What Counts as a Naked Singularity?

A singularity is called naked only relative to a specified notion of singular behavior and a specified receiver of signals. A locally visible incomplete endpoint, a curvature blowup in the causal past of future null infinity, and an incomplete future null infinity are related but not interchangeable conditions. The distinction is essential: weak cosmic censorship concerns exterior visibility, whereas an inner Cauchy horizon can threaten determinism without exposing anything to distant observers.

Required background. The Einstein Equation supplies the field equations and their constraint content. Helpful background. Pseudo-Riemannian Metrics and Signature supplies Lorentzian geometry; Affine and Variational Geodesics supplies completeness language; Causality and Global Structure supplies causal boundaries; Formulations of General Relativity separates constraints, evolution, and gauge; Exact Solutions of Einstein’s Equations supplies comparison geometries; PDEs and Green Functions for Gravitation supplies evolution and boundary-value language.

Singular behavior has no automatic endpoint

Every point of a smooth spacetime manifold is regular. A curvature “singularity at $p$” therefore cannot literally be a point $p\in M$ at which the metric is smooth and the curvature is infinite. What fails is continuation along a curve, extension of the metric, boundedness of a geometric quantity, or some combination of these.

Several candidate diagnoses must be kept separate:

• a future-directed timelike or null geodesic is inextendible in $(M,g)$ but has finite proper or affine length;
• the spacetime has no extension in a named class, such as $C^0$, locally Lipschitz, $C^2$, or smooth;
• curvature scalars or parallelly propagated curvature components diverge along an approach;
• the fields fail to satisfy a weak formulation of the Einstein–matter equations across a boundary; or
• an ideal endpoint appears in a conformal, causal, or other boundary construction.

Geodesic incompleteness alone is not enough. Minkowski spacetime with one point deleted has incomplete geodesics, yet adding the point gives a smooth extension. Conversely, scalar polynomial invariants can remain bounded even when frame components or derivatives obstruct an extension. Geroch’s comparison of singularity criteria explains why no single elementary criterion captures every pathology in Geroch 1968, pp. 526–540.

For censorship, one normally begins with regular constraint-satisfying data and examines their maximal globally hyperbolic development. This removes artificial truncations. One must then state which failure at the boundary is being called singular. A removable boundary of a nonmaximal chart is not a naked singularity produced by collapse.

Local and global visibility

Let $\mathcal S$ denote an ideal singular boundary or an equivalence class of incomplete approaches; it is notation, not a new point of $M$. A locally naked singularity is present when future-directed causal curves can emerge from arbitrarily near $\mathcal S$ into a regular region. More invariantly, there are regular events $q\in M$ whose causal pasts contain sequences tending to the singular boundary. This is a statement about a neighborhood of the endpoint. It need not say that any signal reaches a distant observer.

A globally naked singularity requires a specified distant receiver. In an asymptotically flat spacetime, the usual receiver is future null infinity $\mathscr I^+$. One common causal formulation says that the relevant singular approach lies in the causal past of $\mathscr I^+$: signals from arbitrarily near it can escape to infinity. Another formulation diagnoses global nakedness through future incompleteness of $\mathscr I^+$ in the MGHD generated by complete asymptotically flat data. These formulations become comparable only with additional regularity, asymptotic predictability, and boundary assumptions.

In spherical Einstein–scalar work, Zheng uses an especially sharp version: suitably regular, spherically symmetric, complete asymptotically flat characteristic data generate a spacetime with a globally naked singularity when the MGHD admits a future null infinity that is future incomplete Zheng 2026, Definition 1.5, pp. 6–7. That definition is tailored to that model and characteristic setup; it is not a universal definition for every asymptotic geometry.

Local visibility does not imply global visibility. Outgoing rays may enter a trapped region later, hit an event horizon, or fail to reach the chosen asymptotic component. Global visibility does imply an exterior causal channel, but it still does not determine whether curvature diverges, whether the singular boundary is spacelike, null, or timelike, or which metric extensions are possible.

