Degenerate Killing Horizons

A degenerate Killing horizon is a smooth Killing horizon whose normal generator has zero surface gravity. The condition makes the Killing parameter affine, raises the transverse order of the generator’s norm, and removes the regular boost fixed point that would be a bifurcation surface. None of these facts requires a particular exact black-hole solution. Field equations enter only when one asks which horizon data can actually occur.

Helpful background. Exact Solutions of Einstein’s Equations supplies the extremal Reissner–Nordström and Kerr benchmarks; Causality and Global Structure supplies completeness and extension language; The Einstein Equation supplies the constraints on horizon data; Black-Hole Perturbation Theory treats the dynamical effect of the lost redshift; Black-Hole Geometry and Mechanics supplies Killing horizons, generators, and surface gravity.

Degeneracy in a horizon-regular chart

Let $\mathcal H$ be a smooth null hypersurface and let $K^a$ be a nontrivial Killing field that is normal to it. Choose a spacelike cross-section $H$ and coordinates $x^A$ on $H$. Propagate $x^A$ along the Killing orbits, use their parameter $v$, and shoot transverse affinely parametrized null geodesics with parameter $r$. With $\mathcal H=\{r=0\}$, $K=\partial_v$, and $g_{vr}=1$, the metric takes the Gaussian-null form

$$ds^2=2\,dv\!\left(dr+r h_A(r,x)dx^A +\frac12 r f(r,x)dv\right)+\gamma_{AB}(r,x)dx^A dx^B.$$

Killing symmetry makes all coefficients independent of $v$. On the horizon, the norm-gradient identity

$$\nabla_a(K^2)=-2\kappa K_a$$

gives $f(0,x)=-2\kappa$. If $\kappa=0$, smoothness implies $f(r,x)=rF(r,x)$, and hence

$$ds^2=2\,dv\!\left(dr+r h_A(r,x)dx^A +\frac12 r^2F(r,x)dv\right)+\gamma_{AB}(r,x)dx^A dx^B.$$

Thus $K^2=r^2F$: the norm has at least a double transverse zero. This argument, including the coordinate construction and the smooth factorization, is given in Kunduri and Lucietti 2013, §2.1, pp. 11–12, Eqs. (5)–(7). The sign of $F$ depends on the side of the horizon and the chosen transverse orientation; degeneracy is the order of vanishing, not the sign.

The intrinsic metric $\gamma_{AB}(0,x)$ is Lie-dragged by $K$. Accordingly a Killing horizon has vanishing expansion and shear. Degeneracy does not force $\gamma_{AB}$ to be round, $h_A$ to vanish, or $F$ to be constant. Those are field-equation and symmetry questions addressed by near-horizon classification.

Killing time becomes affine time

For any Killing horizon,

$$K^b\nabla_bK^a=\kappa K^a.$$

Let $v$ be the Killing parameter and $\lambda$ an affine parameter on one generator. Since $dx^a/dv=K^a$, their relation obeys

$$\frac{d^2\lambda}{dv^2}=\kappa\frac{d\lambda}{dv}.$$

If $\kappa\ne0$, then $\lambda-\lambda_0=C e^{\kappa v}/\kappa$; one end of an infinite Killing-time range can occur at finite affine parameter. If $\kappa=0$, then

$$\lambda=Cv+\lambda_0.$$

The Killing parameter is already affine, up to a linear change. If the Killing flow exists for every $v\in\mathbb R$ on a generator, that generator is affine-complete in both directions. Completeness is an extra global assumption: the local equation $\kappa=0$ does not by itself extend the range of $v$.

Why a regular bifurcation surface is impossible

At the bifurcation surface $B$ of a smooth bifurcate Killing horizon, the generator vanishes and its derivative is the infinitesimal boost in the normal two-plane:

$$K^a\big|_B=0, \qquad \nabla_aK_b\big|_B=\kappa\epsilon_{ab}, \qquad \epsilon_{ab}\epsilon^{ab}=-2.$$

Suppose a nontrivial degenerate generator admitted such a regular $B$. Then $\kappa=0$ would make both $K^a$ and $\nabla_aK_b$ vanish at every point of $B$. A Killing field is determined by its value and first derivative at one point: along geodesics these data satisfy a linear propagation equation obtained from $\nabla_a\nabla_bK_c=R_{cbad}K^d$. Zero initial data therefore give the zero Killing field throughout the connected domain, a contradiction.

