Linear Stability of Schwarzschild

Linear stability of Schwarzschild means more than decay of a convenient master variable. A solution of the linearized vacuum Einstein equations must remain bounded on the full exterior and approach a perturbation tangent to the Kerr family after pure-gauge freedom has been fixed. The surviving mass and angular- momentum variations are physical stationary zero modes; requiring the metric perturbation itself to decay to zero would incorrectly discard them. The theorem therefore combines gauge normalization, decay of gauge-invariant curvature, and reconstruction of the metric.

Required background. Analysis for Mathematical Relativity provides the weighted estimates and Sobolev control, while Partial Differential Equations and Green Functions provides hyperbolic evolution and energy methods.

Helpful background. Einstein Equations fixes the linearized system, Causality and Global Structure fixes the exterior domain, and Black-Hole Geometry and Mechanics together with Black-Hole Perturbation Theory provides the Schwarzschild and master-equation setting.

Linearized vacuum geometry

Let $g_M$ be the four-dimensional Schwarzschild metric of mass $M>0$ on its domain of outer communications, including the future event horizon. A smooth one-parameter vacuum family $g(\varepsilon)$ gives $h=\left.\partial_\varepsilon g\right|_0$ satisfying

$$D\operatorname{Ric}_{g_M}(h)=0.$$

An infinitesimal diffeomorphism changes $h$ by $h\mapsto h+\mathcal L_Xg_M$. Such pure-gauge solutions can be large in a coordinate norm while representing no change of geometry. There are also non-gauge stationary solutions: differentiating Schwarzschild with respect to $M$ gives a mass mode, while differentiating Kerr at zero rotation gives the three components of infinitesimal angular momentum.

The theorem begins with sufficiently regular, suitably decaying initial data for the linearized constraints on a hypersurface crossing the horizon and extending to spatial infinity. It concerns the resulting solution throughout the future exterior. In a controlled double-null gauge, one may add a residual pure-gauge solution determined by the initial data so that the normalized perturbation is uniformly bounded and converges at inverse-polynomial rates to a linearized Kerr solution. This is the precise content of Dafermos, Holzegel, and Rodnianski 2019, Main Theorem, pp. 11–19, not a statement that every component of the unnormalized $h$ vanishes.

Radiative and stationary sectors

Spherical harmonics expose the necessary quotient. The $\ell=0$ sector contains the variation of mass. Odd $\ell=1$ modes contain infinitesimal Kerr angular momentum; the remaining dipole freedom is gauge. Radiative degrees of freedom start at $\ell=2$. Their extreme null-curvature components satisfy spin-$\pm2$ Teukolsky equations, but those equations do not directly have the most useful positive energy.

A Chandrasekhar-type transformation produces quantities with Regge–Wheeler-type equations. For a concrete normalization, the odd-parity master amplitude $\Phi_{\ell m}(t,r)$, $\ell\ge2$, obeys the $1+1$ equation

$$\left(-\partial_t^2+\partial_{r_*}^2-U_\ell(r)\right)\Phi_{\ell m}=0,$$

where

$$\begin{aligned} f(r)&=1-\frac{2GM}{r},\qquad \frac{dr_*}{dr}=f^{-1},\\ U_\ell(r)&=f(r)\left[\frac{\ell(\ell+1)}{r^2}-\frac{6GM}{r^3}\right]. \end{aligned}$$

$M$ is physical mass and $r\gt2GM$. The potential has units of inverse length squared. It is positive for $\ell\ge2$ in the exterior: indeed $\ell(\ell+1)r-6GM\ge6r-6GM\gt0$. Thus, for decaying smooth data,

$$E[\Phi]=\frac12\int_{\mathbb R} \left(|\partial_t\Phi|^2+|\partial_{r_*}\Phi|^2 +U_\ell|\Phi|^2\right)dr_*$$

is positive and conserved on complete $t$-slices. This alone is not a nondegenerate horizon estimate or a decay theorem. Redshift, Morawetz, and far-field estimates supply those conclusions. Differential transport identities then return control to the curvature components. The classical master-equation reduction starts with Regge and Wheeler 1957, §§2–4, pp. 1064–1068; the modern geometric proof keeps the horizon, gauge, and reconstruction steps explicit.

