What Singularity Theorems Conclude

The word “singularity” is stronger in ordinary language than in the conclusion of a classical singularity theorem. The theorems prove incomplete causal geodesics under stated hypotheses. They do not place a defective point inside the manifold, calculate curvature at an endpoint, decide whether an extension exists, or determine what a distant observer sees. Those may be natural next questions, but each needs additional geometry or dynamics.

Required background. Congruences and Optical Scalars supplies the focusing input; Einstein Equations explains the conditional role of energy assumptions. Helpful background. Lorentzian Metrics and Causal Cones separates causal types; Geodesics and Jacobi Fields defines affine completeness; Causality and Global Structure supplies global boundaries; Curvature supplies invariant blowup tests.

The exact conclusion

An affinely parametrized geodesic is incomplete when it is inextendible as a geodesic in $(M,g)$ but its maximal affine-parameter interval has a finite endpoint. For a timelike geodesic parametrized by proper time, this means finite proper duration in the incomplete direction. For a null geodesic, affine—not coordinate—parameter is decisive. A nonlinear reparametrization can make a finite coordinate interval infinite without repairing affine incompleteness.

The three theorem conclusions in this chapter are different:

Theorem form	Guaranteed causal type	Guaranteed direction
Penrose with a future trapped surface	at least one null geodesic	future
Hawking with a presently expanding Cauchy slice	at least one timelike geodesic	past
Hawking–Penrose in its combined classical form	at least one timelike or null geodesic	not fixed in the basic statement

These are the minimal existential conclusions. “At least one” cannot be replaced by “every” without an additional argument. The uniform-expansion Hawking proof does supply such extra information for timelike histories starting on its specified slice: their past proper duration is uniformly bounded. That does not make Penrose’s null conclusion universal or identify a common endpoint. Timelike cannot be exchanged with null, and a directional conclusion cannot be inferred when the theorem states none. Steinbauer emphasizes that the classical conclusions are causal-geodesic incompleteness and distinguishes them from extension and curvature questions in Steinbauer 2023, §2.3, pp. 8–9 and §4.1, pp. 21–22.

No singular point is produced

Every point of a Lorentzian manifold has an ordinary smooth neighborhood. A “singular point of spacetime” is therefore not one of the points at which the theorem evaluates a divergent tensor. The incomplete geodesic simply has no endpoint in $M$ at the finite limiting affine parameter. Attaching an ideal endpoint requires a boundary construction, and distinct constructions can encode endpoints differently.

The elementary check is punctured Minkowski spacetime. Remove the origin from otherwise complete flat spacetime. A straight geodesic aimed at the deletion has finite affine length and no endpoint in the punctured manifold. Yet

$$R_{abcd}=0$$

everywhere on $M$, and reinserting the point gives a smooth extension. Incompleteness alone therefore detects a missing continuation, not its cause. Geroch’s analysis of the competing singularity criteria and their limitations remains a useful exact guide Geroch 1968, pp. 526–540.

A theorem applied to an arbitrarily truncated region may diagnose only the truncation. Physical applications therefore normally choose a maximally extended spacetime or, in an initial-value setting, the maximal globally hyperbolic development. “Maximal” must still name a category: a spacetime may be inextendible with a $C^2$ metric yet extendible continuously, or maximal among globally hyperbolic developments yet extendible beyond a Cauchy horizon as a non-globally-hyperbolic spacetime.

Curvature blowup is additional information

None of the Raychaudhuri inequalities requires curvature to become unbounded. The convergence condition is a sign inequality, and strict initial expansion can focus a congruence even in a region where curvature stays finite. A focal point is a failure of a Jacobi map or exponential map to remain nonsingular; it is not a curvature singularity. Radial null rays focus at the center of flat spacetime.

Curvature blowup can independently strengthen an application. If a scalar polynomial invariant such as

$$R_{abcd}R^{abcd}$$

diverges along an incomplete curve, a regular $C^2$ endpoint with bounded curvature cannot be attached along that approach. But absence of scalar-invariant divergence is not a general regularity theorem. Parallelly propagated frame components can diverge while polynomial scalars remain finite, and weak null singularities can have subtler extension properties.

“Strong singularity” is another extra notion. Tipler and Królak give curve-specific Jacobi-field conditions on transverse volume or screen area; Curvature Blowup and Measures of Singularity Strength states their quantifiers explicitly. Curvature-integral tests are derived criteria under additional hypotheses, not the definitions themselves. The classical Penrose, Hawking, and Hawking–Penrose statements do not establish these strength conditions. Exact solutions or separate comparison estimates may.

Schwarzschild illustrates the distinction between theorem and model analysis. The Penrose hypotheses diagnose future null incompleteness once a trapped sphere exists. The explicit Schwarzschild metric then shows that the Kretschmann scalar diverges as $r\to0$. That invariant calculation is not hidden inside the general theorem.

