Regularity Across Cauchy Horizons

Whether a black-hole spacetime is “extendible” has no definite meaning until the differentiability class of the extension is named. A continuous Lorentzian metric, a metric with locally square-integrable connection, and a twice continuously differentiable metric support different geometric and dynamical structures. Strong cosmic censorship is therefore a family of regularity-dependent statements, not a single yes-or-no assertion.

Helpful background. Black-Hole Geometry and Mechanics supplies inner horizons; Black-Hole Perturbation Theory supplies exterior decay assumptions; Formulations of General Relativity supplies weak and gauge-fixed evolution; Causality and Global Structure supplies maximal developments and extensions; Computational Relativity supplies invariant diagnostics.

A hierarchy of extension classes

Let $(\mathcal M,g)$ be the maximal globally hyperbolic development of admissible initial data. A $C^k$ extension consists of a larger manifold $\widetilde{\mathcal M}$, an embedding $\iota:\mathcal M\to\widetilde{\mathcal M}$ onto a proper subset, and a nondegenerate Lorentzian metric $\widetilde g$ of class $C^k$ satisfying $\iota^*\widetilde g=g$. Different $k$ answer different questions.

Class	What remains meaningful	What can fail
$C^0$	continuous causal cones, lengths of piecewise smooth curves	classical connection, curvature, Einstein equation
$C^1$	continuous first derivatives and connection	classical curvature
$C^{1,1}$	locally Lipschitz first derivatives; curvature almost everywhere	pointwise $C^2$ curvature
$W_{\rm loc}^{1,2}$ with nondegenerate continuous metric	Christoffel symbols in $L_{\rm loc}^2$ and some weak formulations	classical pointwise geometry
$C^2$	classical curvature and geodesic equation with differentiable connection	higher derivative control

The inclusions run from stronger to weaker regularity: a $C^2$ extension is also $C^0$, but a $C^0$ extension need not possess a locally square-integrable connection. Therefore $C^2$ inextendibility can coexist with $C^0$ extendibility. The regularity hierarchy and its significance for the Cauchy horizon are explicit in Luk and Oh 2019, §§1 and 3.4–3.6, pp. 1–21, PDF.

A precise spherical theorem

Consider the four-dimensional Einstein–Maxwell–real-scalar system in spherical symmetry with two-ended, asymptotically flat admissible Cauchy data. The proof is a two-paper argument. Part I proves $C^2$ future inextendibility conditional on an $L^2$-averaged inverse-polynomial lower bound for the scalar radiation along both event horizons; Part II proves that lower bound on an open set in a weighted $C^1$ topology whose complement has codimension at least one, and hence on a dense set in a weighted $C^\infty$ topology. Earlier results give continuous extension across a nonempty Cauchy horizon for the corresponding charged subextremal class.

The conditional interior implication is Luk and Oh 2019a, Theorem 5.7, p. 29, with proof in §11, pp. 58–61, PDF. Stability of the Cauchy horizon and blowup are separated in the preceding Theorems 5.1, 5.4, and 5.5. The generic exterior lower bound is established in Luk and Oh 2019b, Theorems 4.1–4.3, pp. 17–18, PDF. The combined censorship conclusions are stated in Part I as Theorems 3.14 and 3.18 and Corollary 3.21, pp. 19–21; their proofs use both papers.

The proof architecture has four pieces:

1. exterior estimates provide a sufficiently sharp lower bound on radiation entering the event horizon;
2. redshift estimates transport controlled data into the black hole;
3. blueshift estimates turn the lower bound into divergence at the Cauchy horizon;
4. a contradiction argument shows that a hypothetical $C^2$ extension would keep a curvature quantity bounded where the established lower bound forces blowup.

Every hypothesis matters. Spherical symmetry suppresses gravitational radiation, the Maxwell charge produces a subextremal inner horizon, two-ended asymptotic flatness fixes the global data class, and “generic” refers to the topology defined in the theorem—not to an informal claim about typical astrophysical collapse.