Proof architecture from curvature to metric

The proof proceeds in a strict order.

1. Normalize the initial perturbation by separating the linearized Kerr and residual gauge sectors.
2. Derive decoupled equations for gauge-invariant extreme curvature components and transform them to Regge–Wheeler quantities.
3. Apply horizon redshift, trapping-degenerate integrated decay, and weighted estimates toward null infinity.
4. Integrate the null structure and Bianchi hierarchies to control the other curvature and connection components.
5. Reconstruct the metric coefficients, using the chosen gauge to eliminate otherwise undetermined integration functions.

The delicate point is the last step. Decay of the Teukolsky scalars alone does not exclude a nondecaying linearized Kerr perturbation, because those extreme curvature components do not see every stationary parameter mode. Nor does it control an arbitrary pure-gauge metric component. The reconstruction and normalization in Dafermos, Holzegel, and Rodnianski 2019, §§7–13, pp. 75–180 are therefore part of the stability theorem, not cosmetic choices after a curvature estimate.

What the theorem does not say

This is a linear theorem about perturbations of Schwarzschild in vacuum. It does not bound quadratic error terms, prove nonlinear Schwarzschild stability, or establish linear stability of every subextremal Kerr metric. Its displayed decay rates are robust rather than claimed sharp Price-law asymptotics. Regularity and asymptotic decay of the initial data are hypotheses, and the conclusion is made only after an explicitly controlled gauge adjustment.

A stationary mass variation is a counterexample to the claim that all regular solutions decay to zero. A pure-gauge solution is a counterexample to treating raw coordinate components as gauge-invariant observables. These examples explain why “bounded modulo gauge and convergent to linearized Kerr” is the correct conclusion.

The regularity of the mass mode is easiest to see in ingoing coordinates:

$$g_M=-\left(1-\frac{2GM}{r}\right)dv^2+2\,dv\,dr+r^2d\Omega^2, \qquad \delta M\,\partial_M g_M=\frac{2G\delta M}{r}\,dv^2.$$

This perturbation is finite at the future horizon and stationary. Apparent divergences obtained by differentiating Schwarzschild-coordinate $g_{rr}=(1-2GM/r)^{-1}$ reflect a chart that is singular at the horizon; they do not disqualify the physical mass variation.

The result nevertheless identifies the analytic ingredients a nonlinear proof would need: a coercive radiative energy, decay strong enough to integrate error terms, a finite-dimensional modulation system, and a gauge in which the constraints propagate. Linear estimates alone do not guarantee that these ingredients survive quasilinear coupling. In particular, derivative loss, motion of the trapped set, and slowly decaying gauge terms must be controlled without assuming that the evolving metric remains exactly Schwarzschild.

Check the stationary remainder

Why is a perturbation $h=\delta M\,\partial_M g_M$ compatible with linear stability even though it does not decay? What would go wrong if the theorem required $h\to0$ without modulation?

Solution

Differentiating the exact vacuum family $g_{M+\varepsilon\delta M}$ at $\varepsilon=0$ gives a regular solution of the linearized vacuum equations. It changes the ADM mass and therefore is not pure gauge under transformations that preserve the asymptotic frame. It is stationary, so it cannot decay to zero. Stability instead allows convergence to a nearby member of the stationary family: subtracting the uniquely determined linearized Kerr remainder leaves the radiative part, which does decay in the normalized gauge.

References

• Dafermos, Mihalis, Gustav Holzegel, and Igor Rodnianski. “The Linear Stability of the Schwarzschild Solution to Gravitational Perturbations.” Acta Mathematica 222 (2019): 1–214. DOI.
• Regge, Tullio, and John A. Wheeler. “Stability of a Schwarzschild Singularity.” Physical Review 108 (1957): 1063–1069. DOI.