Cauchy Horizon Instability

Determinism is general relativity's most valuable and most fragile promise. Lay initial data on a spacelike slice, feed it to the Einstein equations, and the geometry is fixed — but only inside the domain of dependence, the region every past-directed causal curve threads back to the slice. Its edge is the Cauchy horizon, and past it the data no longer decide anything: infinitely many smooth continuations satisfy the same equations. Strong cosmic censorship is the conjecture that Nature never lets a real observer reach such an edge — that for generic data the maximal Cauchy development is a genuine boundary of spacetime, not a smooth gate one drives through into a lawless beyond. The battleground is the interior of a charged or rotating black hole. Solve the equations with exact symmetry and you find a pristine inner horizon: cross it and the maximal analytic extension opens onto other universes, naked singularities, and closed timelike curves — predictability annihilated. The escape is that the exact solution is not generic. Every real black hole is dressed in perturbations — the gravitational-wave tail of its own formation, infalling matter, the cosmic microwave background — and the geometry funnels all of it onto the inner horizon, where radiation that fades to nothing outside is exponentially concentrated in an infalling frame, its energy set blazing by the inner-horizon surface gravity κ₋. Two effects then close the trap. Blueshift sharpens the infalling flux without bound; mass inflation cross-multiplies it against backscattered radiation, driving a quasi-local mass and the curvature to infinity — a null, weak singularity found by Poisson and Israel and by Ori. Whether that singularity is fierce enough to block extension is the sharp modern question, and its answer decides the fate of determinism.

Key takeaways

A Cauchy horizon is the boundary of the region predictable from initial data; beyond it the Einstein equations lose deterministic power.
Inside charged (Reissner-Nordström) and rotating (Kerr) black holes the inner horizon is a Cauchy horizon, where infalling radiation is exponentially blueshifted at the rate set by the inner-horizon surface gravity κ₋.
The blueshift feeds the mass-inflation instability: backscattered radiation crossing the inner horizon drives a quasi-local mass and the curvature scalars to infinity, forming a null curvature singularity.
This is the dynamical engine of strong cosmic censorship — generic data should build a Cauchy horizon too singular to extend, restoring determinism to gravitational collapse.
The singularity is null and weak: the metric can stay continuous (C⁰) while curvature diverges, so the sharp criterion is the Christodoulou-Chruściel demand of inextendibility with square-integrable Christoffel symbols (H¹).
A positive Λ swaps power-law tails for exponential quasinormal decay, and near-extremal de Sitter holes can satisfy the survival condition β > 1/2 — violating the H¹ form of the conjecture.
Key idea: the inner horizon of a charged or rotating black hole is a Cauchy horizon where late-time perturbations are blueshifted exponentially at the rate κ₋; this flux, cross-multiplied with backscattered outflux, inflates a quasi-local mass and the curvature into a null singularity that generically renders the spacetime inextendible — the dynamical content of strong cosmic censorship.

The core insight

The inner horizon of a charged or rotating black hole is a Cauchy horizon where late-time perturbations are exponentially blueshifted at the rate set by the inner-horizon surface gravity κ₋; this blueshifted flux, cross-multiplied with backscattered outflux, inflates a quasi-local mass and the curvature, producing a null singularity that (generically) renders the spacetime inextendible — the dynamical content of strong cosmic censorship.

Where this lives today

The Cauchy-horizon instability has moved from heuristic to theorem to measurement. On the mathematical front, Dafermos and Luk proved in the 2010s that the Cauchy horizon of a perturbed Kerr black hole does form and is C⁰-extendible — the metric really is continuous across it — while remaining generically inextendible in the stronger senses, exactly the C⁰-versus-H¹ dichotomy the physics predicted; numerical-relativity simulations of black-hole interiors now watch mass inflation happen. On the physics front, the 2018 discovery that near-extremal Reissner-Nordström-de Sitter holes can push β past 1/2 turned strong cosmic censorship into a question about quasinormal spectra — the same ringdown modes that LIGO, Virgo and KAGRA extract from merging black holes — so every measured decay rate α is indirectly a datum about a Cauchy horizon no telescope will ever see. Rotating Kerr holes, the astrophysical case, appear to shore censorship back up where the charged de Sitter models threatened it, and whether real spinning black holes preserve determinism at their cores is an active open problem spanning numerical relativity, gravitational-wave astronomy, and global analysis.