Incompleteness is not inextendibility

An extension of $(M,g)$ is an isometric embedding into a larger spacetime of a specified differentiability class whose image is proper. Geodesic incompleteness can be repaired by an extension, as in punctured Minkowski space or a restricted coordinate region. Conversely, an incomplete spacetime may be $C^0$-inextendible; establishing that requires a separate obstruction.

Regularity matters twice. First, the classical theorem assumes enough differentiability for pointwise curvature, geodesic uniqueness, and Jacobi fields. Second, the class of allowed extensions determines what counts as a repair. Divergent curvature may block $C^2$ extension without blocking $C^0$ extension. The hierarchy is developed on Regularity Classes and Inextendibility.

If an extension preserved all hypotheses of a singularity theorem at the same regularity, applying the theorem to the larger spacetime would still force some incomplete causal geodesic. But the original theorem does not prove that every proposed extension preserves global hyperbolicity, trapped geometry, convergence, genericity, or the relevant topology. One cannot turn conditional reapplication into unconditional inextendibility.

No horizon or censorship verdict follows

The Penrose theorem assumes a closed trapped surface, not an event horizon. It can be stated without a conformal future infinity, so the usual event-horizon definition may not even be available. It proves neither that the incomplete geodesic is hidden from infinity nor that a black-hole region forms. Conversely, an event horizon can exist in a spacetime region without the theorem’s trapped-surface hypotheses being invoked.

Weak cosmic censorship asks, roughly and formulation-dependently, whether singular behavior produced by generic asymptotically flat data is hidden from distant observers. Strong cosmic censorship asks whether the maximal globally hyperbolic development is generically inextendible at a chosen regularity. A singularity theorem supplies conditional incompleteness; it does not settle genericity in a data topology, visibility, or extension regularity. Those axes are made explicit in Weak Cosmic Censorship and Strong Cosmic Censorship.

Topology also plays only the role stated in a theorem. Penrose uses a noncompact Cauchy surface to reach a contradiction. Hawking uses a compact Cauchy surface in one cosmological version. Hawking–Penrose admits a compact achronal edgeless set as one alternative. None of these conclusions classifies the topology of an ideal endpoint or proves a universal topology-change rule.

No automatic breakdown scale or quantum remedy

Finite affine length does not identify a curvature scale at which classical general relativity fails. The incomplete segment may encounter Planckian curvature in a model, finite curvature at an extendible boundary, or no scalar blowup at all. A theorem also does not prescribe boundary conditions, quantum evolution, singularity resolution, or a replacement spacetime.

Violations of classical pointwise energy conditions by quantum fields do not simply negate every singularity result. They show that a theorem with that hypothesis cannot be applied without further work. Averaged conditions, quantum inequalities, generalized entropy, or semiclassical backreaction lead to different theorems and must state their own domains. Likewise, low-regularity extensions of the classical theorems change the analytic tools but still conclude a specified form of incompleteness; they do not supply a quantum endpoint.

A safe interpretation protocol

When using a singularity theorem, report conclusions in this order:

1. name the spacetime and regularity class;
2. list the causal, convergence, genericity, and initial/topological hypotheses actually verified;
3. state the causal type and direction of the incomplete geodesic exactly;
4. identify separately any model-specific curvature, strength, visibility, or extension result;
5. state which stronger questions remain open.

This protocol keeps a rigorous theorem physically informative. “The hypotheses force future null incompleteness” is already a powerful global statement. It need not be inflated into “general relativity predicts an infinite-curvature point inside every black hole.”

Exercise

Assume all hypotheses of the future-directed Penrose theorem. Classify the following as guaranteed, excluded, or undecided by that theorem alone: (a) one future-incomplete null geodesic; (b) curvature-scalar divergence; (c) a future event horizon; (d) future timelike incompleteness; (e) absence of every $C^0$ extension.

Solution

(a) is guaranteed: it is the theorem’s exact conclusion. Items (b)–(e) are undecided, not excluded. The theorem neither computes curvature nor assumes an asymptotic infinity needed to define an event horizon. It concludes null rather than timelike incompleteness and says nothing about $C^0$ extensions. Particular spacetimes can satisfy any of these stronger properties, but each requires a separate argument.

References

• Geroch, Robert P., “What Is a Singularity in General Relativity?”, Annals of Physics 48, pp. 526–540, 1968.
• Steinbauer, Roland, “The Singularity Theorems of General Relativity and Their Low Regularity Extensions”, Jahresbericht der Deutschen Mathematiker-Vereinigung 125, §2.3, pp. 8–9; §4.1, pp. 21–22, 2023.