Determinism and weak evolution

The original motivation for strong cosmic censorship is deterministic evolution: generic initial data should determine a unique maximal spacetime in an appropriate solution class Penrose 1969, pp. 1141–1165 in the 2002 reprint. At $C^0$, one can discuss a continuous geometry beyond the Cauchy horizon, but the Einstein equation is not classically defined. At $C^2$, curvature is classical, and blowup blocks extension decisively. Christodoulou’s proposed threshold asks for a connection regular enough to formulate the equations weakly, often expressed by local square integrability of Christoffel symbols.

These thresholds are not ordered by philosophical preference alone. Hyperbolic uniqueness theorems have explicit coefficient regularity requirements. A weak extension is physically relevant only if an evolution equation and a uniqueness class can be specified. Mere existence of many continuous extensions may instead display the breakdown of determinism.

For nonsymmetric vacuum Kerr perturbations, Dafermos and Luk prove a $C^0$ extension across a nontrivial Cauchy-horizon segment under quantified interior closeness assumptions Dafermos and Luk 2025, Theorem 4.24, pp. 50–65, and §16, pp. 165–185, PDF. That theorem neither establishes a $C^2$ extension nor supplies the generic curvature lower bound required for inextendibility. The 2026 exterior result now supplies a narrower rigorous bridge: Hintz reports that its decay estimates meet the Dafermos–Luk hypotheses, giving $C^0$ stability of the Cauchy horizon for the stated small vacuum perturbations of rotating subextremal Kerr Hintz 2026, v2, §1.1, consequence (1), p. 3, PDF. This current preprint does not prove generic curvature blowup, classify the complete singular boundary, or cover arbitrary collapse data.

What a counterexample changes

Exact Reissner–Nordström and Kerr are smooth across their Cauchy horizons, so they are counterexamples to any formulation that omits genericity. Spherical Einstein–Maxwell–scalar solutions with mass inflation can be $C^0$ extendible but $C^2$ inextendible, refuting the implication “continuous extension means regular continuation.” Conversely, proving curvature blowup along one symmetry-restricted family does not prove a generic nonsymmetric theorem.

The correct conclusion always has four labels: model, data class, genericity topology, and extension regularity.

Check your understanding

Suppose a metric extends continuously across a null boundary, while a parallel-propagated curvature component diverges on approach. Which of $C^0$ extendibility and $C^2$ extendibility are decided?

Solution

The exhibited continuous nondegenerate extension establishes $C^0$ extendibility. Divergent curvature along the original spacetime obstructs any $C^2$ extension compatible with that approach, because a $C^2$ metric has continuous, hence locally bounded, curvature components in regular frames. Nothing stated decides $C^1$, $C^{1,1}$, or $W_{\rm loc}^{1,2}$ extendibility; each requires separate derivative or integrability estimates.

References

• Dafermos, Mihalis, and Jonathan Luk, “The Interior of Dynamical Vacuum Black Holes I: The C⁰-Stability of the Kerr Cauchy Horizon”, Annals of Mathematics 202, 309–630, 2025. Open PDF.
• Hintz, Peter. “Nonlinear Stability of Subextremal Kerr Black Holes.” arXiv:2606.28253v2 (3 August 2026). Open PDF.
• Luk, Jonathan, and Sung-Jin Oh, “Strong Cosmic Censorship in Spherical Symmetry for Two-Ended Asymptotically Flat Initial Data I: The Interior of the Black Hole Region”, Annals of Mathematics 190, 1–111, 2019. Open PDF.
• Luk, Jonathan, and Sung-Jin Oh, “Strong Cosmic Censorship in Spherical Symmetry for Two-Ended Asymptotically Flat Initial Data II: The Exterior of the Black Hole Region”, Annals of PDE 5, 6, 2019. Open PDF.
• Penrose, Roger, “Gravitational Collapse: The Role of General Relativity”, General Relativity and Gravitation 34, 1141–1165, 2002 reprint of the 1969 article.