The physical picture

Stand just inside the inner horizon and look back the way you came. The entire infinite future of the outside universe — every photon that will ever fall in — is compressed onto the brief instant of your approach to the horizon, and a signal that decays lazily out there is squeezed into your local now and amplified without limit. The inner horizon is a cosmic lens of infinite magnification, and the exponent that sets it is the inner surface gravity κ₋, not the event-horizon value κ₊ that governs Hawking temperature and the redshift you left behind. The first figure lays out the geography. In the Penrose-Carter diagram of the Reissner-Nordström interior, region II is the shell between the event horizon r = r₊ and the inner horizon r = r₋. Radius is a time coordinate there, so an infalling observer moves up the diamond toward its upper edges — the future Cauchy horizon — where ingoing radiative tails run parallel to the horizons and pile up, and it is this pile-up, read in a frame smooth on the horizon, that diverges.

Figure

Penrose-Carter diagram of the Reissner-Nordström interior: region II lies between the event horizon r = r₊ and the inner horizon r = r₋. An infalling observer moves upward as radius decreases; ingoing tails pile up on the future Cauchy horizon, where the locally measured energy density grows as e^(2κ₋ v) and mass inflation turns the horizon into a null, weak curvature singularity

Mass inflation is best pictured as two beams colliding: an ingoing stream of blueshifted radiation crosses an outgoing stream backscattered off the interior potential, and each beam is amplified by the gravity of the other — a runaway captured by a Misner-Sharp mass that grows without bound, yet reached in finite proper time with the metric still continuous. The instability is a race, and the second figure is the scoreboard: κ₋, which sets the blueshift, plotted against charge. Where κ₋ is large the trap slams shut; as the hole nears extremality κ₋ collapses toward zero and the horizon fights back.

Figure

Surface gravities of the two horizons versus charge for a Reissner-Nordström black hole (M = 1): the inner-horizon κ₋ (red) sets the blueshift and always exceeds the outer κ₊ (blue). At Q = 0.9 M one finds κ₋ = 1.37/M, κ₊ = 0.21/M, a blueshift-to-redshift ratio of 6.5; both vanish at the extremal limit Q = M where the instability degenerates

Notice what this does not require: no new force, no exotic matter. The same Einstein-Maxwell equations that produced the smooth analytic inner horizon guarantee, once realistic radiative tails are switched on, that it is destroyed — the passage to other universes is an artifact of exact symmetry that generic collapse never builds.

The core equations

A Cauchy horizon is defined causally, before any metric is chosen: it is the edge of the region a spacelike data set can determine, where the domain of dependence — the points whose causal past is anchored to the slice — runs out.

D^+(\Sigma) = \{\, p : \text{every past-inextendible causal curve through } p \text{ meets } \Sigma \,\},\qquad H^+(\Sigma) = \overline{D^+(\Sigma)} \setminus I^-\big(D^+(\Sigma)\big) \qquad (1)

where: Σ — a partial Cauchy (initial-data) surface, achronal; $D^+(\Sigma)$ — future domain of dependence, the region fixed by data on Σ; $H^+(\Sigma)$ — future Cauchy horizon; $I^-$ — chronological past. Reading: past $H^+$ the equations still hold but the data no longer pick a unique solution. The concrete stage is the Reissner-Nordström (charged, static, spherical) black hole, the simplest exact geometry with an inner horizon; its Kerr cousin behaves the same way, and charge stands in for spin because it is far easier to compute with.

\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 (2)

where: $M$ — ADM mass in geometric units $G = c = 1$ ; $Q$ — electric charge; $f(r)$ — the lapse; $\mathrm{d}\Omega^2$ — the unit-sphere metric. Reading: the zeros of $f$ give, for $0 < |Q| < M$ , two horizons $r_\pm = M \pm \sqrt{M^2 - Q^2}$ — outer (event) and inner (Cauchy). Each horizon carries a surface gravity, the redshift/blueshift rate read off there — the key number in the story, and, crucially, the inner κ₋ always exceeds the outer κ₊.

