Regularity-Dependent Formulations of Strong Cosmic Censorship

Strong cosmic censorship asks whether generic initial data determine a spacetime that cannot be continued beyond its maximal globally hyperbolic development. The decisive index is the extension regularity. A black-hole interior can extend as a continuous Lorentzian metric while curvature and the connection are too singular for a classical or even weak Einstein evolution. Consequently $C^0$, locally Lipschitz, Christoffel-$L^2$, and $C^2$ formulations are different conjectures with different logical strength.

Required background. Analysis for Mathematical Relativity supplies weak regularity and genericity, while the Einstein equation supplies the initial-value problem.

Helpful background. Causality and Global Structure supplies Cauchy horizons and domains of dependence; PDEs and Green Functions for Gravitation supplies weak-solution classes; and Black-Hole Geometry and Mechanics supplies charged and rotating interior models.

The conjecture as an indexed statement

Fix

$$\mathfrak C_{\mathrm S} =(\mathcal E,\mathcal D,\mathcal T,\mathcal G, \mathcal R,\mathcal W).$$

The entries are an Einstein–matter system $\mathcal E$, constraint-satisfying data space $\mathcal D$, topology $\mathcal T$, generic subset $\mathcal G$, metric extension class $\mathcal R$, and—when required—a weak formulation $\mathcal W$ for the metric and matter equations.

Assume the chosen Einstein–matter system is locally well posed and admits the geometric uniqueness and patching construction. For each $d\in\mathcal D$ in that regularity class, the maximal-development theorem then gives an MGHD, unique up to an isometry fixing the embedded data. Strong censorship at $(\mathcal R,\mathcal W)$ asserts:

For every $d\in\mathcal G$, $\operatorname{MGHD}(d)$ admits no proper extension whose metric belongs to $\mathcal R$ and whose fields satisfy $\mathcal W$ when that equation clause is imposed.

An extension is a connected Lorentzian manifold $(\widetilde M,\widetilde g)$ and an isometric embedding $\iota:M\hookrightarrow\widetilde M$ with proper image. “Maximal” in MGHD does not already prove inextendibility: it says only that the original initial hypersurface cannot remain Cauchy in a larger development. A Cauchy-horizon extension is precisely a larger spacetime in which determinism from that hypersurface fails.

The regularity hierarchy

For classical Hölder differentiability,

$$\{\text{smooth extensions}\} \subset \{C^2\text{ extensions}\} \subset \{C^{0,1}\text{ extensions}\} \subset \{C^0\text{ extensions}\}.$$

Taking complements reverses the arrows:

$$C^0\text{-inextendible} \Longrightarrow C^{0,1}\text{-inextendible} \Longrightarrow C^2\text{-inextendible} \Longrightarrow C^\infty\text{-inextendible}.$$

Thus $C^0$ SCC is the strongest assertion in this chain. A continuous extension can exist even when no $C^2$ extension does. The following thresholds explain why.

• A $C^0$ Lorentzian metric has continuous cone and length structures, but no classical connection or curvature.
• A locally Lipschitz metric has first weak derivatives in $L^\infty_{\mathrm{loc}}$ and Christoffel symbols almost everywhere.
• A $C^{1,1}$ metric has locally bounded weak curvature and a well-posed classical geodesic equation.
• A $C^2$ metric has continuous classical curvature.

The Christodoulou weak-solution threshold is not simply a synonym for one line above. In common formulations the metric and inverse are controlled and the Christoffel symbols lie in $L^2_{\mathrm{loc}}$, so the quadratic connection terms can be interpreted distributionally. Matter fields require matching spaces. The definition is gauge sensitive unless written geometrically, and a continuous extension with connection outside $L^2$ does not count as a weak Einstein extension in this sense.

For a concrete regularity check, consider the local metric

$$g=-dt^2+(1+|t/t_0|^\alpha)^2dx^2+dy^2+dz^2, \qquad 0\lt\alpha\lt1,\quad t_0\gt0.$$

It is continuous and nondegenerate across $t=0$. Writing $a(t)=1+|t/t_0|^\alpha$, with fixed time scale $t_0$, its nonzero connection components include $\Gamma^x{}_{tx}=a'/a$ and $\Gamma^t{}_{xx}=aa'$, of size $|t|^{\alpha-1}$. Their squares are locally integrable precisely for $\alpha\gt1/2$, since $\int_0^\epsilon t^{2\alpha-2}dt$ converges exactly in that range. Thus continuous extendibility does not even determine whether this coordinate representative meets the connection-$L^2$ threshold. This is a regularity example, not an Einstein solution or a proof excluding all alternative extensions; a censorship argument must still establish the field equations and the invariant obstruction.

