Naked Singularities

A spacetime singularity is a place where geodesics run off the edge of the manifold and curvature invariants typically blow up — but nothing in that description says it must be visible. In the Schwarzschild and Kerr black holes the singularity is wrapped in an event horizon, the boundary $\partial J^{-}(\mathscr{I}^{+})$ of the causal past of future null infinity, and no signal from the diverging-curvature region ever crosses it to reach a distant telescope. A naked singularity is the exact negation: a singular point or set that lies inside $J^{-}(\mathscr{I}^{+})$ , so that light rays leaving the neighbourhood of infinite curvature can climb all the way out to $\mathscr{I}^{+}$ and be seen. The stakes are the deterministic character of general relativity itself. Penrose framed the cosmic censorship hypothesis to protect it: if the singularities where the field equations fail are always swept behind horizons, the outside universe evolves smoothly from initial data. The weak form is about visibility from infinity; the strong form — which despite the name does not imply the weak one — is about the inextendibility and global hyperbolicity of the maximal development of generic data. Both remain conjectures, among the deepest open problems in mathematical relativity. Deciding whether a signal can escape is a question in global causal geometry, built from the causal operators $J^{\pm}$ , conformal infinity $\mathscr{I}^{\pm}$ , achronal sets, Cauchy surfaces, and the Penrose–Carter diagrams that render all of this finite. Candidate naked singularities come in three flavours: exact solutions tuned past their black-hole bound (over-spinning Kerr, over-charged Reissner–Nordström); idealized inhomogeneous dust collapse that focuses to a momentarily visible centre; and critical collapse at the black-hole formation threshold, where a naked singularity appears on a measure-zero surface in data space. Whether any generic, physically reasonable example exists is unknown.

Key takeaways

A naked singularity is one that is causally connected to distant observers: the singular set intersects the causal past of future null infinity, $J^{-}(\mathscr{I}^{+})$ , so signals from arbitrarily high curvature can in principle escape — unlike a singularity hidden behind an event horizon.
The event horizon is precisely the boundary $\mathcal{H}=\partial J^{-}(\mathscr{I}^{+})$ ; "no naked singularity" means exactly that this boundary exists and encloses the singular set inside the black-hole region $B=M\setminus J^{-}(\mathscr{I}^{+})$ .
Weak cosmic censorship (Penrose, 1969) conjectures that generic asymptotically flat collapse from regular data produces only horizon-cloaked singularities; naked singularities are the conjectured exceptions whose generic occurrence would falsify it.
Known counterexamples appear fine-tuned or unphysical: over-extremal Kerr–Newman ( $a^{2}+Q^{2}>M^{2}$ ), self-similar inhomogeneous dust (Christodoulou, Eardley–Smarr), and Choptuik critical collapse at the threshold — the word generic is what lets censorship survive them.
Weak censorship (nothing visible from $\mathscr{I}^{+}$ ) and strong censorship (global hyperbolicity of the maximal Cauchy development, nothing visible to any observer) are logically independent statements — the Cauchy horizon inside Reissner–Nordström and Kerr is a strong-censorship problem, not a weak one.
A naked singularity destroys global hyperbolicity: unconstrained signals emerge carrying no antecedent data, so no Cauchy surface covers the future and classical prediction fails.
Key idea: a singularity is naked if and only if it lies in the causal past of future null infinity, $J^{-}(\mathscr{I}^{+})$ ; cosmic censorship is the conjecture that generic collapse never produces this, so the event horizon $\partial J^{-}(\mathscr{I}^{+})$ always cloaks the singular region from distant observers.

Where this lives today

Cosmic censorship is a living research programme, probed from several directions at once. The over-spinning route — Wald's 1974 gedanken experiment of dropping a spinning particle into a near-extremal Kerr hole to push it past $a=M$ — has been refined for fifty years: treatments including the gravitational self-force and radiative back-reaction (Hubeny, Jacobson–Sotiriou, and successors) find that finely tuned processes can momentarily overshoot the bound, yet higher-order effects appear to restore the horizon. On the mathematical front, Christodoulou's proof for the spherically symmetric scalar field established that naked singularities, while they exist, form only on a non-generic set — precisely the statement the conjecture needs — while the analogous problem for the full Einstein equations without symmetry remains open. Observationally, the Event Horizon Telescope images of M87\ and Sgr A\ and the ringdown spectra of LIGO–Virgo–KAGRA mergers are consistent with Kerr horizons and constrain horizonless "naked" alternatives that would show altered shadows, extra photon rings, or high-energy leakage from the near-singular region. And the strong-censorship version has turned quantitative through the blue-shift instability of Reissner–Nordström and Kerr Cauchy horizons — recent near-extremal, cosmological-constant cases even show parameter ranges where strong censorship is marginally violated, keeping the subject unsettled at the frontier where classical relativity hands over to quantum gravity.