\kappa_\pm = \tfrac{1}{2}\,\big|f'(r_\pm)\big| = \frac{|r_\pm - r_\mp|}{2\,r_\pm^2} = \frac{\sqrt{M^2 - Q^2}}{r_\pm^2} \qquad (3)

where: $\kappa_+$ — outer-horizon surface gravity, fixing the Hawking temperature $T = \kappa_+/2\pi$ ; $\kappa_-$ — inner-horizon surface gravity, fixing the blueshift; $f'(r_\pm)$ — radial slope of the lapse. Reading: because $r_- < r_+$ , the same numerator is divided by a smaller radius, so $\kappa_- > \kappa_+$ always — the inner horizon is the fiercer surface. The Cauchy horizon is reached only as advanced time $v = t + r_*$ runs to infinity, yet a coordinate $U$ smooth across it packs that infinite range into a finite one — an exponential relation that is the blueshift, pure geometry before any field enters.

v = t + r_*,\qquad r_* = \int \frac{\mathrm{d}r}{f(r)},\qquad U \sim -\,e^{-\kappa_- v}\ \text{ as } v \to \infty \qquad (4)

where: $v$ — advanced (ingoing Eddington) time, Cauchy horizon at $v \to +\infty$ ; $r_*$ — tortoise coordinate; $U$ — affine coordinate regular on $H^+$ , reaching the horizon at $U \to 0^-$ . Reading: any advanced-time frequency becomes an unbounded local frequency — the definition of blueshift. Nonlinearly, the blueshifted flux drives a quasi-local mass to infinity. The Misner-Sharp mass measures the energy inside a sphere of area radius $r$ ; at the Cauchy horizon it inflates without bound.

m(r) = \frac{r}{2}\Big(1 - g^{ab}\,\partial_a r\,\partial_b r\Big),\qquad m(v) \sim \frac{e^{\kappa_- v}}{v^{\,2p+2}} \xrightarrow{\;v\to\infty\;} \infty \qquad (5)

where: $m(r)$ — Misner-Sharp mass; $g^{ab}\partial_a r\,\partial_b r$ — squared gradient of the area radius, equal to $f$ in static RN; $p \ge 2$ — Price-law tail exponent of the field, $\Phi\sim v^{-p}$ , so the ingoing energy flux decays as $v^{-(2p+2)}$ . Reading: the exponential blueshift overwhelms the tail, so the local mass diverges even though the total mass $M$ is unchanged. Because curvature is built from the mass, it diverges too — but the metric itself need not. This split is the subtle heart of the modern story.

\Psi_2 \sim \frac{m(v)}{r_-^{\,3}} \to \infty,\qquad R_{abcd}R^{abcd} \to \infty,\qquad g_{\mu\nu} \in C^0 \ \text{ on } H^+ \qquad (6)

where: $\Psi_2$ — Coulombic Weyl scalar; $R_{abcd}R^{abcd}$ — Kretschmann invariant; $C^0$ — continuity of the metric. Reading: scalars blow up (a genuine curvature singularity), yet the metric has finite limits, so the singularity is null and weak — tidal forces diverge but integrate to a finite distortion. The decisive question is therefore not whether curvature diverges but whether the spacetime extends across $H^+$ as a weak solution. Christodoulou and Chruściel sharpened strong cosmic censorship into inextendibility with square-integrable Christoffel symbols — the $H^1$ criterion — turning the race into a single inequality.

\beta \equiv \frac{\alpha}{\kappa_-},\qquad \Phi \in H^1_{\mathrm{loc}}(H^+) \iff \beta > \tfrac{1}{2} \qquad (7)

where: α — spectral gap, the slowest decay rate $-\,\mathrm{Im}\,\omega_{\mathrm{QNM}}$ (de Sitter case); $\kappa_-$ — inner surface gravity; β — the ratio that decides the race; $H^1_{\mathrm{loc}}$ — locally square-integrable first derivatives. Reading: if β > 1/2 the field continues across the horizon and predictability fails; if β ≤ 1/2 it is too singular to extend and censorship holds.

Derivation ABlueshift instability of a linear field at the inner horizon

The goal is to show that a generic perturbation — one that fades away as an inverse power of advanced time in the external universe — is amplified without bound when read in the frame of an observer crossing the Cauchy horizon, so that the locally measured energy density diverges even though the background geometry is held perfectly fixed. We work on a frozen Reissner-Nordström background, take the standard late-time Price tail of a massless field, transform to the coordinate smooth on the horizon, and track the energy the crossing observer measures as advanced time runs to infinity.

Step 1. Take the late-time tail $\Phi(v) \sim v^{-p}$ with $p \ge 2$ . A massless field scattering off the long-range Reissner-Nordström potential leaves an inverse-power tail in advanced time along the inner horizon. Why: this is the generic radiative debris of any realistic collapse, not a special choice.

