasnikov later identified a distinct gap in the active proof: a nonempty clopen subset of a connected lifted infinity might be the whole component, so the desired boundary point did not follow Krasnikov 2013, §2, pp. 54–56. The chapter therefore presents the original theorem as history and uses a later sufficient theorem for the proof-level result, rather than treating the two sets of hypotheses as interchangeable.

Horizons, dimensions, and what “positive” means

For a closed two-sided MOTS $\Sigma$ inside a spacelike hypersurface, outward stability makes the principal eigenvalue of the MOTS stability operator nonnegative. Combining its positive principal eigenfunction with the effective dominant energy condition yields a conformal scalar-curvature inequality. Except in a rigid Ricci-flat, shear-free, zero-flux alternative, $\Sigma$ admits a metric of positive scalar curvature. Galloway and Schoen give the theorem and stability extension in Galloway–Schoen 2006, Theorem 2.1, proof, and concluding remarks, pp. 3–6.

In a four-dimensional spacetime, $\Sigma$ is two-dimensional. If it is orientable, Gauss–Bonnet and surface classification turn positive scalar curvature into positive Euler characteristic and hence $\Sigma\simeq S^2$. In five spacetime dimensions, $\Sigma$ is three-dimensional, and the same inference fails: $S^2\times S^1$ admits positive scalar curvature. Thus the black ring is not an exception to a spherical theorem; it demonstrates that the theorem’s higher-dimensional conclusion was always weaker than sphericity.

This theorem concerns stable marginally outer trapped cross-sections, not arbitrary slices of an event horizon. Connecting an event-horizon cut to the MOTS hypotheses requires additional global and slicing information. A surface that meets a timelike boundary or corner also requires a theorem with boundary terms; the interior theorem is applied only after an explicit restriction to a boundaryless tubular neighborhood of an interior cross-section.

Controls and failure modes

Several examples keep the logical boundaries sharp.

A globally hyperbolic cylinder has Cauchy surface $S^1$ and nontrivial fundamental group. It satisfies the product theorem and refutes the claim that global hyperbolicity alone enforces simple connectivity.

An Einstein–Rosen bridge can appear on a Cauchy slice of maximally extended Schwarzschild without providing a causal path from one infinity through the throat to another. Censorship constrains the path, not the mere existence of a bridge in spatial geometry.

A traversable Morris–Thorne throat needs defocusing of its threading null congruence and violates pointwise null convergence at the throat; standard versions also fail the relevant averaged condition. This is a failure of a stated premise, not a contradiction of the conclusion.

A timelike conformal boundary changes the theorem. In asymptotically locally AdS settings the boundary may itself have nontrivial topology. The appropriate result is then often surjectivity

$$i_*:\pi_1(\mathscr I)\longrightarrow\pi_1(D'),$$

not absolute simple connectivity. Boundary loops can account for exterior loops; the exterior becomes simply connected only when the boundary component is simply connected. The boundary-capable theorem and this algebraic consequence are proved in Galloway–Schleich–Witt–Woolgar 1999, Theorems 2.1–2.2 and Proposition 3.1, pp. 7–12.

A topology-changing pair of pants is a valid cobordism but not a smooth causally compact chronological evolution between Cauchy boundaries. In Borde’s formulation, chronology forces the causally compact Lorentzian interpolating cobordism to be a product Borde 1994, Theorem 1 and proof, pp. 17–18. Topology change must therefore pay through chronology violation, loss of causal compactness, metric degeneracy, singular or incomplete behavior, or a boundary that is not regular Cauchy data.

Review the chapter

For each situation, identify the strongest conclusion supported by the stated data and name one tempting conclusion that does not follow.

