Strong Cosmic Censorship

General relativity keeps a strict promise: hand it constraint-satisfying data on a spacelike slice and the Einstein equations return a unique maximal globally hyperbolic development, the largest spacetime that slice can determine. The theorem behind that promise — Choquet-Bruhat and Geroch, 1969 — is what makes the theory deterministic. The trouble is that the tidiest solutions we know break the promise on paper. Reissner-Nordström and Kerr both possess a maximal analytic extension strictly larger than their globally hyperbolic core, separated by a Cauchy horizon at the inner radius $r_-$ . Cross it and the metric is no longer fixed by the data: infinitely many smooth continuations fit the same past, and prediction dies. Strong Cosmic Censorship, which Penrose first raised in the late 1960s and gave its modern shape in 1979, conjectures that this failure is a measure-zero accident of exact symmetry. For generic initial data the maximal development should be inextendible across any Cauchy horizon, so the physical spacetime stays globally hyperbolic and determinism survives. The word generic is load-bearing: nobody claims Cauchy horizons never form, only that they are unstable — a perturbation destroys them. The conjecture lives where global Lorentzian geometry meets the analysis of Einstein's equations as quasilinear hyperbolic PDEs. The physical engine is the blue-shift at inner horizons. An observer crossing a Cauchy horizon receives, in finite proper time, the entire infinite future history of the outside universe, piled up and exponentially frequency-shifted by $e^{\kappa_- v}$ . The energy flux diverges, the local mass inflates, and curvature becomes singular — yet often only weakly, the metric extending continuously ( $C^0$ ) even as its connection tears. That subtlety forced a sharpening: Christodoulou rephrased censorship as inextendibility with locally square-integrable Christoffel symbols, and the modern verdict rides on a single ratio $\beta=\alpha/\kappa_-$ crossing $\tfrac12$ . He proved the conjecture for scalar-field collapse; near-extremal de Sitter black holes violate it — a tension that keeps SCC at the live frontier of mathematical relativity.

Key takeaways

Strong Cosmic Censorship (SCC) conjectures that for generic initial data the maximal globally hyperbolic development of the Einstein equations is inextendible as a suitably regular spacetime — so black-hole interiors and the whole universe stay determined by the data.
Operationally, the smooth Cauchy horizons of Reissner-Nordström and Kerr at $r_-=M-\sqrt{M^2-Q^2}$ are conjectured to be non-generic: an arbitrarily small perturbation should destroy the extension across them and restore global hyperbolicity.
The mechanism is the blue-shift instability — infalling radiation crossing the inner horizon is amplified by $e^{\kappa_- v}$ , producing the mass-inflation singularity where the renormalized Hawking mass diverges.
Determinism must be phrased against a regularity class: the metric can extend continuously ( $C^0$ ) even when SCC holds, so Christodoulou pins it to a non- $H^1_{\mathrm{loc}}$ connection — the threshold below which the weak Einstein equations lose meaning.
The verdict is a race of exponentials captured by $\beta=\alpha/\kappa_-$ : Christodoulou-SCC holds when $\beta\le\tfrac12$ and is violated when $\beta>\tfrac12$ , which is why a positive cosmological constant and near-extremality can flip the answer.
SCC is logically independent of Weak Cosmic Censorship: WCC hides singularities behind horizons (visibility from infinity), SCC forbids Cauchy horizons (determinism); neither implies the other.
Key idea: determinism in general relativity is the inextendibility of the maximal globally hyperbolic development. Cauchy horizons in idealized symmetric solutions threaten it, but the blue-shift instability is conjectured to convert them into genuine weak null singularities for generic data — so the boundary of predictability becomes the boundary of spacetime itself.

The core insight

Determinism in general relativity is the statement that the maximal globally hyperbolic development is inextendible. Cauchy horizons in idealized symmetric solutions threaten this, but the blue-shift instability is conjectured to convert them into genuine weak null singularities for generic data — so the boundary of predictability becomes the boundary of spacetime itself.