Step 2. Change to the regular coordinate. The coordinate smooth on $H^+$ obeys $U \propto -\,e^{-\kappa_- v}$ , so $v = -\kappa_-^{-1}\ln(-U)$ , compressing the whole range $v \in (v_0,\infty)$ into a finite one ending at $U \to 0^-$ . Why: physical quantities must be read in a frame that does not itself blow up at the horizon.

Step 3. Read the local frequency. A finite advanced-time frequency $\omega_v$ becomes $\omega_{\mathrm{loc}} = |\mathrm{d}v/\mathrm{d}U|\,\omega_v = \omega_v/[\kappa_-(-U)] \to \infty$ as $U \to 0^-$ . Why: this is the blueshift made quantitative — the crossing observer sees the oscillation tick infinitely fast.

Step 4. Assemble the energy density. The flux on crossing is $\rho_{\mathrm{loc}} \sim (\partial_U\Phi)^2 = (\mathrm{d}v/\mathrm{d}U)^2 (\partial_v\Phi)^2 \sim e^{2\kappa_- v}\,v^{-2p-2}$ ; the exponential prefactor beats any power-law decay, so $\rho_{\mathrm{loc}} \to \infty$ . Why: a slowly dying field delivers an infinite local energy density at the horizon.

Step 5. Read the verdict. At linear order this divergence is only the seed, because the background was held fixed; it signals that backreaction — the next derivation — must convert the seed into a true singularity. Why: a test-field calculation flags the instability but cannot by itself prove a curvature singularity. Boxed result: $\rho_{\mathrm{loc}} \sim e^{2\kappa_- v}\,v^{-2p-2} \xrightarrow{\,v\to\infty\,} \infty$ — infinite blueshift.

Go deeper · Derivation B · 2 minMass inflation from the Ori cross-flow model

Derivation BMass inflation from the Ori cross-flow model

The goal is to promote the linear blueshift seed into a genuine curvature singularity by restoring backreaction, showing that when an ingoing blueshifted flux crosses an outgoing backscattered flux the Misner-Sharp mass diverges in finite proper time while the metric stays continuous — a null, weak singularity rather than the spacelike crush of Schwarzschild. We use Ori's two-stream picture: a contracting null shell carrying the outflux separates two ingoing-Vaidya regions, and we follow the mass function as the shell nears the inner horizon.

Step 1. Encode the tail as a slowly varying mass. The ingoing Vaidya region has $m_{\mathrm{in}}(v) = M - c\,v^{-(2p+1)}$ , approaching the constant $M$ as a power law. Why: this repackages the Price tail as a time-dependent mass parameter, the natural variable for the Einstein-Maxwell equations with null dust.

Step 2. Add the outflux and let the streams cross. Backscattered radiation supplies an outgoing null flux; where in- and outflux overlap, each is gravitationally amplified by the energy of the other. Why: this cross-flow is the essential nonlinearity — a single Vaidya region shows nothing; the divergence lives in the interaction of the two streams.

Step 3. Track the mass near the horizon. Matching across the shell gives $m(v) \propto |\partial_v r|^{-1}\,m_{\mathrm{in}}(v)\,\times\,(\text{outflux})$ , and as $r \to r_-$ the contraction rate $\partial_v r \sim e^{-\kappa_- v}$ becomes exponentially small. Why: the vanishing of $\partial_v r$ is where κ₋ re-enters and injects the exponential growth.

Step 4. Collect the divergence. Combining the pieces, $m(v) \sim e^{\kappa_- v}/v^{\,2p+2} \to \infty$ : exponential growth defeats the power-law suppression and the quasi-local mass inflates without bound. Why: this is mass inflation — the interior energy runs to infinity while the hole's external mass $M$ is untouched.

Step 5. Convert mass to curvature. The Coulombic Weyl scalar tracks the mass, $\Psi_2 \sim m(v)/r_-^3 \to \infty$ , so the Kretschmann invariant diverges. Why: curvature scalars are coordinate-invariant, so their blow-up is a real feature of the geometry.

Step 6. Check that it is null and weak. The metric components still tend to finite limits ( $g_{\mu\nu} \in C^0$ ), the observer reaches the horizon in finite proper time $\int\mathrm{d}\tau < \infty$ , and the doubly integrated tidal distortion stays finite. Why: the singularity is real but gentle enough that inextendibility is not automatic — the subtlety at the heart of strong cosmic censorship. Boxed result: $m(v) \sim e^{\kappa_- v}/v^{\,2p+2} \to \infty$ with $\tau_{\mathrm{sing}} < \infty$ and $g_{\mu\nu} \in C^0$ — mass inflation.

