Gauge, Modulation, and Bootstrap Closure

The nonlinear perturbation cannot be defined as “the dynamical metric minus the initial Kerr metric” and then simply estimated to zero. Part of that difference is a coordinate change, and part is motion along the Kerr family as flux changes the final mass and spin. Gauge fixing, parameter modulation, and bootstrap improvement are the three mechanisms that separate those neutral directions from genuinely radiative degrees of freedom.

Required background. PDEs and Green Functions for Gravitation supplies hyperbolic gauge reduction and bootstrap estimates. Helpful background. Black-Hole Geometry and Mechanics supplies Kerr parameters; Black-Hole Perturbation Theory supplies gauge and zero modes; Formulations of General Relativity supplies generalized harmonic evolution and constraints; Computational Relativity supplies gauge-sensitive nonlinear examples.

Hyperbolic gauge and constraint propagation

The vacuum equation is diffeomorphism invariant. Introduce a reference connection and a gauge vector

$$C^\mu =g^{\alpha\beta} \left(\Gamma^\mu_{\alpha\beta}(g) -\widehat\Gamma^\mu_{\alpha\beta}\right) -H^\mu,$$

where $H^\mu$ is a prescribed or dynamically selected gauge source. The reduced equation has principal part

$$g^{\alpha\beta}\partial_\alpha\partial_\beta g_{\mu\nu} =\mathcal Q_{\mu\nu}(g,\partial g) +\text{gauge-source terms}.$$

Einstein evolution is recovered when $C^\mu=0$ and the constraint propagation equation preserves that condition. For example, the undamped reduction $R_{\mu\nu}-\nabla_{(\mu}C_{\nu)}=0$, where parentheses include a factor $1/2$, gives $\Box C_\nu+R_\nu{}^\mu C_\mu=0$ by the contracted Bianchi identity. Zero compatible initial data for this homogeneous equation remain zero. Strong constraint damping modifies this subsystem to improve control of nonphysical modes; it cannot replace the Hamiltonian and momentum constraints on the initial slice. The Schwarzschild-family theorem uses a different geometric gauge and provides an independent nonlinear stability construction Dafermos et al. 2021, Theorem I.3.1 and §V, pp. 5–8 and 21–28, PDF.

In the full-subextremal proof, the generalized wave-map gauge is modified by finite-dimensional source terms determined together with the solution. This is necessary because large zero-energy gauge states otherwise decay too slowly Hintz 2026, §1.4, pp. 24–32, and §§6–8, pp. 113–162, PDF.

Modulation of physical zero modes

Write the solution relative to a time-dependent reference family:

$$g(t)=\Phi_t^*g_{b(t)}+h(t), \qquad b(t)=(M(t),\mathbf a(t)).$$

Here $b(t)$ contains the Kerr mass and spin, while $\Phi_t$ handles the coordinates, boost, and center-of-mass motion used for comparison. A useful schematic way to determine the parameters is to impose projection conditions

$$\langle h(t),\psi_A^*(b(t))\rangle=0$$

against adjoint zero states $\psi_A^*$, with a pairing appropriate to the function spaces. Differentiating these conditions produces modulation equations for $\dot b(t)$. The radiative flux makes the physical parameters drift, but decay should render $\dot b$ integrable so that $b(t)\to b_\infty$. This projection picture explains the purpose of modulation; it is not a replacement for the particular geometric or spectral construction in either cited proof.

The decomposition is not unique until the gauge, pairing, asymptotic frame, and zero-state basis are fixed. A coordinate center-of-mass drift must not be mistaken for emitted momentum, while a true mass change must not be gauged away. Center-of-mass position need not converge in the original frame: even a constant final boost gives linear motion, and long-range effects can produce logarithmic corrections. Hintz’s asymptotic construction explicitly tracks these effects Hintz 2026, §§11.4–11.8, pp. 238–295, PDF.

Bootstrap improvement

Measure time in units of the fixed background scale $GM_0$. Assume on a maximal interval $[0,T]$ that a high-order quadratic energy $E_N$ and a lower-order decay quantity $D_{N-2}$ satisfy

$$E_N(t)\le 2C\varepsilon^2(1+t)^\eta, \qquad D_{N-2}(t)\le 2C\varepsilon^2(1+t)^{-2+\eta},$$

with small $\eta\gt 0$. Linear estimates on the modulated background and the null structure of nonlinear errors aim to improve the factors $2C$ to $C$. Schematically,

$$E_N(t) \le C_0\varepsilon^2 +C_1\int_0^t \|\nabla h(s)\|_{L^\infty}E_N(s)\,ds.$$

If the lower-order hierarchy makes the coefficient integrable with sufficiently small integral and $C_0\lt C$, Grönwall’s inequality improves the assumed energy bound. One must improve the other bounds too and verify the local continuation criterion; only then does continuity exclude a finite endpoint of the bootstrap interval. The hierarchy is essential: top derivatives may grow mildly while lower derivatives decay strongly enough to control products.

Klainerman and Szeftel organize their slowly rotating proof through a bootstrap region and geometrically normalized structures that identify the final Kerr parameters Klainerman and Szeftel 2022, §§4.1–4.3, pp. 23–32, PDF. Hintz instead uses a Nash–Moser iteration because the tame inverse loses derivatives. Smoothing compensates for that loss as successive corrections reduce the equation’s defect. The logical purpose is analogous to bootstrap closure, but ordinary contraction mapping is not claimed Hintz 2026, §§1.5.1–1.5.4, pp. 33–42, and §13.1, pp. 318–323, PDF.

Hypotheses and proof boundary

The argument assumes four-dimensional vacuum Einstein evolution, a fixed strictly subextremal Kerr background, smooth constraint data small in high weighted regularity, and the asymptotic expansion allowed by the cited theorem. It proves exterior relaxation. It does not imply stability at extremality, with massive matter, for large perturbations, or across the complete Cauchy-horizon interior.

Dropping modulation gives an immediate failed converse: a nearby exact Kerr metric with mass $M+\delta M$ is a perfectly stable solution, but its difference from $g_M$ does not decay. Dropping gauge control gives another: $\mathcal L_Xg$ can grow in coordinates while all curvature invariants remain stationary.

Bootstrap diagnostic

Suppose $\|\nabla h(t)\|_{L^\infty}\le\varepsilon(1+t)^{-1}$ and the only estimate is $E'(t)\le C\|\nabla h\|_\infty E(t)$. Does this close a uniform bootstrap?

Solution

Direct integration gives

$$E(t)\le E(0)(1+t)^{C\varepsilon}.$$

This allows mild growth but not a uniform bound. Closure may still be possible if the bootstrap permits that growth and lower-order decay remains integrable, or if null structure, renormalization, or weighted estimates improve the effective coefficient. The bare borderline estimate alone does not prove uniform stability.

References

• Dafermos, Mihalis, Gustav Holzegel, Igor Rodnianski, and Martin Taylor, “The Nonlinear Stability of the Schwarzschild Family of Black Holes”, arXiv:2104.08222v1, 2021. Open PDF.
• Hintz, Peter, “Nonlinear Stability of Subextremal Kerr Black Holes”, arXiv:2606.28253v2, 2026. Open PDF.
• Klainerman, Sergiu, and Jérémie Szeftel, “Brief Introduction to the Nonlinear Stability of Kerr”, arXiv:2210.14400v2, 2022. Open PDF.