Where this lives today

Strong Cosmic Censorship is one of the most active problems in mathematical relativity, its center of gravity shifting from heuristics to theorems. On the "holds" side stands Christodoulou's proof, for the spherically symmetric Einstein-scalar-field system, that generic data either disperse or collapse to a hole whose terminal boundary is a spacelike singularity with inextendible ( $H^1$ ) connection — the strongest evidence that censorship is true for dynamical, asymptotically flat collapse. On the interior, Dafermos and Luk proved the $C^0$ -stability of the Kerr Cauchy horizon (2017): a perturbed rotating hole's metric does extend continuously across $r_-$ , so the singularity there is weak and the debate moves onto the finer $H^1_{\mathrm{loc}}$ threshold. The sharpest twist is on the "fails" side — the 2018 discovery by Cardoso, Costa, Destounis, Hintz and Jansen that near-extremal Reissner-Nordström-de Sitter holes admit $H^1_{\mathrm{loc}}$ extensions across $r_-$ , because a positive $\Lambda$ speeds quasinormal decay relative to the shrinking $\kappa_-$ and pushes $\beta$ above $\tfrac12$ . That drove a semiclassical program asking whether the renormalized $\langle T_{\mu\nu}\rangle$ diverges at the inner horizon even where classical fields stay finite — reconnecting censorship to the information problem and to the deepest question it guards: whether classical general relativity is, in the end, deterministic.

The physical picture

Stand just inside a charged black hole's inner horizon and look back out. That surface is a Cauchy horizon: a null wall which, in finite proper time, catches every signal the exterior universe will ever emit over its entire infinite future. The wavefronts pile onto the horizon and are exponentially blue-shifted — an eternity compressed into an instant. A late signal emitted at frequency $\omega_{\mathrm{em}}$ arrives at $\omega_{\mathrm{rec}}\sim\omega_{\mathrm{em}}\,e^{\kappa_- v}$ , with $\kappa_-$ the inner-horizon surface gravity and $v$ the advanced time. Because the whole future is received in a finite span, the integrated energy diverges — the seed of everything SCC relies on. The first figure fixes the geometry for a concrete hole and puts a number on the amplification.

Figure

Interior causal diagram of a Reissner-Nordström hole with Q = 0.8M, so r± = M ± √(M² − Q²) gives event horizon r₊ = 1.6M and inner horizon r₋ = 0.4M. Infalling radiation from region I with a Price tail v⁻ᵖ crosses r₊ and piles onto the Cauchy horizon r₋, blue-shifted by e^(κ₋ v) with κ₋ = 3.75/M. The pile-up drives mass inflation and turns the smooth inner horizon into a weak null singularity

The mathematics is a competition of two exponentials. Exterior perturbations fade — as power-law Price tails $v^{-p}$ when $\Lambda=0$ , or as exponential quasinormal ringdown $e^{-\alpha v}$ when $\Lambda>0$ — while the inner horizon amplifies whatever arrives by $e^{\kappa_- v}$ . If amplification wins, the connection is too singular to continue the field equations and censorship holds; if decay wins, a finite-energy extension survives and censorship fails. The fate of determinism thus hangs on the single ratio $\beta=\alpha/\kappa_-$ against the threshold $\tfrac12$ : small $\kappa_-$ (near extremality) and fast decay (positive $\Lambda$ ) both push $\beta$ up and endanger the conjecture. The second figure is the punchline — the phase line on which every known case sits.

Figure

The decisive ratio β = α/κ₋ on a number line with threshold β = ½. For β ≤ ½ the perturbation derivative fails to be square-integrable, no weak extension exists, and Christodoulou Strong Cosmic Censorship holds. For β greater than ½ the connection stays in H¹, a finite-energy extension survives, and censorship is violated. Asymptotically flat Reissner-Nordström and Kerr sit near β = 0 with power-law tails, while near-extremal Reissner-Nordström-de Sitter has κ₋ → 0 and lands beyond β = ½

The core equations

Determinism begins as an existence-and-uniqueness theorem: constraint-satisfying data on a slice evolve into one largest globally hyperbolic spacetime, the maximal globally hyperbolic development (MGHD) whose inextendibility SCC asserts.

\mathrm{MGHD}(\Sigma,h,K)=\text{unique maximal globally hyperbolic }(M,g)\ \text{with}\ R_{\mu\nu}-\tfrac12 R\,g_{\mu\nu}+\Lambda g_{\mu\nu}=8\pi T_{\mu\nu} \qquad (1)

where: $\Sigma$ — the initial Cauchy surface (spacelike hypersurface); $h_{ij}$ — its induced metric; $K_{ij}$ — its extrinsic curvature; $(M,g)$ — the developed spacetime; $\Lambda$ — cosmological constant. Reading: the data fix $(M,g)$ uniquely up to isometry; SCC conjectures this predicted region is, generically, all of spacetime. The MGHD stops at the Cauchy horizon — what survives of the closure of the future domain of dependence after removing every point that still has determined events to its future.