The physical picture

Picture stellar collapse as the closing of a light-cone trap. As matter concentrates, the outgoing light cones tip inward; once they tip past vertical even a radially outgoing photon falls in, and that surface of no return is the horizon. A singularity is naked when, by contrast, the cones at the singular centre still tilt enough to let an outward ray race ahead of the trap and reach infinity. Cosmic censorship is the claim that nature always slams the trap shut first — that the horizon forms before, or together with, the singularity, never after it. The cleanest way to see the two outcomes side by side is on a conformal diagram, where every causal line is drawn at exactly $45°$ and the whole infinite spacetime is squeezed into a finite triangle.

Figure

Two conformal diagrams. Left, censored collapse: the spacelike singularity zigzag sits above the event horizon 𝓗, an interior ray hits it and is trapped, and no causal curve reaches future null infinity 𝓘⁺. Right, a naked singularity: the timelike singularity is in the causal past of 𝓘⁺, so an outgoing 45° null ray escapes to 𝓘⁺ with no horizon to stop it — the condition sing ∩ J⁻(𝓘⁺) ≠ ∅

The whole question is one of communication with the screen $\mathscr{I}^{+}$ on which distant astronomers watch the universe: does any null geodesic originate in the limit at the singular set and terminate on that screen? For an exact solution the test reduces to arithmetic. A charged (Reissner–Nordström) or spinning (Kerr) hole has horizons wherever the metric function $f(r)=1-2M/r+Q^{2}/r^{2}$ vanishes, and those roots are real only up to the extremal bound $Q^{2}\le M^{2}$ (or $a^{2}\le M^{2}$ ). Push the charge or spin past the bound and the roots go complex: the horizon vanishes, $f$ stays positive everywhere, and the curvature singularity at the centre is left exposed. The second figure is that test made visible — the lapse $f(r)$ for three charges straddling the bound.

Figure

The Reissner–Nordström lapse f(r) = 1 − 2M/r + Q²/r² in units M = 1 for three charges. At Q/M = 0.8 (blue) f dips below zero, giving two horizons r± = M ± √(M²−Q²) = 0.4M and 1.6M. At Q/M = 1 (amber, extremal) f just touches zero at r = M. At Q/M = 1.2 (red, over-charged) f stays positive with minimum f = 1 − M²/Q² = 0.31 at r = Q²/M = 1.44M — no horizon, the singularity is naked

Beneath the arithmetic sits the real worry: determinism. In a globally hyperbolic spacetime, data on one Cauchy surface fix the entire future through the field equations. A naked singularity behaves like an uncontrolled boundary in the middle of spacetime, radiating signals that were fixed by no initial data — a hole in predictability. And the word generic is what keeps the conjecture alive against its counterexamples: over-extremal Kerr resists being fed enough angular momentum to cross the bound, dust collapse ignores pressure, and critical collapse sits on a codimension-one surface in data space. Every known nakedness is fine-tuned, leaving room for censorship to hold for generic, physically reasonable matter.

The core equations

Everything begins with the operational definition. A singularity is naked exactly when its causal past reaches out to future null infinity, so that a distant observer sits in causal contact with the diverging-curvature region.

\text{Naked singularity}\iff(\text{singular set})\cap J^{-}(\mathscr{I}^{+})\neq\varnothing \qquad (1)

where:

$J^{-}(\mathscr{I}^{+})$ — causal past of future null infinity: every event from which some future-directed causal curve reaches $\mathscr{I}^{+}$ .
$\mathscr{I}^{+}$ — future null infinity, the endpoint of outgoing null rays in the conformal completion $(\bar M,\bar g)$ .

Reading: nakedness is a statement about causal curves alone, so it is conformally invariant and can be read straight off a Penrose–Carter diagram. The complement of that causal past is the region from which nothing escapes; its boundary is the event horizon. This one line is the geometric definition of a black hole, and the exact negation of nakedness.

B=M\setminus J^{-}(\mathscr{I}^{+}),\qquad \mathcal{H}=\partial J^{-}(\mathscr{I}^{+})\cap M=\partial B \qquad (2)

where:

$B$ — the black-hole region: events that cannot signal $\mathscr{I}^{+}$ .
$\mathcal{H}$ — the future event horizon, an achronal null hypersurface generated by null geodesics with no future endpoints.

Reading: a singularity in the interior of $B$ is censored; one touching $J^{-}(\mathscr{I}^{+})$ leaks light to $\mathscr{I}^{+}$ and is naked. The horizon is the membrane between the two. The canonical laboratory is the static charged (Reissner–Nordström) black hole, whose single metric function decides everything about horizons and nakedness.

\mathrm{d}s^{2}=-f(r)\,\mathrm{d}t^{2}+f(r)^{-1}\mathrm{d}r^{2}+r^{2}\mathrm{d}\Omega^{2},\qquad f(r)=1-\frac{2M}{r}+\frac{Q^{2}}{r^{2}} \qquad (3)

where:

$M$ — ADM mass; $Q$ — electric charge, both in geometrized units $G=c=1$ .
$\mathrm{d}\Omega^{2}=\mathrm{d}\theta^{2}+\sin^{2}\theta\,\mathrm{d}\varphi^{2}$ — the unit two-sphere metric.