Why the formulations can disagree

Suppose an MGHD has a null boundary across which $g$ extends continuously, but the Hawking mass and transverse derivatives diverge. Then:

• $C^0$ inextendibility is false for that datum;
• curvature blowup may establish $C^2$ inextendibility;
• failure of every extension to have Christoffel symbols in $L^2_{\mathrm{loc}}$ establishes the selected weak-solution inextendibility;
• none of those one-solution facts supplies genericity.

This pattern is called a weak null singularity. “Weak” does not mean the curvature is small; it records survival of a low-regularity metric and possibly finite integrated tidal deformation. The rotating-vacuum construction of Dafermos and Luk starts from suitably decaying spacelike Cauchy data on a hypersurface already inside a dynamical black hole. Its maximal development has a nonempty Cauchy-horizon segment across which the metric admits a $C^0$ extension Dafermos–Luk 2025, Theorems 4.2 and 4.24, pp. 50–52 and 64–65 of the open manuscript. It does not provide a $C^2$ extension or a universal genericity verdict.

Conversely, $C^0$ inextendibility need not arise from curvature. Schwarzschild curvature diverges at $r=0$, but proving that no continuous Lorentzian extension exists requires a low-regularity geometric argument. Sbierski controls intrinsic spacelike diameter and causal geometry to establish $C^0$ inextendibility Sbierski 2018, Theorem 1.1 and §§3–5. Trying to evaluate the curvature tensor of a hypothetical $C^0$ extension would be ill posed.

Genericity and theorem scope

A complete SCC formulation attaches the regularity claim to a generic subset of a named data space. Isenberg’s survey records nonlinear vacuum results for $T^3$ Gowdy spacetimes: within that symmetry class, appropriate open/dense sets have $C^2$-inextendible MGHDs Isenberg 2015, §§4–5, pp. 7–11. These are genuine model theorems, but the two-Killing-field subspace is not open in unrestricted vacuum data.

Exact Reissner–Nordström, Kerr, and Taub–NUT Cauchy horizons likewise do not falsify a generic conjecture. One must perturb within $\mathcal D$, control the exterior tail entering the black hole, prove the nonlinear interior behavior, and then translate it to $\mathcal R$. A linear test field can identify a blue-shift mechanism but cannot establish metric backreaction or an MGHD extension theorem.

The topology and extension class interact. A property open in two-ended spherical Einstein–Maxwell–scalar characteristic data need not be open among one-ended collapse data. A result for smooth perturbations may fail in a rough topology that contains self-similar singular backgrounds. The word “generic” does not repair a missing choice of phase space.

Determinism at several thresholds

If a $C^0$ extension exists, continuous causal geometry can be appended beyond the Cauchy horizon in more than one way, so metric-level determinism fails. If no extension has an $L^2$ connection, the original evolution may still be maximal among admissible weak solutions. If a $C^2$ metric extension exists, its classical curvature and geodesic equations are defined across the boundary. It continues an Einstein solution only if the extended matter and field equations are also required and satisfied.

There is no theorem selecting one threshold as the only meaningful physical one. The mathematically honest practice is to carry $\mathcal R$ and $\mathcal W$ in every claim. Inextendibility Criteria develops the proof tools for each class; the current theorem matrix keeps their status separate through 2026-08-12.

Check the hierarchy

Suppose one MGHD admits a proper $C^0$ extension, every such extension has Christoffel symbols outside $L^2_{\mathrm{loc}}$, and a scalar curvature invariant diverges along every approach to the boundary. Classify the $C^0$, locally Lipschitz, Christodoulou weak-solution, $C^2$, and smooth statements for this datum.

Solution

$C^0$-inextendibility is false because an extension is given. A locally Lipschitz metric has locally bounded first weak derivatives and hence an $L^2_{\mathrm{loc}}$ connection on compact sets, so no such extension exists. The stipulated Christodoulou class is also excluded. Divergent scalar curvature rules out $C^2$ extension and therefore smooth extension. These conclusions concern one datum; an SCC theorem still needs them on $\mathcal G$ in the named data topology.

References

• Dafermos, Mihalis, and Jonathan Luk. “The Interior of Dynamical Vacuum Black Holes I: The $C^0$-Stability of the Kerr Cauchy Horizon.” Annals of Mathematics 202 (2025): 309–630. DOI. Open PDF.
• Isenberg, James. “On Strong Cosmic Censorship.” Surveys in Differential Geometry 20 (2015): 17–36. arXiv.
• Sbierski, Jan. “The $C^0$-Inextendibility of the Schwarzschild Spacetime and the Spacelike Diameter in Lorentzian Geometry.” Journal of Differential Geometry 108 (2018): 319–378. arXiv.