H^{+}(\Sigma)=\overline{D^{+}(\Sigma)}\setminus I^{-}\!\big(D^{+}(\Sigma)\big),\qquad D^{+}(\Sigma)\cup D^{-}(\Sigma)=M_{\mathrm{MGHD}} \qquad (2)

where: $D^{\pm}(\Sigma)$ — future and past domains of dependence; $H^{+}(\Sigma)$ — the future Cauchy horizon, an achronal null hypersurface beyond which data no longer determine the geometry; $I^{-}$ — chronological past. Reading: $H^{+}(\Sigma)$ is the exact frontier where prediction ends, and SCC asks that generic data make it empty or singular. The concrete arena is the charged black hole: its metric function has two roots, and the smaller is the Cauchy horizon of any exterior slice.

f(r)=1-\frac{2M}{r}+\frac{Q^{2}}{r^{2}},\qquad r_{\pm}=M\pm\sqrt{M^{2}-Q^{2}} \qquad (3)

where: $M$ — ADM mass in units $G=c=1$ ; $Q$ — electric charge; $r_{+}$ — outer event horizon; $r_{-}$ — inner (Cauchy) horizon, real and distinct when $|Q|<M$ . Reading: data fix the geometry only down to $r_{-}$ ; the continuation across $r_{-}$ is non-unique — exactly the SCC concern. The blue-shift rate is the inner-horizon surface gravity, half the slope of $f$ at the horizon.

\kappa_{\pm}=\tfrac12\left|f'(r_{\pm})\right|=\frac{r_{\pm}-r_{\mp}}{2r_{\pm}^{2}} \qquad (4)

where: $\kappa_{+}$ — outer-horizon surface gravity (Hawking temperature); $\kappa_{-}$ — inner-horizon surface gravity, the exponent in $e^{\kappa_- v}$ ; $v$ — advanced (ingoing) null coordinate. Reading: as $|Q|\to M$ the horizons merge and $\kappa_-\to0$ , switching the linear instability off. The linear seed is the competition between exterior decay and inner-horizon amplification, seen by re-expressing a decaying perturbation in a coordinate regular at $r_{-}$ .

\Psi\sim\Psi_{\mathrm{CH}}+(\text{regular})+A\,e^{-\kappa_- v}\,v^{-p}\ \Longrightarrow\ \partial_v\Psi\sim e^{\kappa_- v}\,v^{-p} \qquad (5)

where: $\Psi$ — a linearized perturbation; $v^{-p}$ — the late-time Price power-law tail; $e^{\kappa_- v}$ — amplification in a Kruskal-type coordinate regular at $r_{-}$ ; $A$ — amplitude from the exterior. Reading: the transverse derivative, not the field, blows up — the metric may stay continuous while its gradient diverges. Nonlinearly the amplified flux drives the local mass function to infinity — mass inflation — and with it the Weyl curvature.

m(v)\sim\frac{C}{v^{\,p}}\,e^{\kappa_- v}\longrightarrow\infty\quad(v\to\infty),\qquad \Psi_2\propto\frac{m}{r^{3}}\to\infty \qquad (6)

where: $m(v)$ — renormalized Misner-Sharp/Hawking mass at $r_{-}$ ; $C$ — a constant from the crossing fluxes; $\Psi_2$ — the Coulombic Weyl scalar. Reading: for the physical range of $p$ the exponential beats the power law, so curvature diverges at $r_{-}$ and the smooth inner horizon becomes a null singularity. Because that singularity can be weak, censorship is stated against a regularity class: Christodoulou's threshold is a non-square-integrable connection, and the verdict compresses into one ratio.

\Gamma\notin L^{2}_{\mathrm{loc}}\ (g\notin H^{1}_{\mathrm{loc}})\iff\text{SCC holds},\qquad \beta\equiv\frac{\alpha}{\kappa_-},\quad \text{Christodoulou-SCC}\iff\beta\le\tfrac12 \qquad (7)

where: $\Gamma^{\lambda}_{\mu\nu}$ — the metric connection; $L^{2}_{\mathrm{loc}}$ , $H^{1}_{\mathrm{loc}}$ — locally square-integrable functions and the Sobolev space of metrics with one square-integrable derivative; $\alpha$ — slowest exterior decay rate (Price exponent, or $|\mathrm{Im}\,\omega_0|$ for the lowest de Sitter quasinormal mode); $\beta$ — the decisive ratio. Reading: $\beta>\tfrac12$ gives an $H^{1}_{\mathrm{loc}}$ connection and a weak continuation past the horizon (censorship fails); $\beta\le\tfrac12$ forbids it (determinism safe).

