Blueshift at the Cauchy Horizon

Radiation falling through an event horizon can decay at late exterior time and still acquire unbounded locally measured frequency at an inner Cauchy horizon. The reason is geometric: outgoing inner-horizon generators accumulate an infinite interval of advanced time while a freely falling observer reaches the horizon in finite proper time. Exponential compression beats polynomial exterior decay.

Helpful background. Black-Hole Geometry and Mechanics supplies stationary horizons and surface gravity; Black-Hole Perturbation Theory supplies late-time tails; Formulations of General Relativity supplies characteristic variables; Causality and Global Structure supplies Cauchy horizons; Computational Relativity supplies invariant diagnostics and convergence concepts.

Frequency measured by an infalling observer

For any geometric-optics ray with wave covector $k_\mu=\nabla_\mu S$ and observer four-velocity $u^\mu$, the measured angular frequency is the scalar

$$\omega_{\rm obs}=-k_\mu u^\mu.$$

This definition is coordinate independent. Consider the outgoing branch of a subextremal stationary black hole’s future Cauchy horizon. Late ingoing radiation intersects outgoing rays approaching that branch. Let $v$ be advanced time normalized by the stationary exterior. Near the inner horizon, a regular Kruskal-like coordinate has the asymptotic form

$$V=-C e^{-\kappa_- v}, \qquad C\gt 0,$$

where $\kappa_-\gt 0$ denotes the magnitude of the inner-horizon surface gravity. Hence

$$\frac{dv}{dV}=-\frac{1}{\kappa_-V}.$$

A freely falling observer crosses $V=0$ with finite nonzero $dV/d\tau$. For an ingoing geometric-optics component with phase $S(v)$, the frequency magnitude is

$$\omega_{\rm obs} =\left|\partial_vS\right| \left|\frac{dv}{d\tau}\right|.$$

For fixed nonzero exterior phase frequency $\lvert\partial_vS\rvert$, the amplification factor is proportional to $e^{\kappa_-v}$. A nonoscillatory late-time tail need not have a single optical frequency; the field-derivative calculation below describes its blueshift without that assumption.

The sign of the Killing-horizon surface gravity depends on convention at the inner horizon; the amplification rate uses its positive magnitude. In a Reissner–Nordström benchmark, use the geometrized radii $r_\pm$ and ingoing coordinates $ds^2=-f\,dv^2+2\,dv\,dr+r^2d\Omega^2$, where $f=(r-r_+)(r-r_-)/r^2$. An outgoing radial null ray satisfies

$$\frac{dr}{dv}=\frac f2 \simeq-\kappa_-(r-r_-), \qquad \kappa_-=\frac{r_+-r_-}{2r_-^2}\gt0.$$

Thus $r-r_-\propto e^{-\kappa_-v}$: the inner horizon is approached while $v$ tends to infinity. This checks the exponential’s sign and explains its origin without assuming divergent curvature in the exact background. The geometric estimates are developed for the charged system in Dafermos 2005, §§7 and 9, pp. 23–40, PDF and for Kerr in Dafermos and Luk 2025, §§1.2.2–1.2.4, pp. 11–13, PDF.

Decay outside versus growth inside

Suppose an exterior tail along the event horizon behaves schematically as

$$\partial_v\phi\sim v^{-p}, \qquad p\gt 0.$$

The field derivative decays in the exterior normalization. A regular inner-horizon derivative instead contains

$$\partial_V\phi =\frac{dv}{dV}\,\partial_v\phi \sim \frac{v^{-p}}{V} \sim e^{\kappa_-v}v^{-p}.$$

Thus every fixed polynomial power loses to the exponential factor. This does not say that $\phi$ itself must diverge. Indeed, $\int^\infty v^{-p}dv$ converges when $p\gt 1$, so the scalar field and even metric coefficients sourced by integrable fluxes can approach finite limits while transverse derivatives blow up.

The invariant stress measured by the observer makes the physical point:

$$\mathcal E_{\rm obs}=T_{\mu\nu}u^\mu u^\nu.$$

For a scalar wave in geometric optics, $\mathcal E_{\rm obs}$ contains $\omega_{\rm obs}^2$ times the local amplitude. The energy density can therefore grow without bound. In a self-gravitating problem that growth feeds the geometry and leads to mass inflation rather than remaining a test-field effect; the original counterstreaming model makes that feedback explicit Poisson and Israel 1990, §§II–IV, pp. 1797–1807.

The argument depends on an infinite advanced-time limit, a subextremal inner horizon with $\kappa_-\gt 0$, and a nonzero tail. It does not apply unchanged at extremality, where the redshift/blueshift hierarchy degenerates, or when cosmological asymptotics replace polynomial tails by exponential decay; there the decay exponent must be compared quantitatively with $\kappa_-$ rather than assumed to win or lose. Those differences are central to modern formulations of strong cosmic censorship.

Independent checks

Three checks prevent the coordinate calculation from becoming a coordinate claim.

1. Compute $\omega_{\rm obs}=-k\mathbin{\cdot}u$, not merely a component such as $k_v$.
2. Change from $v$ to the regular coordinate $V$ and verify that the divergence moves into a physical contraction or a transverse derivative rather than disappearing.
3. Compare the exponential rate with the actual exterior decay rate in the same spacetime and field model.

The amplification also has a simple consistency limit. If the incoming tail vanished identically after some finite time, the schematic late-time contribution above would be absent. Generic scattering produces tails, but the claim of instability requires that input; the stationary metric alone cannot manufacture radiative stress.

Check your understanding

Let $\partial_v\phi=A v^{-4}$ and $V=-e^{-\kappa_-v}$ with $\kappa_-\gt 0$. Determine the behavior of $\partial_V\phi$ as $V\to0^-$ and decide whether $\phi$ itself must diverge.

Solution

Since $dv/dV=-1/(\kappa_-V)$,

$$\partial_V\phi =-\frac{A}{\kappa_-V}\,v^{-4}.$$

Using $v=-\kappa_-^{-1}\log(-V)$,

$$\left|\partial_V\phi\right| \sim \frac{\kappa_-^3\lvert A\rvert} {|V|\,|\log(-V)|^4},$$

which diverges. But $\int^\infty v^{-4}dv$ converges, so $\phi$ can tend to a finite limit. Bounded field value and divergent transverse derivative are compatible.

References

• Dafermos, Mihalis, “The Interior of Charged Black Holes and the Problem of Uniqueness in General Relativity”, Communications on Pure and Applied Mathematics 58, 445–504, 2005. Open PDF.
• Dafermos, Mihalis, and Jonathan Luk, “The Interior of Dynamical Vacuum Black Holes I: The C⁰-Stability of the Kerr Cauchy Horizon”, Annals of Mathematics 202, 309–630, 2025. Open PDF.
• Poisson, Eric, and Werner Israel, “Internal Structure of Black Holes”, Physical Review D 41, 1796–1809, 1990.