Reading: the horizons live where $f(r)=0$ ; where $f>0$ everywhere the coordinate $r$ stays spacelike and no trapped region forms. Setting $f=0$ and clearing $r^{2}$ turns the horizon condition into a quadratic, and its discriminant is the entire content of cosmic censorship for this solution.

r^{2}-2Mr+Q^{2}=0\ \Rightarrow\ r_{\pm}=M\pm\sqrt{M^{2}-Q^{2}} \qquad (4)

where:

$r_{+},r_{-}$ — outer (event) and inner (Cauchy) horizon radii.
$M^{2}-Q^{2}$ — the discriminant, positive below the extremal bound and negative above it.

Reading: real roots require $Q^{2}\le M^{2}$ ; for $Q^{2}>M^{2}$ the roots are complex, no horizon exists, and the $r=0$ singularity is naked. For Kerr the identical statement holds with $Q\to a=J/M$ . That the centre is a genuine singularity, not a coordinate artifact, is settled by a curvature invariant. The Kretschmann scalar is coordinate-independent, so its behaviour cannot be gauged away.

R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}=\frac{48M^{2}}{r^{6}}\ \xrightarrow{\ r\to0\ }\ \infty \qquad (5)

where:

$R_{\mu\nu\rho\sigma}$ — the Riemann tensor (this is the Schwarzschild value; the charged case adds $Q$ -dependent terms with the same $r\to0$ divergence).
$r$ — the areal radius.

Reading: the divergence at $r=0$ marks true geodesic incompleteness; at $r=2M$ the scalar is finite, confirming the Schwarzschild horizon is a coordinate effect, not a singularity. Why the horizon hides what is behind it follows from the radial null condition — the slope of an outgoing light ray in the $(t,r)$ plane.

\mathrm{d}s^{2}=0\ \Rightarrow\ \frac{\mathrm{d}t}{\mathrm{d}r}\bigg|_{\text{out}}=\frac{1}{f(r)}\ \xrightarrow{\ r\to r_{+}\ }\ \infty \qquad (6)

where:

$\mathrm{d}t/\mathrm{d}r$ — coordinate slope of an outgoing radial null ray; $f(r)$ — the lapse, vanishing at $r_{+}$ .

Reading: as $f\to0$ the outgoing ray needs infinite coordinate time to climb to $r_{+}$ — it never reaches $\mathscr{I}^{+}$ . Remove the root (over-charge) and $f$ never vanishes, so the slope stays finite and rays from the centre do escape. Dynamical nakedness needs matter, not just an exact vacuum solution. Inhomogeneous pressureless dust — the Lemaître–Tolman–Bondi model — is the classic testbed for whether a forming singularity can be seen.

\mathrm{d}s^{2}=-\mathrm{d}t^{2}+\frac{{R'}^{2}}{1+2E(r)}\,\mathrm{d}r^{2}+R^{2}(t,r)\,\mathrm{d}\Omega^{2},\qquad \dot R^{2}=\frac{2m(r)}{R}+2E(r) \qquad (7)

where:

$R(t,r)$ — areal radius of the dust shell labelled $r$ ; $R'=\partial R/\partial r$ , $\dot R=\partial R/\partial t$ .
$m(r)$ — Misner–Sharp mass inside shell $r$ ; $E(r)$ — energy (binding) function fixing the shell profile.

Reading: a central shell-focusing singularity forms where $R\to0$ ; whether an outgoing null ray leaves it with a positive tangent — and so becomes visible — depends on the profiles $m(r),E(r)$ , i.e. on how centrally condensed the dust is. The third route is the black-hole formation threshold. Choptuik's numerical collapse of a scalar field shows the horizon mass obeying a universal power law that vanishes exactly at criticality.

M_{\rm BH}\propto|p-p_{*}|^{\gamma},\qquad \gamma\approx0.374,\qquad \lim_{p\to p_{*}}M_{\rm BH}=0 \qquad (8)

where:

$p$ — amplitude of a one-parameter family of initial data; $p_{*}$ — the critical value separating dispersal from collapse.
$\gamma$ — universal critical exponent, $\approx0.374$ for a massless scalar; the self-similar solution echoes with period $\Delta\approx3.44$ .

Reading: tuning $p\to p_{*}$ makes the horizon shrink to nothing while curvature diverges — a naked singularity, but only on the codimension-one critical surface, hence non-generic.

Derivation AOver-extremal Reissner-Nordström has no horizon so its singularity is naked

The goal is to prove, by locating the roots of the lapse and testing its sign, that pushing the charge past the mass bound $Q^{2}>M^{2}$ eliminates both horizons and leaves the $r=0$ curvature singularity in causal contact with $\mathscr{I}^{+}$ . We work in geometrized units $G=c=1$ with the static spherically symmetric charged metric, treat the horizon condition as a quadratic in $r$ , and complete the square to read off the sign of the lapse everywhere outside the centre.