Derivation AThe blue-shift instability of the Cauchy horizon

The goal is to show that a freely falling observer crossing the inner horizon receives exterior radiation amplified by $e^{\kappa_- v}$ , the exponent set entirely by the inner-horizon surface gravity. We track one ingoing wave from distant emission to reception at the horizon and read its frequency off a coordinate that stays regular where the usual ones fail — the linear precursor of mass inflation and the reason a generic perturbation is expected to shred the smooth horizon.

Step 1. Introduce ingoing and outgoing Eddington-Finkelstein coordinates $v=t+r_*$ and $u=t-r_*$ with the tortoise coordinate $r_*=\int\mathrm{d}r/f(r)$ . Near the inner horizon expand $f(r)\approx f'(r_-)(r-r_-)=-2\kappa_-(r-r_-)$ . Why: the surface gravity is nothing but the slope of $f$ at the horizon, so every near-horizon rate will be expressed through $\kappa_-$ alone.

Step 2. Integrate the tortoise coordinate through that linear zero to get $r_*\approx-\frac{1}{2\kappa_-}\ln|r-r_-|+\text{const}$ , which diverges logarithmically at $r_-$ . The Cauchy horizon therefore sits at $v\to\infty$ in these coordinates even though an infalling observer reaches it in finite proper time. Why: an infinite coordinate range is squeezed into a finite physical interval, which is exactly how an unbounded stretch of exterior history gets delivered in one moment.

Step 3. Remove the coordinate singularity with a Kruskal-type coordinate $V=-e^{\kappa_- v}$ , smooth and non-degenerate across $r_-$ . A wave of fixed frequency in $v$ has frequency $\propto \mathrm{d}v/\mathrm{d}V=e^{-\kappa_- v}/\kappa_-$ relative to $V$ . Why: physical frequencies must be measured against a clock that runs smoothly at the horizon, and $V$ is that clock.

Step 4. A locally inertial observer crossing the horizon measures frequency with respect to $V$ , so a late signal ( $v$ large) emitted at fixed $\omega_{\mathrm{em}}$ arrives at $\omega_{\mathrm{rec}}\sim\omega_{\mathrm{em}}\,e^{\kappa_- v}$ , unbounded as $v\to\infty$ . This is the boxed result $\omega_{\mathrm{rec}}/\omega_{\mathrm{em}}\sim e^{\kappa_- v}$ . Why: reading a fixed-frequency wave on the exponentially compressed clock $V$ is precisely what multiplies its measured frequency by $e^{\kappa_- v}$ .

Step 5. Because the entire infinite exterior future is received in finite proper time, the blue-shifted, integrated flux can diverge — the linear signature that goes nonlinear as mass inflation and motivates the whole conjecture. Why: a diverging energy density at the horizon is what forbids a smooth extension, tying the linear blue-shift to the loss of determinism.

Go deeper · Derivation B · 1 minThe decay-versus-blueshift criterion and the one-half threshold

Derivation BThe decay-versus-blueshift criterion and the one-half threshold

The goal is to derive when a linear perturbation crosses the Cauchy horizon with locally square-integrable derivatives — the Christodoulou regularity permitting a weak extension — and to show the boundary lands exactly at $\beta=\alpha/\kappa_-=\tfrac12$ . We combine exponential exterior decay with exponential blue-shift and test square-integrability of the field's transverse derivative in a regular coordinate, since that integral is the finite-energy condition separating extendible from inextendible.

Step 1. Take the slowest-decaying exterior mode to fall off as $\Psi\sim e^{-\alpha v}$ , where for $\Lambda>0$ the rate $\alpha=|\mathrm{Im}\,\omega_0|$ is the spectral gap of the lowest quasinormal mode, and for $\Lambda=0$ the decay is a polynomial Price tail with effectively zero exponential rate. Why: the exterior is the source of everything hitting the horizon, so its decay rate sets one side of the race.