Limiting cases

The character of the Cauchy horizon turns sharply on charge or spin, on the cosmological constant, and on the regularity demanded of an extension; the instability is at full strength only in the sub-extremal, asymptotically flat middle ground.

Case	Result	Meaning
Schwarzschild (Q = 0, a = 0)	r₋ → 0, no inner horizon	The singularity is spacelike and the spacetime is globally hyperbolic; there is no Cauchy horizon at all, so censorship is trivially satisfied and the instability is moot.
Sub-extremal RN / Kerr (0 < \	Q\	< M)	κ₋ = √(M²−Q²)/r₋² > κ₊ > 0	A genuine Cauchy horizon exists with nonzero inner surface gravity; the blueshift is exponential and mass inflation proceeds. This is the canonical regime where the instability operates.
Extremal limit (\	Q\	→ M, a → M)	r₊ = r₋ = M, κ₋ → 0	The horizons merge and the surface gravity vanishes, degrading the exponential blueshift to a mild power law; extremal Cauchy horizons are the hardest, most stable cases for censorship.
de Sitter (Λ > 0)	β = α/κ₋, tails ∼ e^(−αv)	Power-law tails give way to exponential quasinormal decay; if β > 1/2 the field survives in H¹ across the horizon, violating the Christodoulou form of strong cosmic censorship — reachable for near-extremal RN-de Sitter.

Common trapCurvature diverges, therefore observers are destroyed and the spacetime cannot be extended

This is the central trap, and it packs three errors. First, a diverging Kretschmann scalar does not by itself shred an infalling body: the singularity is null and weak, the metric stays continuous (C⁰), and the tidal distortion integrated along a worldline is finite, so a small object can in principle cross with finite total deformation. Second, "the metric is continuous" does not mean "the spacetime is extendible" in the sense that matters — strong cosmic censorship demands inextendibility with square-integrable Christoffel symbols (the H¹ criterion), strictly stronger than C⁰-continuity, and the two can disagree. Third, the blueshift exponent is the inner surface gravity κ₋, never the outer κ₊ that sets Hawking temperature; and in de Sitter the competing rate is the quasinormal spectral gap α, not a power-law tail. Keep the three numbers — κ₋, κ₊, α — and the three regularity classes — C⁰, H¹, curvature — straight, and the paradoxes dissolve.

Worked examples

1 · Surface gravities and blueshift of a Q = 0.9 M charged hole (fully worked)

A Reissner-Nordström black hole has charge Q = 0.9 M. Locate its horizons, compute both surface gravities, and estimate by how much the local energy density of an infalling tail is amplified over ten units of advanced time as it approaches the Cauchy horizon.

Step 1. Horizons. With M = 1, $\sqrt{M^2 - Q^2} = \sqrt{0.19} = 0.43589$ , so r₊ = 1.43589 M and r₋ = 0.56411 M — two distinct roots, a sub-extremal hole with a real inner horizon.

Step 2. Surface gravities from equation (3). $\kappa_- = \sqrt{M^2-Q^2}/r_-^2 = 0.43589/0.31822 = \mathbf{1.370\,/M}$ , and $\kappa_+ = 0.43589/2.06177 = \mathbf{0.211\,/M}$ . As promised, the inner value is larger.

Step 3. The blueshift ratio. $\kappa_-/\kappa_+ = 1.370/0.211 = \mathbf{6.5}$ : the inner horizon amplifies infalling radiation more than six times faster than the event horizon redshifts escaping radiation — exactly why the tail wins the race.

Step 4. Energy amplification over Δv = 10 M. From Derivation A the local energy density scales as $\rho_{\mathrm{loc}} \propto e^{2\kappa_- v}$ , so over ten units of advanced time it grows by $e^{2\kappa_- \Delta v} = e^{2\times 1.370\times 10} = e^{27.4} \approx \mathbf{8\times10^{11}}$ . In one Price-tail decade the local flux jumps by nearly twelve orders of magnitude — and the factor itself runs to infinity as v → ∞, leaving the tail $v^{-2p-2}$ negligible.

2 · Does de Sitter let censorship fail? Evaluating β near extremality

A near-extremal Reissner-Nordström-de Sitter black hole has inner surface gravity κ₋ = 0.19/M, and its dominant quasinormal family has spectral gap α = 0.11/M. Decide whether the H¹ form of strong cosmic censorship survives.

