Inner Horizons and Quantum Instability

The inner horizon of an exactly stationary charged or rotating black hole is a Cauchy horizon: data on the original Cauchy surface no longer determine a unique continuation beyond it. Perturbations change that idealized picture. Radiation that decays outside can be exponentially blueshifted inside, driving mass, curvature, or a renormalized quantum stress tensor large. The ordering of those effects is not universal; it depends on exterior decay, surface gravity, state, field content, regularity criterion, and whether backreaction has been included.

Helpful background. Semiclassical Gravity supplies backreaction and renormalized stress tensors; Quantum Information supplies the determinism question; Quantum Fields in Curved Spacetime supplies Hadamard-state observables; Black-Hole Information supplies the asymptotic unitarity problem; Quantum Gravity supplies candidate completions; and Causality and Global Structure supplies Cauchy horizons and extension classes.

Why a decaying tail can become singular

Let $v$ be advanced time and let $V$ be a dimensionful coordinate regular at a subextremal inner horizon, writing $\kappa_-\gt0$ for the magnitude of its surface gravity. For a fixed scale $V_0\gt0$, choose near the horizon

$$V=-V_0\exp(-\kappa_-v), \qquad \partial_V=\frac{\exp(\kappa_-v)}{\kappa_-V_0}\partial_v.$$

Suppose an incoming field tail behaves schematically as $\partial_v\phi\sim v^{-p}$. A freely falling derivative then scales, up to a constant prefactor, like $\partial_V\phi\sim \exp(\kappa_-v)v^{-p}$. Polynomial decay loses to the exponential blueshift. The local energy density can therefore grow roughly as $\exp(2\kappa_-v)v^{-2p}$ even though the flux measured at the event horizon tends to zero.

Exponential exterior decay gives a different competition. If instead $\partial_v\phi\sim e^{-\alpha v}$, then $\partial_V\phi\sim e^{(\kappa_- -\alpha)v}$. The pointwise derivative diverges for $\alpha\lt\kappa_-$ and stays bounded at this schematic order for $\alpha\ge\kappa_-$. Integrability and nonlinear extendibility require their own thresholds. This is why a power-law asymptotically flat tail cannot simply be imported into a de Sitter problem.

This estimate identifies an amplifier, not a complete solution. In spherical Einstein–Maxwell–scalar gravity, counterstreaming ingoing and outgoing fluxes make a charge-renormalized mass aspect diverge: mass inflation. Poisson and Israel derive the mechanism in a perturbed charged interior in Poisson and Israel 1989, pp. 1663–1666 and develop its geometry in Poisson and Israel 1990, §§ II–IV, pp. 1796–1804. Ori’s analytic model shows how a null curvature singularity can replace the smooth Cauchy horizon without immediately collapsing its area radius to zero Ori 1991, pp. 789–792.

What nonlinear classical results actually establish

The strongest rigorous statements name a matter model and extension class. For spherical Einstein–Maxwell–real-scalar evolution, Dafermos proves Cauchy- horizon formation and mass inflation for specified characteristic data. A continuous metric extension may coexist with blowup strong enough to obstruct a smoother classical solution Dafermos 2005, main theorems, pp. 445–504. This is not a contradiction: $C^0$ extendibility asks less than $C^2$ extendibility, and curvature contains two derivatives of the metric.

Those theorems do not by themselves cover vacuum Kerr, one-ended stellar collapse, quantum matter, or an evaporating geometry. The exact division of classical responsibility is developed in Cauchy Horizons in Charged and Rotating Models. Here the result is input to a different question: before quantum gravity is invoked, which semiclassical observable first leaves its domain of validity?

Three candidate thresholds must be kept separate:

1. a classical mass aspect or curvature component becomes unbounded;
2. $8\pi G\lvert\langle T_{\hat a\hat b}\rangle\rvert$ in a declared orthonormal frame becomes comparable to the background curvature scale;
3. smeared, renormalized stress fluctuations induce metric fluctuations large enough to invalidate the mean-field geometry at the resolution of interest.

The first threshold can occur while the metric remains continuous. The second requires renormalization and a state. The third requires a noise kernel and the gravitational response; a variance divided by a possibly vanishing mean squared is not a universal criterion. None alone derives the geometry after breakdown.

Quantum fields sharpen, but do not finish, the verdict

For a scalar field on Reissner–Nordström–de Sitter, Hollands, Wald, and Zahn analyze Hadamard states and find a leading divergence of the renormalized stress tensor at the Cauchy horizon under their generic conditions. Its leading coefficient is insensitive to the choice among the admissible initial Hadamard states considered, so changing a regular exterior state does not generally remove the instability Hollands, Wald, and Zahn 2020, §§ 2–5. This is powerful fixed-background evidence precisely in a regime where rapid de Sitter exterior decay can weaken some classical linear instabilities.