Step 2. Transport the residual to the horizon in the regular coordinate: after accounting for the blue-shift factor $e^{\kappa_- v}$ , the leftover behaves as $\Psi-\Psi_{\mathrm{CH}}\sim e^{-(\alpha-\kappa_-)v}=e^{-\kappa_-(\beta-1)v}$ with $\beta=\alpha/\kappa_-$ . Why: subtracting the amplification exponent from the decay exponent is what turns two separate rates into the single competition parameter $\beta$ .

Step 3. Differentiate in the regular coordinate $V=-e^{\kappa_- v}$ , for which $\mathrm{d}v/\mathrm{d}V\propto e^{-\kappa_- v}$ , so $\partial_V\Psi\sim e^{-\kappa_-(\beta-1)v}\,e^{-\kappa_- v}=e^{-\kappa_-\beta v}$ . Why: the connection, not the metric, is what must be tested, and the connection lives in derivatives of the field — hence the transverse derivative.

Step 4. Test square-integrability against the regular measure: $\int^{\infty}|\partial_V\Psi|^{2}\,\frac{\mathrm{d}V}{\mathrm{d}v}\,\mathrm{d}v\sim\int^{\infty}e^{-2\kappa_-\beta v}\,e^{\kappa_- v}\,\mathrm{d}v=\int^{\infty}e^{-\kappa_-(2\beta-1)v}\,\mathrm{d}v$ , which converges if and only if $2\beta-1>0$ , i.e. $\beta>\tfrac12$ . Why: convergence of this one integral is the definition of a finite-energy ( $H^{1}_{\mathrm{loc}}$ ) perturbation at the horizon.

Step 5. Interpret: for $\beta>\tfrac12$ the perturbation is $H^{1}_{\mathrm{loc}}$ , a weak extension of the Einstein equations exists, and Christodoulou-SCC fails; for $\beta\le\tfrac12$ no such extension exists and censorship holds. Asymptotically flat holes ( $\Lambda=0$ ) give $\beta$ effectively at or below $\tfrac12$ ; near-extremal de Sitter holes can push $\beta>\tfrac12$ . Why: the boxed criterion $\int|\partial_V\Psi|^{2}\,\mathrm{d}v<\infty\iff\beta>\tfrac12$ converts a deep conjecture into a computable comparison of two rates.

Limiting cases

SCC reads differently across the parameter space of known solutions; the single dial is the ratio of exterior decay rate to inner-horizon surface gravity.

Case	Result	Meaning
Schwarzschild	$r_-=0$ , no Cauchy horizon	The singularity is spacelike and genuine; the MGHD is already inextendible, so SCC holds trivially with no subtlety.
Reissner-Nordström / Kerr, asymptotically flat	$\Lambda=0$ : Price tails, $\beta\le\tfrac12$	Power-law decay is too slow to beat the blue-shift, so mass inflation makes the connection non- $H^1$ ; Christodoulou-SCC is expected to hold, rigorously controlled for Kerr by Dafermos-Luk.
Near-extremal Reissner-Nordström-de Sitter	$\beta=\dfrac{\lvert\mathrm{Im}\,\omega_0\rvert}{\kappa_-}>\tfrac12$	Positive $\Lambda$ gives exponential ringdown while $\kappa_-\to0$ near extremality, so $\beta$ exceeds $\tfrac12$ : perturbations extend in $H^1_{\mathrm{loc}}$ and the Christodoulou criterion is violated — a genuine counterexample.
Extremal limit	$\kappa_-\to0$ , $r_+\to r_-\to M$	The horizons merge and the blue-shift shuts off; the Aretakis instability appears on the horizon instead, and the censorship picture changes character entirely.

Common trapStrong and Weak Cosmic Censorship are the same conjecture, and "inextendible" just means the curvature blows up.

Both halves are wrong. Weak Cosmic Censorship concerns causal structure seen from infinity — singularities from collapse should stay hidden behind event horizons so null infinity is complete (no naked singularities). Strong Cosmic Censorship concerns local determinism — the maximal globally hyperbolic development should be inextendible, so no Cauchy horizon survives for generic data. Neither implies the other; they are logically independent statements about visibility versus predictability. And "inextendible" is not "curvature diverges": the singularity at a perturbed Cauchy horizon is typically weak and null, with curvature scalars and the renormalized mass diverging while the metric continues continuously ( $C^0$ ) across it. If censorship meant only "no $C^0$ extension" it would be false wherever such a continuation exists — which is why Christodoulou pins it to a non- $H^1_{\mathrm{loc}}$ connection, the regularity below which the weak Einstein equations have no solution and determinism is genuinely lost.