Visibility in different asymptotic settings

The phrase “visible from infinity” changes meaning with the conformal boundary.

Setting	Natural receiver	Necessary formulation data	Main caution
Asymptotically flat	future null infinity $\mathscr I^+$	conformal completion, chosen asymptotic end, completeness or causal-past condition	the standard weak-censorship setting
Asymptotically de Sitter	a component of spacelike $\mathscr I^+$ or a chosen observer’s worldline	cosmological horizons and observer/end selection	there is no asymptotically flat null infinity
Asymptotically anti-de Sitter	the timelike conformal boundary	boundary conditions, reflecting/absorbing choice, boundary causal structure	evolution is an initial-boundary value problem

In de Sitter asymptotics, a source can be hidden from one observer by a cosmological horizon while influencing another portion of spacelike future infinity. “Naked to whom?” is therefore part of the definition. In anti-de Sitter asymptotics, signals can reach the timelike boundary and return; the boundary condition helps define both the evolution and the receiver. An asymptotically flat theorem about complete $\mathscr I^+$ cannot be relabeled as a de Sitter or AdS theorem.

Even within asymptotic flatness, multiple ends matter. A curve may escape to one component of $\mathscr I^+$ but not another. A precise statement quantifies over the relevant end or all ends and states whether future timelike infinity is also controlled.

Horizons do not define nakedness by themselves

For a chosen asymptotically flat end, the black-hole region is

$$\mathcal B=M\setminus J^-(\mathscr I^+),$$

and its event horizon is $\partial J^-(\mathscr I^+)$. A singular approach confined to $\mathcal B$ cannot signal that $\mathscr I^+$; one in $J^-(\mathscr I^+)$ may be globally visible. This definition is global and teleological: an event horizon cannot generally be located from one time slice.

An apparent horizon or marginally trapped tube is instead foliation-dependent and quasi-local. Its absence on a selected slice does not prove global nakedness, and its presence does not by itself identify the final event horizon. A trapped surface strongly constrains future evolution under energy and causality hypotheses, but it is not the definition of censorship.

A Cauchy horizon is different again. It bounds a domain of dependence. The inner horizons of exact Reissner–Nordström and Kerr spacetimes lie inside black holes and motivate strong cosmic censorship. They may permit extensions of the MGHD while exterior singular behavior remains hidden from $\mathscr I^+$. Thus “has a Cauchy horizon” and “has a naked singularity visible from infinity” answer different questions.

Examples and nonexamples

Negative-mass Schwarzschild. The curvature singularity at $r=0$ is timelike and visible in the exact asymptotically flat solution. But the solution is not the demonstrated outcome of generic regular collapse data of the class used in weak censorship. It tests the visibility definition, not the generic collapse conjecture.

Overextremal Reissner–Nordström or Kerr. With parameters beyond the black-hole bound, the exact metric has no event horizon shielding its central pathology. Again, writing down an exact exterior is not the same as evolving admissible complete data through a regular collapse. Charge, angular momentum, matter model, and initial topology must be included before calling it a counterexample.

Subextremal Reissner–Nordström or Kerr. Their inner Cauchy horizons are not visible from the chosen $\mathscr I^+$. They are central examples for extension-dependent strong censorship, not naked exterior singularities in the weak sense.

Spherical massless-scalar collapse. Christodoulou constructed globally naked outcomes from regular asymptotically flat spherical data and later proved a precise instability statement. In the space $BV\cap L^1$ of his reduced initial function, the exceptional set left after the trapped-surface criteria has “positive codimension” in the specific sense that through each exceptional datum passes a two-dimensional affine plane meeting the exceptional set only there Christodoulou 1999, Theorem 4.1, pp. 214–216. This is neither an unrestricted theorem nor automatically an open-dense statement in a smooth-data topology.