Step 1. Write the lapse and its roots: $f(r)=1-2M/r+Q^{2}/r^{2}$ , and $f=0$ gives $r^{2}-2Mr+Q^{2}=0$ , so $r_{\pm}=M\pm\sqrt{M^{2}-Q^{2}}$ . Why: horizons are the null surfaces where $f$ vanishes — equivalently where $\nabla r$ becomes null — so the whole question is the reality of these two roots.

Step 2. Impose the over-extremal condition $Q^{2}>M^{2}$ , so the discriminant $M^{2}-Q^{2}<0$ . The square root is imaginary and $r_{\pm}$ are complex conjugates: there is no real radius anywhere at which $f$ vanishes. Why: a complex root is not a place in the spacetime, so there is simply no horizon to find.

Step 3. Complete the square in the numerator: $f(r)=\dfrac{r^{2}-2Mr+Q^{2}}{r^{2}}=\dfrac{(r-M)^{2}+(Q^{2}-M^{2})}{r^{2}}$ . Since $Q^{2}-M^{2}>0$ , the numerator is a sum of a square and a strictly positive constant, so $f(r)>0$ for every $r>0$ . Why: a manifestly positive lapse is the algebraic certificate that no trapped region exists, without having to solve any geodesic equation.

Step 4. With $f>0$ the coordinate $r$ is spacelike and $t$ timelike for all $r>0$ : the light cones never tip past vertical, so no outgoing ray is ever turned back. Why: trapping is exactly the tipping of cones that $f\le0$ would signal, and here it never happens.

Step 5. The curvature still diverges — $R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}\to\infty$ as $r\to0$ — but with no cloaking surface, outgoing radial null rays from near $r=0$ run out to $\mathscr{I}^{+}$ . Hence the singular set lies in $J^{-}(\mathscr{I}^{+})$ and, by definition (1), it is naked; being at fixed $r=0$ with $r$ spacelike, it is a timelike world-tube, not a moment of time. Why: this closes the argument by connecting the positive lapse directly to the operational definition of nakedness. *Verified: solving $r^{2}-2Mr+Q^{2}=0$ and checking the discriminant sign, and the completed-square positivity of $f$ , reproduce Wald §12.3 and the Reissner–Nordström conformal diagram in Hawking & Ellis §5.5.*

Go deeper · Derivation B · 2 minThe event horizon as the boundary of the causal past of infinity

Derivation BThe event horizon as the boundary of the causal past of infinity

The goal is to show that the existence of an event horizon is precisely the statement that the singular set is excluded from the causal past of $\mathscr{I}^{+}$ , so that "no naked singularity" and "a horizon exists that hides it" are the same proposition. We work on the conformally completed asymptotically flat spacetime $(\bar M,\bar g)$ with boundary $\mathscr{I}^{\pm}$ , using only the global causal operators $J^{\pm}$ , and build the black-hole region, its boundary, and the predictability of the exterior in turn.

Step 1. Define the black-hole region $B=M\setminus J^{-}(\mathscr{I}^{+})$ — the events that cannot send any causal signal to future null infinity. Why: this makes "region of no escape" a precise set rather than a metaphor, defined purely by causal reachability.

Step 2. Take its boundary $\mathcal{H}=\partial B=\partial J^{-}(\mathscr{I}^{+})\cap M$ . Standard results (achronality, the null generators lemma) show $\mathcal{H}$ is an achronal null hypersurface generated by null geodesics that have past endpoints but no future endpoints — the outgoing photons that asymptote to the horizon yet never cross it. Why: identifying $\mathcal{H}$ as a null hypersurface is what makes it a one-way membrane, the geometric content of "horizon".

Step 3. Characterize nakedness causally: a singular endpoint $s$ is naked iff $s\in\overline{J^{-}(\mathscr{I}^{+})}$ , i.e. some sequence of causal curves from neighbourhoods of $s$ reaches $\mathscr{I}^{+}$ ; equivalently $s\notin\mathrm{int}\,B$ . Why: this ties the abstract definition (1) to the concrete question of whether curves from $s$ can climb out.

Step 4. State weak censorship in these terms: for generic regular data the singular set lies in $\mathrm{int}\,B$ , entirely inside the horizon. The horizon's nonempty existence is then the geometric guarantee of hiddenness. Why: this is what makes the horizon, not some separate object, the precise witness that the conjecture holds.