Worked examples

1 · Horizons, surface gravities, and blue-shift of a charged hole (fully worked)

A non-extremal Reissner-Nordström black hole has $Q=0.8\,M$ in units $G=c=1$ . Locate both horizons, compute the inner and outer surface gravities, identify the Cauchy horizon, and estimate the blue-shift amplification over an advanced-time span $\Delta v=10\,M$ .

Step 1. Discriminant and horizons. $\sqrt{M^{2}-Q^{2}}=M\sqrt{1-0.64}=M\sqrt{0.36}=0.6\,M$ , so $r_{+}=M+0.6M=\mathbf{1.6\,M}$ (event horizon) and $r_{-}=M-0.6M=\mathbf{0.4\,M}$ (inner horizon). The Cauchy horizon of an exterior slice is $r_{-}=0.4\,M$ , well inside $r_{+}$ .

Step 2. Surface gravities from Eq. (4), $\kappa_{\pm}=(r_{+}-r_{-})/(2r_{\pm}^{2})$ with $r_{+}-r_{-}=1.2M$ . Outer: $\kappa_{+}=1.2M/(2\cdot(1.6M)^{2})=1.2M/(5.12M^{2})=\mathbf{0.234/M}$ . Inner: $\kappa_{-}=1.2M/(2\cdot(0.4M)^{2})=1.2M/(0.32M^{2})=\mathbf{3.75/M}$ . As a check the ratio is $\kappa_{-}/\kappa_{+}=(r_{+}/r_{-})^{2}=(1.6/0.4)^{2}=\mathbf{16}$ .

Step 3. Blue-shift over $\Delta v=10M$ . The amplification factor is $e^{\kappa_{-}\Delta v}=e^{(3.75/M)(10M)}=e^{37.5}\approx\mathbf{1.9\times10^{16}}$ . A signal that took a modest slice of exterior history to arrive is boosted by sixteen orders of magnitude in frequency — and the true horizon lives at $v\to\infty$ , so the idealized factor is unbounded.

Step 4. Reading. For $Q=0.8M$ the inner surface gravity $\kappa_{-}=3.75/M$ dwarfs $\kappa_{+}=0.234/M$ by a factor 16, and even a $10M$ span blue-shifts infall by $\sim10^{16}$ — the seed of the mass-inflation singularity that SCC conjectures seals the horizon.

2 · The censorship verdict from the ratio β (fully worked)

Suppose the slowest exterior decay rate is $\alpha=0.5/M$ . For an inner horizon with $\kappa_{-}=2.0/M$ (far from extremal) and then $\kappa_{-}=0.2/M$ (near extremal), compute $\beta=\alpha/\kappa_{-}$ , decide whether Christodoulou-SCC holds, and find the critical $\kappa_{-}$ that separates the two verdicts.

Step 1. Far from extremality: $\beta=\alpha/\kappa_{-}=(0.5/M)/(2.0/M)=\mathbf{0.25}$ . Since $\beta\le\tfrac12$ , the transverse derivative fails to be square-integrable, no weak extension exists, and Christodoulou-SCC holds — determinism is safe.

Step 2. Near extremality: $\beta=(0.5/M)/(0.2/M)=\mathbf{2.5}$ . Now $\beta>\tfrac12$ , the connection stays in $H^{1}_{\mathrm{loc}}$ , a finite-energy extension survives, and Christodoulou-SCC is violated — this is the de Sitter-style counterexample in miniature.

Step 3. Critical value. Setting $\beta=\tfrac12$ gives $\kappa_{-}=2\alpha=2\cdot(0.5/M)=\mathbf{1.0/M}$ . For $\kappa_{-}>1.0/M$ censorship holds; for $\kappa_{-}<1.0/M$ it fails. Because $\kappa_{-}\to0$ at extremality while a positive $\Lambda$ keeps $\alpha$ finite, near-extremal de Sitter holes are driven straight into the violating regime.

Step 4. Reading. A large blue-shift ( $\kappa_{-}$ ) protects determinism; exponential exterior decay ( $\alpha$ large, from $\Lambda>0$ ) and near-extremality ( $\kappa_{-}$ small) endanger it, knife-edge at $\beta=\tfrac12$ . For $\Lambda=0$ the tails are only polynomial, effectively $\alpha\to0$ , so $\beta\le\tfrac12$ and asymptotically flat Reissner-Nordström and Kerr stay safe.