It is still not a self-consistent quantum interior. Once $8\pi G\langle T_{\hat a\hat b}\rangle$ rivals the characteristic curvature scale in a regular frame, the prescribed metric used to compute the expectation value is no longer controlled. A divergence can signal a singular backreacted solution, a shifted inner horizon, a spacelike endpoint, or entry into a nonperturbative regime. Choosing among those outcomes requires solving the coupled dynamics with controlled initial data. A calculation of the mean also does not establish that fluctuations are small.

Results and open questions through 2026-08-19

This discussion reflects results available through 2026-08-19. A conservative position says the common result is only loss of the stationary Cauchy horizon as a smooth predictable boundary; the post-breakdown geometry is unresolved. A more permissive position reads classical mass inflation and quantum stress growth together as evidence that generic inner horizons become singular. These competing positions agree on the amplifier but differ on how far it may be extrapolated.

Evidence or calculation	Result	Scope or limitation
Perturbative counterstreaming model in Poisson and Israel 1990, §§ II–IV, pp. 1796–1804	Exponential blueshift drives an internal mass parameter large.	Charged spherical model with assumed fluxes; it is not a generic rotating quantum solution.
Nonlinear analysis in Dafermos 2005, main theorems, pp. 445–504	A Cauchy horizon, continuous extension, and mass-inflation blowup can coexist.	Classical spherical Einstein–Maxwell–scalar hypotheses; regularity conclusions must not be exported unchanged.
Quantum-field calculation in Hollands, Wald, and Zahn 2020, §§ 2–5	A generic leading stress-tensor divergence appears at the Reissner–Nordström–de Sitter Cauchy horizon.	Fixed background, selected field theory and state class; no solved semiclassical backreaction.
Exact stationary Reissner–Nordström or Kerr geometry	The unperturbed inner horizon is smooth.	Nongeneric absence of perturbing tails is an adverse check, not evidence that a dynamical horizon is stable.

The adverse comparison with the exact stationary solution is instructive. It removes the incoming tail and therefore removes the mass-inflation source; it does not refute a generic instability claim. Conversely, adding a tail to a spherical charged model does not prove the same exponent, singularity type, or quantum endpoint for Kerr. Regime and formulation are part of the conclusion.

What a completion must calculate

A credible quantum completion should evolve a state from the event horizon to the would-be inner horizon, report renormalized one- and two-point stress observables, and identify when their backreaction invalidates the starting geometry. It should then propagate gauge-invariant or relational observables through the replacement region and show whether exterior evolution remains a normalized linear channel. Stability must be tested against nonspherical perturbations and against small changes of the initial quantum state.

The unresolved issue is not whether blueshift exists, but what replaces the semiclassical Cauchy horizon after control is lost. The conclusions could change after a backreacted quantum calculation evolves generic Hadamard data through that region, or a theorem changes the relevant regularity conclusion for rotating interiors, or an adverse model keeps every stress observable and fluctuation bounded despite a nonzero physical tail.

Exercises

An event-horizon tail obeys $\partial_v\phi=A v^{-4}$ and the inner horizon has $\kappa_-\gt0$. Using the kinematic estimate above, decide whether a freely falling squared derivative remains bounded as $v\to\infty$, and state one conclusion that the estimate does not justify.

Solution

The regular derivative scales as $\partial_V\phi\sim A\exp(\kappa_-v)v^{-4}/(\kappa_-V_0)$, so its square grows as $A^2\exp(2\kappa_-v)v^{-8}/(\kappa_-V_0)^2$ and is unbounded for every positive $\kappa_-$. This establishes the blueshift competition within the assumed fixed near-horizon geometry. It does not determine the backreacted causal structure, prove a Kerr result, or show whether the metric remains continuous; those require the coupled equations and an explicit extension criterion.

References

• Dafermos, Mihalis. 2005. “The Interior of Charged Black Holes and the Problem of Uniqueness in General Relativity.” Communications on Pure and Applied Mathematics 58, 445–504. arXiv:gr-qc/0307013.
• Hollands, Stefan, Robert M. Wald, and Jochen Zahn. 2020. “Quantum Instability of the Cauchy Horizon in Reissner–Nordström–de Sitter Spacetime.” Classical and Quantum Gravity 37, 115009. arXiv:1912.06047.
• Ori, Amos. 1991. “Inner Structure of a Charged Black Hole: An Exact Mass-Inflation Solution.” Physical Review Letters 67, 789–792. DOI.
• Poisson, Eric, and Werner Israel. 1989. “Inner-Horizon Instability and Mass Inflation in Black Holes.” Physical Review Letters 63, 1663–1666. DOI.
• Poisson, Eric, and Werner Israel. 1990. “Internal Structure of Black Holes.” Physical Review D 41, 1796–1809. DOI.