The 2026 localized Hölder result. Zheng studies the same one-parameter family of continuously self-similar spherical Einstein–scalar naked singularities, whose characteristic data have localized regularity $C^{1,k^2/(1-k^2)}$ at the sphere where the initial cone meets the past light cone of the singularity and are smooth away from it. For sufficiently small nonzero $k$, small perturbations in the explicitly defined localized Hölder space $\overline C_N^{p_k,\delta}$, with $p_k=(1-k^2)^{-1}$, $N\ge6$, and $0<\delta\le\delta_0$, generate asymptotically flat MGHDs with no trapped surface and incomplete $\mathscr I^+$ Zheng 2026, Theorem 3.1, pp. 23–25. The norm controls the $C^{1,p_k-1}$-type singular behavior only at the distinguished sphere, requires smoothness away from it, and includes weighted decay at infinity Zheng 2026, Definitions 2.7 and 2.11, pp. 18–20.

That 2026 preprint supplies nonlinear stability in a spherically symmetric, characteristic, localized Hölder topology, for finite-regularity backgrounds and $0<k^2\ll1$. Its $N\ge6$ perturbations have controlled $C^N$ behavior away from the distinguished sphere, not unrestricted smoothness. It does not establish stability under nonspherical perturbations, does not cover arbitrary smooth collapse data, and does not settle every intermediate topology. It sharpens the formulation question rather than producing a universal weak-censorship verdict.

A definition checklist

Before accepting “a naked singularity forms,” ask:

1. What data are evolved, in which matter model and symmetry class?
2. Is the data surface complete, regular, constraint satisfying, and of which asymptotic type?
3. What is singular: an incomplete geodesic, incomplete $\mathscr I^+$, curvature blowup, or inextendibility at a named regularity?
4. Is visibility local, visible to a finite observer, visible from a named conformal-boundary component, or visible from asymptotically flat $\mathscr I^+$?
5. Does the claim concern one solution, an open set, an open dense set, a residual set, a measure, or a codimension statement?

The evidence status on this page is fixed at 2026-08-12. Later theorems, corrections, or withdrawals are not incorporated.

Formulation diagnostic

Problem. Consider three proposed examples: (A) a flat region made incomplete by deleting one regular point from Minkowski spacetime; (B) an asymptotically flat collapse MGHD with a past-incomplete outgoing null geodesic from a singular boundary that reaches regular points but later falls into a black hole; (C) an asymptotically flat collapse MGHD whose $\mathscr I^+$ is future incomplete. Classify what can be concluded about nakedness and identify which example fails to represent formation from complete regular Cauchy data.

Solution

(A) is an artificial deletion, not singularity formation in the MGHD of complete regular Minkowski data; that MGHD is full Minkowski spacetime. (B) has local visibility: the past-incomplete end emits causal influence into a regular region. The information given does not establish global visibility from $\mathscr I^+$. (C) satisfies the future-null-infinity definition of global nakedness used in the cited spherical Einstein–scalar setting, provided its hypotheses—complete asymptotically flat data and the stated regularity—hold. None of the three statements alone specifies genericity or the regularity of possible extensions.

With the receiver now fixed, continue to Weak Cosmic Censorship.

References

• Christodoulou, Demetrios. “The Instability of Naked Singularities in the Gravitational Collapse of a Scalar Field.” Annals of Mathematics 149 (1999): 183–217. arXiv:math/9901147.
• Geroch, Robert P. “What Is a Singularity in General Relativity?” Annals of Physics 48 (1968): 526–540. https://doi.org/10.1016/0003-4916(68)90144-9.
• Isenberg, James. “On Strong Cosmic Censorship.” Surveys in Differential Geometry 20 (2015): 17–36. arXiv:1505.06390.
• Zheng, Weihao. “Nonlinear Stability of Continuously Self-Similar Naked Singularities for the Einstein-Scalar Field Equations I: Main Results.” Preprint, 15 May 2026. arXiv:2605.16235.