Step 5. Draw the predictability payoff: if the singularity is inside $B$ , the exterior $M\setminus B$ is globally hyperbolic with $\mathscr{I}^{+}$ as its future boundary, so distant physics is fully fixed by Cauchy data. A naked singularity breaches this, injecting signals no data constrained. Why: this is why censorship matters — it is exactly the hypothesis that keeps the exterior deterministic. *Verified: the definitions of $B$ and $\mathcal{H}$ and the null-generator property follow Wald §12.1 and Hawking & Ellis §9.2–9.3.*

Go deeper · Derivation C · 2 minWhen an outgoing ray can escape a collapsing singularity

Derivation CWhen an outgoing ray can escape a collapsing singularity

The goal is to set up the nakedness test for a dynamically forming central singularity and show it reduces to an algebraic condition: a naked singularity exists iff the outgoing radial null geodesic equation admits a positive real root for the limiting tangent as the singularity is approached. We work in the Lemaître–Tolman–Bondi dust model, expand the areal radius near the centre, and extract the root condition that separates a black hole from at least a locally naked singularity.

Step 1. Write the outgoing radial null condition from metric (7): $\mathrm{d}t/\mathrm{d}r=R'(t,r)/\sqrt{1+2E(r)}$ , an ODE for the ray $t(r)$ whose right side is set by the dust profiles $m(r),E(r)$ . Why: nakedness is decided by the behaviour of this one first-order equation as it approaches the singular centre.

Step 2. Locate the central shell-focusing singularity where the areal radius $R(t,r)\to0$ as $r\to0$ ; the marginally bound case $E=0$ gives $R=(9m/2)^{1/3}(t_{s}(r)-t)^{2/3}$ with singularity time $t_{s}(r)$ set by the enclosed mass. Why: an explicit near-centre form for $R$ is what lets the ray equation be solved in a self-similar limit.

Step 3. Change to the self-similar variable and ask for the limiting tangent $X_{0}=\lim_{r\to0,\,R\to0}R/r$ of an outgoing null ray with a past endpoint at the singularity. Substituting the expansion turns the null ODE into an algebraic equation for $X_{0}$ whose lowest coefficient is fixed by the leading density inhomogeneity. Why: reducing a differential to an algebraic condition is what makes the test computable for a given profile.

Step 4. Apply the root criterion: if that algebraic equation has a positive real root $X_{0}>0$ , an outgoing null geodesic emanates from the singularity into its future — the singularity is at least locally naked. If no positive root exists, every outgoing ray is swallowed and the singularity is covered; homogeneous Oppenheimer–Snyder dust falls in this class and forms a black hole. Why: the sign of a root, controlled by how centrally condensed the dust is, is the entire dividing line.

Step 5. Distinguish local from global nakedness: a locally naked singularity emits into a finite region; it is globally naked — the case that bears on weak censorship — only if such a ray actually reaches $\mathscr{I}^{+}$ . Christodoulou showed sufficiently centrally condensed dust is globally naked, but the set of such data is non-generic. Why: weak censorship is about $\mathscr{I}^{+}$ , so only global nakedness can falsify it, and only if it is generic. *Verified: the root condition for $X_{0}$ and the failure of homogeneous dust to produce a positive root follow Joshi, Global Aspects in Gravitation and Cosmology, Ch. 6, and Christodoulou, Commun. Math. Phys. 93, 171 (1984).*

Limiting cases

How the horizon condition and nakedness depend on the parameters of the canonical exact solutions:

Case	Result	Meaning
Schwarzschild ( $Q=a=0$ )	$r_{+}=2M$ , single root	One horizon for any $M>0$ ; the $r=0$ singularity is always hidden — never naked.
Reissner–Nordström	$r_{\pm}=M\pm\sqrt{M^{2}-Q^{2}}$	Horizons exist iff $Q^{2}\le M^{2}$ ; over-charged $Q^{2}>M^{2}$ gives no real root and a naked timelike singularity.
Kerr	$r_{\pm}=M\pm\sqrt{M^{2}-a^{2}}$	Horizons exist iff $a\le M$ ( $a=J/M$ ); over-spinning $a>M$ exposes the ring singularity.
Kerr–Newman	$r_{\pm}=M\pm\sqrt{M^{2}-a^{2}-Q^{2}}$	The general bound $a^{2}+Q^{2}\le M^{2}$ ; violate it and the singularity is naked.
Extremal limit	$M^{2}=a^{2}+Q^{2}$	Degenerate horizon $r_{+}=r_{-}=M$ ; the razor's edge between black hole and naked singularity, with zero Hawking temperature.
Choptuik threshold	$M_{\rm BH}\to0$ at $p=p_{*}$	A naked singularity on a codimension-one surface in data space — non-generic, consistent with censorship.

Common trapWeak and strong cosmic censorship are different conjectures — do not conflate them.