Physics of the portrait

Cosmic censorship is inseparable from one person: the geometer who reinvented how relativists reason about singularities, causal boundaries, and the global shape of spacetime, and who then asked the two questions — weak and strong — that have organized the field ever since.

Origins · 1 minRoger Penrose and how this was discovered

Roger Penrose

1931– · global methods, singularity theorems, and cosmic censorship

Roger Penrose turned general relativity from a subject about explicit metrics into one about global causal structure. His 1965 singularity theorem introduced trapped surfaces and used topological, coordinate-free arguments to prove that generic gravitational collapse forces a singularity — work recognized with the 2020 Nobel Prize in Physics. To make such statements precise he built the machinery this cluster runs on: conformal (Penrose-Carter) diagrams that compactify infinity onto a finite page, the achronal-boundary theorems that give Cauchy and event horizons their null, geodesically ruled character, and the language of domains of dependence and global hyperbolicity. Out of this came the two conjectures that bear his name. Weak Cosmic Censorship (1969) proposes that collapse singularities stay hidden behind event horizons, keeping null infinity free of naked singularities. Strong Cosmic Censorship, sharpened through the 1970s and stated in inextendibility form in his 1979 essay "Singularities and Time-Asymmetry", proposes that the maximal globally hyperbolic development of generic data is inextendible — that determinism itself, not merely modesty about singularities, is preserved. The engine he identified for the strong form was the blue-shift instability at inner horizons, later made quantitative by Poisson-Israel mass inflation and by the rigorous programs of Christodoulou, Dafermos, Luk and Rodnianski. It is this global, causal viewpoint — and the censorship questions — that make him the presiding figure of this article.

History

1965

— Penrose proves his singularity theorem using trapped surfaces and global causal methods

1969

— Choquet-Bruhat and Geroch establish the unique maximal globally hyperbolic development

1979

— Penrose gives strong cosmic censorship its modern inextendibility formulation

1990

— Poisson and Israel identify the mass-inflation singularity at the inner horizon

2018

— Cardoso, Costa, Destounis, Hintz and Jansen exhibit near-extremal de Sitter violations

Going further

Texts. Misner, Thorne & Wheeler, Gravitation (1973), Ch. 33-34 · Wald, General Relativity (1984), Ch. 12 and Sec. 9.5-9.6 · Hawking & Ellis, The Large Scale Structure of Space-Time (1973), Ch. 5 and 8 · Penrose, "Gravitational collapse", Riv. Nuovo Cimento 1, 252 (1969), and "Singularities and Time-Asymmetry" (1979) · Christodoulou, Ann. Math. 149, 183 (1999) · Dafermos & Luk, "The $C^0$ -stability of the Kerr Cauchy horizon" (2017) · Cardoso, Costa, Destounis, Hintz & Jansen, PRL 120, 031103 (2018) · Poisson & Israel, Phys. Rev. D 41, 1796 (1990). The Reissner-Nordström horizon radii $r_{\pm}=M\pm\sqrt{M^{2}-Q^{2}}$ and surface gravities $\kappa_{\pm}=(r_{\pm}-r_{\mp})/(2r_{\pm}^{2})$ follow from the roots of $r^{2}-2Mr+Q^{2}=0$ via Vieta's relations $r_{+}+r_{-}=2M$ and $r_{+}r_{-}=Q^{2}$ , and reduce to the Schwarzschild value $\kappa=1/(4M)$ as $Q\to0$ . Direct substitution at $Q=0.8M$ gives $r_{+}=1.6M$ , $r_{-}=0.4M$ , $\kappa_{-}=3.75/M$ , $\kappa_{+}=0.234/M$ , $\kappa_{-}/\kappa_{+}=16$ , and $e^{37.5}\approx1.9\times10^{16}$ .

Prerequisites. Domain of Dependence and Cauchy Horizons · Global Hyperbolicity · Cauchy Surfaces · Causal Structure of Spacetime · Penrose-Carter Conformal Diagrams. Leads to. Asymptotic Flatness · Conformal Infinity and Asymptotic Structure · Causality Conditions Hierarchy.

Ask the tutor: Want to walk through how mass inflation turns the Kerr inner horizon into a weak null singularity, or how a positive cosmological constant pushes β past ½ to break Christodoulou censorship in the de Sitter case?

