Mode Stability and Decay on Kerr

Mode stability and time-domain decay answer different questions. Mode stability excludes a separated solution of the Kerr wave equation that grows exponentially while satisfying the physical horizon and infinity conditions. Decay controls every solution arising from an appropriate finite-energy Cauchy datum. The first is a necessary spectral statement; the second additionally requires uniform frequency estimates, control of trapping and superradiance, and reconstruction of arbitrary wave packets. Confusing them hides precisely the estimates on which black-hole stability depends.

Required background. Analysis for Mathematical Relativity supplies spectral and energy estimates, while Partial Differential Equations and Green Functions supplies hyperbolic evolution and resolvent ideas.

Helpful background. Einstein Equations fixes the geometric setting, Causality and Global Structure fixes the exterior boundaries, and Black-Hole Geometry and Mechanics together with Black-Hole Perturbation Theory supplies Kerr geometry and separation theory.

Separated modes on subextremal Kerr

Write the four-dimensional vacuum Kerr parameters as a physical mass $M_{\mathrm{phys}}>0$, specific angular momentum $a=J/M_{\mathrm{phys}}$, and dimensionless spin $\chi=a/(GM_{\mathrm{phys}})$. The subextremal range is $|\chi|<1$. With the site convention $e^{-i\omega t}$, a scalar mode has the form

$$\psi=e^{-i\omega t}e^{im\phi}S_{m\ell}(\theta;a\omega)R_{m\ell}(r;\omega).$$

The radial and angular ordinary differential equations are coupled through a separation constant. Physical boundary conditions select a solution ingoing at the future horizon and outgoing at future null infinity. If $\operatorname{Im}\omega>0$, the time factor grows exponentially. Mode stability says that no nonzero solution obeys both boundary conditions in that upper half-plane; the real-axis statement excludes corresponding nontrivial radiative modes after stationary exceptional frequencies are treated.

Whiting’s transformation converts the separated radial problem to one with a positive integral identity, contradicting a growing mode. Its exact domain and boundary conditions are stated in Whiting 1989, theorem and pp. 1301–1305. This is an ODE theorem for each frequency, not yet a norm estimate for the Cauchy evolution.

Why absence of unstable poles is insufficient

An arbitrary solution is a superposition over continuous real frequencies and angular modes. Pointwise invertibility of the separated radial operator does not bound the inverse uniformly. Its Wronskian could become arbitrarily small along a sequence of frequencies, producing large transient responses even without an upper-half-plane zero. Quantitative mode stability supplies bounds for the inverse Wronskian on controlled frequency sets; see Shlapentokh-Rothman 2015, Theorem 1.1 and §§1.2–1.3.

Three further effects remain. First, the real-frequency spectrum is continuous, so low frequency determines polynomial tails rather than an exponential gap. Second, the trapped set makes high-frequency estimates lose coercivity or derivatives. Third, superradiant frequencies satisfy $0<\omega<m\Omega_H$ for $m\Omega_H>0$, and their stationary horizon flux has the opposite sign from the incoming flux. None of these phenomena produces an exponentially growing separated mode, yet each obstructs a naive positive- energy proof.

From spectral information to decay

The time-domain proof combines several regimes. Near the horizon, a redshift current controls the transverse derivative. In bounded frequency ranges, quantitative mode stability and frequency-localized currents handle superradiance. At high frequency, microlocal or separated estimates control the trapped set with an explicit degeneration. Near zero frequency, elliptic and transport arguments isolate stationary behavior. An $r^p$ hierarchy then pushes energy toward null infinity.

After summing the frequency pieces, one obtains bounded nondegenerate energy, integrated local-energy decay, and inverse-polynomial decay for sufficiently regular data throughout the general subextremal scalar range. The complete argument and its hypotheses are given in Dafermos, Rodnianski, and Shlapentokh-Rothman 2016, Main Theorem and §§5–13. The proof architecture shows the logical direction:

$$\text{mode stability} \;\Longrightarrow\; \text{uniform frequency control} \;\Longrightarrow\; \text{integrated decay} \;\Longrightarrow\; \text{time decay},$$

where every arrow needs additional estimates.

Boundedness, integrated decay, and pointwise decay should also remain separate. Boundedness prevents energy growth but permits a persistent compact-region field. Integrated local-energy decay rules out such persistence in an averaged sense, with degeneration at trapping. Quantitative pointwise decay then uses commuted estimates and Sobolev control, and its rate depends on regularity and asymptotic weights of the data. A statement at one level cannot be silently promoted to the next.

Limits of the comparison

These scalar results do not by themselves prove decay of linearized gravity. Spin-$\pm2$ Teukolsky fields require their own mode analysis, and metric stability also requires reconstruction, constraint propagation, removal of pure gauge, and extraction of linearized Kerr modes. Nor does mode stability establish a complete quasinormal-mode expansion: branch cuts and late-time tails are part of the continuous spectral problem.

An elementary matrix example makes the issue quantitative. In dimensionless time, consider $i\dot z=H_bz$, where

$$H_b=\begin{pmatrix}-i&b\\0&-i\end{pmatrix}, \qquad e^{-itH_b}=e^{-t} \begin{pmatrix}1&-ibt\\0&1\end{pmatrix}.$$

Both eigenvalues are always $-i$. Nevertheless $z(0)=(0,1)$ has $\|z(1)\|=e^{-1}\sqrt{1+b^2}$, arbitrarily large as $b$ increases. Each fixed matrix eventually decays; the point is that eigenvalue locations alone give no bound uniform in $b$. Similarly, estimates for separated wave equations must control constants uniformly over frequency and angular parameters before they can be summed. This model explains the gap without asserting that Kerr evolution is a finite-dimensional matrix problem. Extremality introduces a separate limitation by changing the horizon estimate.

Check the logical gap

Assume every separated Kerr scalar mode with $\operatorname{Im}\omega>0$ is absent. Does it follow that a uniform constant $C$ satisfies $E(\tau)\le C E(0)$ for all finite-energy data and all $\tau\ge0$?

Solution

No. Mode stability gives pointwise invertibility of the separated problem, but the constant in that inverse may diverge with frequency. A time-domain bound requires quantitative Wronskian or resolvent estimates, summability over angular modes, and control at zero and high frequency. It must also overcome the negative stationary horizon flux in the superradiant range and the loss at trapping. Those extra ingredients, not mode exclusion alone, produce a uniform energy estimate.

References

• Dafermos, Mihalis, Igor Rodnianski, and Yakov Shlapentokh-Rothman. “Decay for Solutions of the Wave Equation on Kerr Exterior Spacetimes III: The Full Subextremal Case $|a|<M$.” Annals of Mathematics 183 (2016): 787–913. arXiv.
• Shlapentokh-Rothman, Yakov. “Quantitative Mode Stability for the Wave Equation on the Kerr Spacetime.” Annales Henri Poincaré 16 (2015): 289–345. arXiv.
• Whiting, Bernard F. “Mode Stability of the Kerr Black Hole.” Journal of Mathematical Physics 30 (1989): 1301–1305. DOI.