So every nontrivial smooth bifurcate Killing horizon is nondegenerate. This is also why the Rácz–Wald enlargement theorem assumes a nonzero constant surface gravity Rácz and Wald 1996, §4, Theorem 4.2, pp. 548–550. An extremal metric can be smooth across its future horizon; what it lacks is the ordinary two-branch boost fixed point.

Infinite static throats: a consequence, not a definition

For a static metric written near a horizon as

$$ds^2=-f(R)dt^2+\frac{dR^2}{f(R)}+q_{AB}(R,x)dx^A dx^B,$$

a nondegenerate root has $f\sim 2\kappa(R-R_H)$ and lies at finite proper distance on a $t={\rm constant}$ slice. A degenerate double root has $f\sim C(R-R_H)^2$, so

$$\ell\sim\frac{1}{\sqrt C}\int\frac{dR}{|R-R_H|}$$

diverges logarithmically. This is the familiar infinite extremal throat. It is not a curvature singularity, and the argument is tied to a static slicing; it should not be promoted to a definition for every rotating or nonstatic degenerate horizon.

Nor is every degenerate Killing horizon a black-hole event horizon. The Poincaré horizon of an $\mathrm{AdS}_2$ factor supplies a local counterexample: it has the same quadratic norm behavior but need not bound a black-hole region in a specified global spacetime.

A degenerate horizon is not merely a parameter limit

The definition concerns one smooth pair $(\mathcal H,K)$, not a sequence of nondegenerate solutions. A family may have $\kappa(\delta)\to0$ as a parameter $\delta$ approaches a boundary, but the associated bifurcation surfaces, static slices, and normalized horizon coordinates need not converge uniformly. In Reissner–Nordström, every subextremal member has a finite-distance bifurcation sphere on an appropriate maximal slice, whereas the extremal static slice develops the logarithmically infinite throat just calculated. The endpoint structure changes at the limit.

This distinction gives a reliable comparison procedure. First identify the horizon and normalized generator in each member. Next evaluate the invariant $\kappa$ and the order of the zero of $K^2$ in a chart regular at the horizon. Finally compare global extensions only after checking that the coordinate maps and affine ranges themselves converge. Sending a parameter to its extremal value inside a Kruskal coordinate built from $e^{\kappa v}$ fails the last test: the coordinate definition becomes singular precisely when $\kappa\to0$. The regular degenerate Gaussian-null chart must be constructed separately.

Conversely, a smooth degenerate Killing horizon need not lie in any known one-parameter family. Its local horizon data are meaningful even when no subextremal deformation, asymptotic end, or conserved mass has been supplied. This is why the zero-surface-gravity definition precedes family-specific extremality bounds.

Solved diagnostic: extremal Reissner–Nordström

Let $L=GM$ and introduce an ingoing coordinate $v$ for the extremal Reissner–Nordström solution:

$$ds^2=-\left(1-\frac{L}{R}\right)^2dv^2+2\,dv\,dR +R^2d\Omega_2^2.$$

Set $r=R-L$. Identify the Gaussian-null data and determine the proper distance from $R=L+\rho_0$ toward the horizon on a static slice.

Solution

Here $K=\partial_v$, $h_A=0$, and

$$F(r)=-\frac{1}{(L+r)^2}, \qquad \gamma_{AB}(r)=(L+r)^2\Omega_{AB}.$$

Therefore $K^2=-r^2/(L+r)^2$, so $\partial_rK^2|_{r=0}=0$ and $\kappa=0$. Moreover $K_a|_{r=0}=(dR)_a$, confirming that $K$ is normal to the horizon. On a static slice,

$$\ell=\int\frac{dR}{1-L/R} =\int\left(1+\frac{L}{R-L}\right)dR,$$

which contains $L\ln(R-L)$ and diverges as $R\to L$. The horizon is regular in the ingoing chart even though the static slice develops an infinite throat.

The near-horizon scaling limit now uses the quadratic Gaussian-null form. Questions about decay, conserved transverse derivatives, and backreaction remain with Black-Hole Perturbation Theory.

References

• Kunduri, Hari K., and James Lucietti. “Classification of Near-Horizon Geometries of Extremal Black Holes.” Living Reviews in Relativity 16, 8 (2013). DOI.
• Rácz, István, and Robert M. Wald. “Global Extensions of Spacetimes Describing Asymptotic Final States of Black Holes.” Classical and Quantum Gravity 13 (1996): 539–553. DOI. Open PDF.