The two statements protect different observers and neither implies the other. Weak cosmic censorship forbids singularities visible from $\mathscr{I}^{+}$ : it protects observers at infinity, and its natural setting is asymptotically flat collapse with an event horizon $\partial J^{-}(\mathscr{I}^{+})$ . Strong cosmic censorship is a statement about determinism everywhere: the maximal Cauchy development of generic initial data should be inextendible, i.e. globally hyperbolic, so that no singularity is visible even to an observer who falls through a horizon. The classic trap is the Cauchy horizon inside Reissner–Nordström and Kerr — the inner horizon $r_{-}$ , beyond which the metric is no longer fixed by the data. That is a strong-censorship problem (about the fate of an infalling observer's predictability), not a weak one: the exterior of those black holes is perfectly well behaved and nothing there is naked. A second trap is forgetting the word generic; without it every fine-tuned counterexample would already falsify the conjecture.

Worked examples

1 · The electron as a would-be naked singularity (fully worked)

Two questions. (a) For a toy Reissner–Nordström object with charge-to-mass ratio $Q/M=1.2$ , show there is no horizon and locate the closest approach of the lapse to zero. (b) If the electron's actual charge and mass are inserted into the Reissner–Nordström bound, is the resulting geometry a black hole or a naked singularity?

Step 1. Test the toy object against the bound. The discriminant is $M^{2}-Q^{2}=M^{2}(1-1.44)=-0.44M^{2}<0$ , so $r_{\pm}=M\pm\sqrt{-0.44}\,M$ are complex — no real horizon exists. The horizon quadratic has no solution in the spacetime.

Step 2. Find where the lapse comes closest to zero. Minimizing $f(r)=1-2M/r+Q^{2}/r^{2}$ gives $f'(r)=2M/r^{2}-2Q^{2}/r^{3}=0$ , so $r_{\min}=Q^{2}/M=1.44M$ , and $f(r_{\min})=1-M^{2}/Q^{2}=1-1/1.44=\mathbf{0.31}$ . The lapse dips to $0.31$ and climbs back — strictly positive everywhere, so the $r=0$ singularity is naked.

Step 3. Now the real electron. The Reissner–Nordström bound in geometrized units is $Q^{2}\le M^{2}$ , which restored to SI reads $\dfrac{Q^{2}}{M^{2}}=\dfrac{q^{2}}{4\pi\varepsilon_{0}G\,m^{2}}$ — the ratio of electric to gravitational coupling. Insert $q=1.60\times10^{-19}\,\mathrm{C}$ , $m=9.11\times10^{-31}\,\mathrm{kg}$ , $G=6.67\times10^{-11}$ , $4\pi\varepsilon_{0}=1.11\times10^{-10}$ .

Step 4. Evaluate: $\dfrac{Q^{2}}{M^{2}}=\dfrac{(1.60\times10^{-19})^{2}}{(1.11\times10^{-10})(6.67\times10^{-11})(9.11\times10^{-31})^{2}}\approx\mathbf{4.2\times10^{42}}$ , so $Q/M\approx\mathbf{2.0\times10^{21}}$ . The electron overshoots the extremal bound by twenty-one orders of magnitude: its would-be Reissner–Nordström geometry is wildly naked, not a black hole. This is the standard reason the electron cannot be modelled as a Reissner–Nordström black hole — and a vivid illustration that nakedness is generic for fundamental charged particles yet apparently forbidden for macroscopic collapse, which is exactly the tension cosmic censorship lives in.

2 · Critical collapse and the vanishing horizon (fully worked)

In Choptuik collapse of a massless scalar field the black-hole mass obeys $M_{\rm BH}=C\,|p-p_{*}|^{\gamma}$ with $\gamma\approx0.374$ . (a) By what factor does the horizon mass shrink when the initial data are tuned from $|p-p_{*}|/p_{*}=10^{-3}$ to $10^{-6}$ ? (b) What happens exactly at $p=p_{*}$ , and why is it consistent with cosmic censorship?

Step 1. Take the ratio of the scaling law at the two tunings so the unknown constant $C$ cancels: $\dfrac{M_{\rm BH}(10^{-6})}{M_{\rm BH}(10^{-3})}=\left(\dfrac{10^{-6}}{10^{-3}}\right)^{\gamma}=(10^{-3})^{0.374}$ .

Step 2. Evaluate the exponent: $(10^{-3})^{0.374}=10^{-1.122}=\mathbf{0.076}$ . Tuning three decades closer to threshold makes the black hole about 13 times lighter — you can form an arbitrarily small horizon by tuning arbitrarily close.

Step 3. Take the limit. As $p\to p_{*}$ , $M_{\rm BH}=C\,|p-p_{*}|^{0.374}\to\mathbf{0}$ : the horizon radius shrinks to zero while the central curvature diverges. Exactly at $p=p_{*}$ there is a zero-mass singularity with no horizon — a naked singularity. Along the way the self-similar solution echoes, repeating on spatial scales that shrink by $e^{\Delta}\approx e^{3.44}\approx\mathbf{31}$ per echo, so tuning closer reveals ever more echoes before a (tiny) horizon forms.