Step 1. Form the ratio. The censorship criterion (equation 7) rests on $\beta = \alpha/\kappa_- = 0.11/0.19 = \mathbf{0.58}$ .

Step 2. Compare with the threshold. Since β = 0.58 > 1/2, the field behaves as $(-U)^{0.58}$ near the horizon, and $\int (-U)^{-0.84}\,\mathrm{d}U$ converges: the first derivative is square-integrable.

Step 3. State the verdict. The perturbation extends across the Cauchy horizon as an H¹ weak solution, so the Christodoulou-Chruściel form of strong cosmic censorship is violated for this hole — an observer could in principle cross into a region where the equations no longer fix the future. This is the effect Cardoso, Costa, Destounis, Hintz and Jansen found for near-extremal RN-de Sitter in 2018. Remove Λ and the power-law tails restore censorship — the cosmological constant, not the hole alone, opens the door.

Physics of the portrait

The instability is the machinery; the question it answers belongs to one person. Roger Penrose gave general relativity both its singularity theorems and the cosmic-censorship conjectures — and he was the first to notice that the inner horizon of a charged black hole cannot be trusted.

Origins · 1 minRoger Penrose and how this was discovered

Roger Penrose

1931– · cosmic censorship and the global structure of spacetime

Roger Penrose reshaped general relativity by importing global, topological reasoning into a subject that had lived on local differential equations. His 1965 singularity theorem showed that once a trapped surface forms, geodesic incompleteness — a singularity — is unavoidable, independent of symmetry; that work, and its Hawking-Penrose successors, won him a share of the 2020 Nobel Prize. Having proved singularities are generic, he asked whether they are always hidden. Weak cosmic censorship (1969) conjectures that collapse hides its singularities behind event horizons, so the asymptotic world stays predictable; strong cosmic censorship, sharpened through the 1970s, makes the deeper claim that no observer — not even one who has fallen in — ever reaches a Cauchy horizon and watches determinism fail. Penrose supplied the first physical reason to believe it: already in 1968 he pointed out that the inner horizon of Reissner-Nordström is a surface of infinite blueshift, so the smooth passage to other universes is a fiction the slightest perturbation destroys. The Penrose-Carter conformal diagrams that make all this visible are his too. The instability elaborated here — blueshift, mass inflation, the H¹ criterion — is the detailed vindication of an intuition Penrose stated in a single sentence more than half a century ago.

History

1968

— Penrose argues that the inner horizon of Reissner-Nordström is unstable to infinite blueshift

1973

— Simpson and Penrose sharpen the blueshift instability of the charged black hole

1990

— Poisson and Israel discover mass inflation at the Cauchy horizon

1991

— Ori models the horizon as a null, weak curvature singularity via a cross-flow

2018

— Cardoso and collaborators show strong cosmic censorship can fail in de Sitter

Going further

Texts. Misner, Thorne and Wheeler, Gravitation (1973), Ch. 33–34 (Reissner-Nordström and Kerr global structure) · Wald, General Relativity (1984), Ch. 6, 12 (the Cauchy problem and cosmic censorship) · Hawking and Ellis, The Large Scale Structure of Space-Time (1973), Ch. 5–6 · Poisson, A Relativist's Toolkit (2004), Ch. 5 (mass inflation) · Carroll, Spacetime and Geometry (2004), Ch. 6.

Papers. Simpson and Penrose, "Internal instability in a Reissner-Nordström black hole", Int. J. Theor. Phys. 7, 183 (1973) · Poisson and Israel, "Internal structure of black holes", Phys. Rev. D 41, 1796 (1990) · Ori, "Inner structure of a charged black hole", Phys. Rev. Lett. 67, 789 (1991) · Dafermos, "Stability and instability of the Reissner-Nordström Cauchy horizon", Commun. Pure Appl. Math. 58, 445 (2005) · Cardoso, Costa, Destounis, Hintz and Jansen, "Quasinormal modes and strong cosmic censorship", Phys. Rev. Lett. 120, 031103 (2018). The horizon radii and surface gravities follow directly from the Reissner-Nordström lapse $f(r) = 1 - 2M/r + Q^2/r^2$ ; the inner surface gravity $\kappa_- = \sqrt{M^2-Q^2}/r_-^2$ is obtained from $f'(r_-)$ and satisfies κ₋ > κ₊ for all 0 < |Q| < M. The blueshift relation $U \propto -e^{-\kappa_- v}$ and Price tails $\Phi \sim v^{-p}$ are standard results. Mass inflation was established by Poisson and Israel (1990) and Ori (1991); the H¹ criterion and its de Sitter violation follow Christodoulou and Cardoso et al. (2018), with rigorous results from Dafermos (2005) and Dafermos–Luk. The β = 0.58 example uses illustrative rates in the reported range; the surface-gravity values were computed from equation (3) at M = 1.