1. A smooth globally hyperbolic spacetime without boundary has a compact Cauchy hypersurface diffeomorphic to $T^3$. No asymptotic structure or energy condition is supplied.
2. A four-dimensional time-oriented spacetime contains a closed connected two-sided orientable strictly stably outermost MOTS, and the effective dominant energy condition holds near it.
3. A globally hyperbolic exterior completed by a connected timelike conformal boundary obeys the boundary theorem’s compact-cut, null-completeness, and averaged Ricci hypotheses, but the boundary has nontrivial fundamental group.
4. A compact Lorentzian cobordism has two spacelike boundary components of different topology, is causally compact, and its metric is everywhere smooth and nondegenerate.

Solution

In case 1, smooth Cauchy splitting gives $M\simeq\mathbb R\times T^3$, and every smooth spacelike Cauchy hypersurface is diffeomorphic to $T^3$. Simple connectivity does not follow: indeed $\pi_1(T^3)=\mathbb Z^3$. Without an asymptotic end and a focusing condition, topological censorship is not applicable.

In case 2, strict stability and the effective dominant energy condition give positive Yamabe type without the rigid exception. Because the cross-section is a closed orientable two-manifold, Gauss–Bonnet and classification make it diffeomorphic to $S^2$. This does not prove that the domain of outer communications is simply connected, nor does it identify the surface as an event-horizon cut.

In case 3, the inclusion-induced map from the boundary fundamental group onto the exterior fundamental group is surjective. Exterior loop classes must come from boundary loop classes. The exterior need not be simply connected because the boundary was not assumed simply connected; surjectivity from a nontrivial group is weaker than triviality of the target.

In case 4, the product obstruction says that chronology cannot also hold. Since the two boundary topologies differ, at least one premise of a chronological product evolution must fail. Under the data given, the necessary failure is chronology, so a closed timelike curve occurs. The cobordism itself is not topologically impossible; it is incompatible with the full causal package.

Onward use

The chapter’s conclusions feed into different later questions. Exterior simple connectivity is an input to classification and uniqueness arguments, but it is not Weak Cosmic Censorship or Strong Cosmic Censorship. A product Cauchy development clarifies what fixed initial-data topology means, while topology-change obstructions clarify which hypotheses must fail if a proposed interpolation changes it. Horizon positive-Yamabe type supplies a curvature-topology restriction whose classification content depends strongly on dimension.

The practical habit to carry forward is to ask four questions before invoking a topology theorem: Which space is being constrained—the whole spacetime, a Cauchy slice, the domain of outer communications, or a horizon cut? What kind of boundary is present? Which causal compactness and asymptotic assumptions produce the global curve? Which pointwise or averaged energy condition produces focusing? Once those are explicit, apparent paradoxes usually become ordinary failures of hypotheses.

References

• Aké Hau, Luis, José L. Flores, and Miguel Sánchez, “Structure of Globally Hyperbolic Spacetimes-with-Timelike-Boundary”, Revista Matemática Iberoamericana 37, no. 1, pp. 45–94, 2021. Open PDF.
• Bernal, Antonio N., and Miguel Sánchez, “Smoothness of Time Functions and the Metric Splitting of Globally Hyperbolic Spacetimes”, Communications in Mathematical Physics 257, pp. 43–50, 2005. Open PDF.
• Borde, Arvind, “Topology Change in Classical General Relativity”, preprint, 1994.
• Chruściel, Piotr T., and Gregory J. Galloway, “Roads to Topological Censorship”, preprint UWThPh-2019-17, 2019.
• Friedman, John L., Kristin Schleich, and Donald M. Witt, “Topological Censorship”, Physical Review Letters 71, pp. 1486–1489, 1993; erratum, volume 75, p. 1872, 1995. 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.
• Galloway, Gregory J., Kristin Schleich, Donald M. Witt, and Eric Woolgar, “Topological Censorship and Higher Genus Black Holes”, Physical Review D 60, 104039, 1999. Open PDF.
• Krasnikov, Sergey V., “‘Topological Censorship’ Is Not Proven”, Gravitation and Cosmology 19, no. 1, pp. 54–56, 2013. Open PDF.