Step 4. Reconcile with censorship. The naked singularity appears only at $p=p_{*}$ , a single point in the one-parameter family — a codimension-one surface in the full infinite-dimensional space of initial data. It is a measure-zero, non-generic configuration, so it does not violate weak cosmic censorship as stated, which speaks only of generic data. This example is the sharpest demonstration of why the word "generic" is indispensable: without it, Choptuik's threshold alone would already refute the conjecture.

Physics of the portrait

Naked singularities exist as a subject because one physicist, having just helped prove that singularities are unavoidable in collapse, immediately asked whether they could be seen — and turned the question of their visibility into the deepest open conjecture in classical relativity.

Origins · 1 minRoger Penrose and how this was discovered

Roger Penrose

1931– · singularity theorems and cosmic censorship

Roger Penrose reshaped general relativity by importing global, topological reasoning into a field that had been dominated by local differential geometry. His 1965 singularity theorem introduced the notion of a trapped surface and, with it, a proof that gravitational collapse generically produces geodesic incompleteness — a singularity — independent of any symmetry assumption. That result, later sharpened with Stephen Hawking into the Hawking–Penrose theorems, made singularities a robust prediction of the theory rather than an artifact of idealized models, and it won him a share of the 2020 Nobel Prize. Having established that singularities form, Penrose posed in 1969 the question that defines this article: are they always hidden? His cosmic censorship hypothesis — that generic collapse cloaks singularities behind event horizons — was offered candidly as a conjecture, and he spent decades probing it, formulating the Penrose inequality (a necessary condition relating horizon area to mass) as a concrete test. His toolkit is everywhere in the subject: conformal (Penrose–Carter) diagrams that compress infinite spacetimes into finite pictures, the conformal boundary $\mathscr{I}$ , spinor and twistor methods, and the causal-structure machinery of $J^{\pm}$ that makes the very words "naked" and "hidden" precise. Characteristically, he framed a physical worry — can the breakdown of the field equations infect the outside world? — as an exact geometric statement about $J^{-}(\mathscr{I}^{+})$ , and then left it standing as a challenge that remains unmet.

History

1939

— Oppenheimer and Snyder model homogeneous dust collapse to a black hole with a hidden singularity

1965

— Penrose proves the first singularity theorem using trapped surfaces

1969

— Penrose states the cosmic censorship hypothesis

1973

— Hawking and Ellis codify the global causal machinery of horizons and singularities

1979

— Eardley and Smarr find naked singularities in inhomogeneous dust collapse

1984

— Christodoulou proves cosmic censorship is violated in self-similar dust collapse

1993

— Choptuik discovers critical collapse and the naked singularity at threshold

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), §34 and Box 34.2 · Wald, General Relativity (1984), Ch. 12, §12.1–12.3 · Hawking & Ellis, The Large Scale Structure of Space-Time (1973), Ch. 8–9 · Carroll, Spacetime and Geometry (2004), Ch. 5–6 · Hartle, Gravity (2003), Ch. 12 · Schutz, A First Course in General Relativity (2nd ed., 2009), Ch. 11 · Penrose, "Gravitational collapse: the role of general relativity", Riv. Nuovo Cimento 1, 252 (1969) · Christodoulou, Commun. Math. Phys. 93, 171 (1984) · Choptuik, Phys. Rev. Lett. 70, 9 (1993) · Joshi, Global Aspects in Gravitation and Cosmology (1993), Ch. 6.

Provenance. The definitions of $J^{\pm}$ , $\mathscr{I}^{\pm}$ , the black-hole region $B$ , and the event horizon $\mathcal{H}=\partial J^{-}(\mathscr{I}^{+})$ follow Wald §12.1 and Hawking & Ellis §9.2–9.3. The Reissner–Nordström and Kerr horizon roots ( $r_{\pm}=M\pm\sqrt{M^{2}-Q^{2}}$ , $r_{\pm}=M\pm\sqrt{M^{2}-a^{2}}$ ) and the Schwarzschild Kretschmann scalar $48M^{2}/r^{6}$ are standard textbook results re-derived from the metrics in geometrized units. Penrose (1969) introduces cosmic censorship; Eardley–Smarr and Christodoulou (1984) establish naked dust singularities; Choptuik (1993) gives $\gamma\approx0.374$ and the echo period $\Delta\approx3.44$ for a massless scalar. A verifier should: (1) solve $r^{2}-2Mr+Q^{2}=0$ and check the discriminant sign; (2) confirm $f(r)=[(r-M)^{2}+(Q^{2}-M^{2})]/r^{2}>0$ for $Q^{2}>M^{2}$ , with $f_{\min}=1-M^{2}/Q^{2}=0.31$ at $r=Q^{2}/M=1.44M$ for $Q/M=1.2$ ; (3) verify the electron ratio $Q^{2}/M^{2}=q^{2}/(4\pi\varepsilon_{0}Gm^{2})\approx4.2\times10^{42}$ , so $Q/M\approx2.0\times10^{21}$ ; (4) confirm the critical scaling factor $(10^{-3})^{0.374}\approx0.076$ and $e^{3.44}\approx31$ ; (5) cross-check the weak/strong censorship distinction against Wald §12.1 and Hawking & Ellis Ch. 9.