What is the difference between Strong and Weak Cosmic Censorship?

They guard different things. Weak Cosmic Censorship is about visibility from infinity: singularities from collapse should stay hidden behind event horizons, so a distant observer never sees a naked singularity and null infinity stays complete. Strong Cosmic Censorship is about determinism: the maximal globally hyperbolic development of generic data should be inextendible, so no Cauchy horizon survives to make the future ambiguous. One asks whether you can see the singularity, the other whether the field equations still predict. Neither implies the other — a spacetime could hide its singularity yet remain extendible, or be inextendible yet expose one — which is why they are studied as separate conjectures.

Why can't Strong Cosmic Censorship simply say "the curvature blows up"?

Because the singularity that forms at a perturbed Cauchy horizon is usually weak. It is a null singularity where curvature scalars and the renormalized mass diverge, yet the metric can be extended continuously ( $C^0$ ) across it — an infalling observer is not crushed the way they are at the spacelike Schwarzschild singularity. If "inextendible" meant only "no continuous extension", the conjecture would be false exactly where such an extension exists (Kerr, by Dafermos-Luk). Christodoulou therefore rephrased it as inextendibility with a non-square-integrable connection, $g\notin H^1_{\mathrm{loc}}$ — the precise regularity below which the weak Einstein equations have no solution, and so the honest threshold at which determinism is lost.

What exactly is mass inflation?

Mass inflation is the nonlinear mechanism that turns the blue-shift instability into a genuine singularity at $r_-$ . Two streams of radiation meet near the inner horizon: an ingoing flux with a Price tail $v^{-p}$ and an outgoing flux from backscatter and collapsing matter. Each is blue-shifted, and their product feeds the local Misner-Sharp mass nonlinearly, $m(v)\sim (C/v^{p})\,e^{\kappa_- v}$ . For the physical range of $p$ the exponential overwhelms the power law, $m(v)\to\infty$ , and the Weyl scalar $\Psi_2\propto m/r^3$ diverges with it. The smooth Cauchy horizon of the idealized solution is thereby replaced, in any perturbed hole, by a curvature singularity — the realization of what SCC needs to restore determinism.

How does a positive cosmological constant threaten the conjecture?

With $\Lambda>0$ exterior perturbations do not fade as slow Price tails; they decay exponentially through quasinormal ringdown at a rate $\alpha=|\mathrm{Im}\,\omega_0|$ . For near-extremal black holes this decay can be fast while $\kappa_-$ is small, so $\beta=\alpha/\kappa_-$ climbs above $\tfrac12$ . Past that threshold the perturbation reaches the Cauchy horizon with square-integrable derivatives, a finite-energy ( $H^1_{\mathrm{loc}}$ ) extension exists, and the Christodoulou formulation is violated. This is what Cardoso, Costa, Destounis, Hintz and Jansen found in 2018 for Reissner-Nordström-de Sitter — the sharpest challenge to the classical conjecture, and the reason attention turned to whether quantum effects restore censorship.

Has Strong Cosmic Censorship been proved anywhere?

Yes, in an important model. Christodoulou proved both Weak and (his formulation of) Strong Cosmic Censorship for the spherically symmetric Einstein-scalar-field system: generic data either disperse or collapse into a black hole whose terminal boundary is a spacelike singularity with $H^1$ -inextendible connection. Completed in the late 1990s, it is the strongest evidence that censorship holds for genuinely dynamical, asymptotically flat collapse, and its techniques now underpin the field. Beyond spherical symmetry the picture is partial: Dafermos and Luk proved $C^0$ -stability of the Kerr Cauchy horizon (the singularity there is weak), while full $H^1_{\mathrm{loc}}$ -inextendibility for vacuum Kerr, and the fate of censorship with $\Lambda>0$ , remain open and contested.

What does the Choquet-Bruhat-Geroch theorem contribute?

It makes "determinism" precise. Given data $(\Sigma,h,K)$ satisfying the Einstein constraints, there exists a unique — up to isometry — maximal globally hyperbolic development. Uniqueness is the exact statement that general relativity predicts: two developments of the same data agree on their common globally hyperbolic region. Maximality means you cannot enlarge it without leaving global hyperbolicity, and the surface where you would leave it is a Cauchy horizon. SCC is then the conjecture, layered on top, that for generic data this maximal development is inextendible — that the region physics can determine is, generically, the whole spacetime, with no smooth continuation beyond a Cauchy horizon.