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

Ask the tutor: Ask me to walk through why the inner-horizon surface gravity κ₋ — and not the Hawking value κ₊ — is the exponent that decides whether determinism survives inside a black hole.

Is a Cauchy horizon the same thing as an event horizon?

No — they answer different questions. An event horizon is the boundary of the region from which light can never escape to infinity: it is about what can get out. A Cauchy horizon is the boundary of the region that initial data can predict: it is about what the equations can determine. In Schwarzschild there is no inner horizon at all. In a charged or rotating black hole the event horizon sits at r₊ and the Cauchy horizon at the inner radius r₋, and it is only the inner one, with its larger surface gravity κ₋, that threatens determinism. Confusing the two misplaces where the instability actually lives.

Why does slowly decaying radiation become infinitely blueshifted?

Because the Cauchy horizon is reached only in the infinite future of advanced time, v → ∞, while a coordinate smooth on the horizon compresses that whole infinite range into a finite one through U ∝ −e^(−κ₋v). A field that fades gently as a power of v is read against a clock running exponentially fast near the horizon. The local frequency an infalling observer measures is the advanced-time frequency times |dv/dU|, which diverges, so the energy density is amplified without limit. The power-law decay is real but hopelessly outmatched, which is why the inner-horizon surface gravity κ₋, not the tail exponent, controls the outcome.

If the metric stays continuous, why can't an observer just fly through?

Continuity of the metric (C⁰) means the geometry has finite limits at the horizon, and in that narrow sense a small body could cross with finite total tidal distortion — which is why the singularity is called weak. But "flying through" requires the spacetime to continue as a solution of the Einstein equations on the far side, and that demands more than continuity: the connection coefficients must be locally square-integrable (the H¹ condition). Generically they are not, so there is no weak solution to fly into — the analytic "other universe" of the textbook diagram does not exist for a perturbed black hole. The observer meets the ragged edge of predictability itself.

Does mass inflation change the black hole's mass as measured from outside?

No. The ADM mass M — what a distant astronomer weighs by watching orbits — is fixed by the asymptotic geometry and is untouched. What inflates is a quasi-local mass, the Misner-Sharp function m(r), which measures the energy inside a sphere of given area radius deep in the hole. Near the Cauchy horizon this internal energy runs to infinity because the blueshifted ingoing flux and the backscattered outgoing flux amplify one another, but none of it leaks out to change M. The inflation is an interior phenomenon that curves spacetime into a singularity; from outside the hole looks exactly as massive as ever.

Why does a cosmological constant change the censorship verdict?

Because censorship is a race between decay and blueshift, and Λ changes the decay. In an asymptotically flat spacetime, fields leave feeble power-law (Price) tails v^(−p); the inner-horizon blueshift overwhelms them, mass inflation wins, and censorship holds. Turn on a positive Λ and the exterior develops a cosmological horizon that replaces those tails with exponential quasinormal decay e^(−αv). Now the softening rate α is a finite number that can be tuned — by approaching extremality — until β = α/κ₋ exceeds 1/2, at which point the field continues across the horizon in H¹ and the Christodoulou form of censorship fails. The constant changes the exterior spectrum that feeds the interior.

Is the charged Reissner-Nordström model relevant, given real black holes spin instead?

Real astrophysical black holes carry negligible charge but substantial spin, so Kerr is the physical case. Reissner-Nordström is used because it shares the crucial global feature — a smooth inner Cauchy horizon with its own surface gravity κ₋ — while being spherically symmetric and far more tractable. The blueshift, the mass-inflation cross-flow, and the null weak singularity all carry over to Kerr, as the Dafermos-Luk work established. The one delicate point is the de Sitter verdict: charged de Sitter models can violate the H¹ criterion whereas rotating models appear to restore it, so the astrophysical conclusion is drawn from Kerr, with Reissner-Nordström as the transparent laboratory.