Prerequisites. Causal Structure of Spacetime · Chronological and Causal Futures and Pasts · Conformal Infinity and Asymptotic Structure · Global Hyperbolicity · Null Geodesics and Light Cones. Leads to. Domain of Dependence and Cauchy Horizons · Penrose–Carter Conformal Diagrams · Cauchy Surfaces.

Ask the tutor: Want to walk through Christodoulou's dust nakedness criterion in detail, or see exactly why Wald's spinning-particle gedanken fails to over-spin a near-extremal Kerr hole?

What exactly makes a singularity "naked" rather than hidden?

It is a purely causal criterion, and it has nothing to do with how strong the curvature is. A singularity is naked when its causal past reaches future null infinity — formally, when the singular set intersects $J^{-}(\mathscr{I}^{+})$ — so that at least one null ray from the neighbourhood of diverging curvature can climb all the way out to a distant observer. A hidden singularity is one buried in the black-hole region $B=M\setminus J^{-}(\mathscr{I}^{+})$ , from which no causal signal escapes. The event horizon $\partial J^{-}(\mathscr{I}^{+})$ is exactly the boundary between these two fates, which is why "no naked singularity" is the same statement as "an event horizon cloaks it".

Does an over-charged or over-spinning black hole really exist in nature?

Almost certainly not as an endpoint of collapse, and that is the whole point of cosmic censorship. Over-extremal Reissner–Nordström ( $Q^{2}>M^{2}$ ) and over-spinning Kerr ( $a>M$ ) are perfectly good solutions of the Einstein equations with naked singularities, but every attempt to create one dynamically seems to fail: a particle carrying enough charge or angular momentum to push a near-extremal hole past the bound is electrostatically or centrifugally repelled and cannot be captured (Wald's 1974 argument), and back-reaction and radiation appear to close the remaining loopholes. So the over-extremal metrics exist on paper, but nature resists assembling them — which is the empirical shadow of censorship.

How is a naked singularity different from the Cauchy horizon inside a charged black hole?

They belong to the two different censorship conjectures. A naked singularity visible from $\mathscr{I}^{+}$ would violate weak censorship — it is seen by observers who never fall in. The Cauchy horizon $r_{-}$ inside Reissner–Nordström or Kerr is invisible from infinity; the exterior is completely normal. It matters only to an observer who has already crossed the event horizon: beyond $r_{-}$ the spacetime is no longer determined by the initial data, so prediction fails for that infalling observer. That is a strong-censorship issue about global hyperbolicity of the maximal development, not about anything a distant telescope could see. Conflating the two is the single most common error in the subject.

Why does a naked singularity threaten determinism?

Because it acts as a boundary in the middle of spacetime that emits data no initial surface controls. In a globally hyperbolic spacetime, a Cauchy surface $\Sigma$ plus the field equations fix the entire future. A naked singularity lies to the future of $\Sigma$ (or radiates into regions to its future) while being a place where the field equations break down, so signals can emerge from it that were fixed by no data on $\Sigma$ . A Cauchy horizon forms, the maximal development stops being globally hyperbolic, and the future becomes underdetermined. Cosmic censorship is, at bottom, the hypothesis that classical general relativity stays a deterministic theory — that this never happens for generic collapse.

Is the singularity at the centre of a naked case timelike or spacelike, and does it matter?

It matters a great deal, and the two cases look completely different on a conformal diagram. The Schwarzschild singularity is spacelike: it is a moment of time in the future of everything inside the horizon, like a final instant no one can avoid, and it is drawn as a horizontal line above the horizon. The over-charged Reissner–Nordström singularity is timelike: with the lapse $f>0$ everywhere, $r=0$ is a world-tube persisting through time, drawn as a vertical line, and you can in principle steer around it or receive signals from it. Nakedness is closely tied to this: a timelike singularity sitting in $J^{-}(\mathscr{I}^{+})$ can send rays to infinity throughout its history, which a spacelike one buried behind a horizon cannot.

Could we ever observe a naked singularity, and would it help physics?

If one existed astrophysically it would be spectacular and diagnostic: lacking a true event horizon, it might cast a different shadow, show extra photon rings, or leak high-energy radiation from near the exposed high-curvature region, and current Event Horizon Telescope and gravitational-wave data already constrain such horizonless alternatives — so far consistent with ordinary Kerr horizons. Ironically, an observed naked singularity would be a gift: it is exactly at the singularity that classical general relativity fails and quantum gravity must take over, so a visible one would be a window onto Planck-scale physics that a hidden singularity forever denies us — part of why proving or refuting censorship carries such weight.

Naked Singularities

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · The electron as a would-be naked singularity (fully worked)

2 · Critical collapse and the vanishing horizon (fully worked)

Physics of the portrait

Going further

Before this

Where this leads