Computational Black-Hole Perturbation Theory

Computational black-hole perturbation theory turns accepted linear field equations into finite, checkable spectra, Green functions, and waveforms. The central discipline is to keep analytic branch choices and numerical evidence inseparable: a small residual has meaning only for the intended horizon, infinity, harmonic, and Fourier conventions. This chapter follows one chain from separated boundary conditions through time evolution and resonance solvers to sourced radiation, while leaving the continuum derivations and physical interpretation to the companion volumes.

Helpful background. Formulations of General Relativity supplies the geometric variables, Ordinary Differential Equations supplies boundary-value and evolution methods, Black-Hole Perturbation Theory develops the master equations and spectra, and Gravitational Radiation and Waveforms supplies the radiative observables.

Enter this chapter

The common computational object is a linear operator on a stationary black-hole background, accompanied by a complete convention package. With $e^{-i\omega t}$, physical quasinormal decay requires $\operatorname{Im}\omega\lt 0$. Horizon-ingoing and infinity-outgoing behavior must be translated consistently into asymptotic factors, recurrence coefficients, spectral boundary rows, and time-domain extraction. Berti, Cardoso, and Starinets 2009, §§3–4 surveys the continuum resonance setting; these pages focus on how numerical methods can fail or corroborate it.

Let $M_{\rm phys}$ be the physical Schwarzschild mass and $M_{\rm geom}=G M_{\rm phys}$ its geometric length. The recurring axial gravitational fundamental-mode benchmark is

$$G M_{\rm phys}\omega=M_{\rm geom}\omega \simeq0.3736715-0.0889625i.$$

This precision follows from converting Leaver 1985, Table 1, p. 293, where $2M_{\rm geom}=1$. It is useful precisely because a wrong mass normalization, Fourier sign, angular label, or endpoint branch moves the answer in a recognizable way. It is not sufficient by itself: each method also needs an equation residual, boundary or jump residual, and an independent representation.

Route through the six computations

Begin with Frequency-Domain Separation and Boundary Conditions, which factors endpoint behavior and makes the physical branches explicit. Then Time-Domain Master-Equation Evolution develops a causal evolution with grid refinement, boundary movement, QNM fits, and late-time tail checks.

Continued-Fraction Quasinormal Modes turns a minimal recurrence solution into a complex root condition. Spectral Quasinormal-Mode Solvers computes a larger spectrum but must distinguish converged modes from nonnormal spectral pollution through off-grid residuals, precision changes, and an independent solver.

Green Functions, Branch Cuts, and Tails reconstructs prompt, pole, and cut contributions and checks them against time evolution. Finally, Point-Source Integration and Perturbative Waveforms treats delta and delta-derivative sources through analytic jumps and flux normalization. This order moves from homogeneous boundary data to the full sourced retarded response.

A practical first calculation is the axial Schwarzschild $\ell=2$ mode. The frequency-domain page states its potential and boundary phases, the continued-fraction page supplies the actual recurrence coefficients and backward evaluation, and the spectral page derives the factored polynomial operator from the same equation. Agreement then tests two numerical representations. For time-domain tail studies, switch explicitly to the scalar potential and reduced field defined on that page: the scalar monopole tail and the gravitational quadrupole frequency belong to different wave equations.

Common checks, different failure modes

Every computation should pin its environment, background, harmonic labels, radial and angular conventions, precision, endpoint series, numerical controls, and code revision. Refinement must vary at least three relevant controls: resolution or tolerance, boundary placement or expansion order, and precision or transform window. Outputs include unrounded modes or waveforms, residuals, convergence tables, and sufficient metadata for a clean rerun. The pages specify such interfaces but do not claim an absent program was executed.

The independent check changes method: shooting versus collocation, continued fraction versus a spectral pencil, contour reconstruction versus causal evolution, or analytic jumps versus a resolved source treatment. Shared symbolic coefficients should be identified because they create correlated error. Leaver 1985, pp. 286–293 provides a benchmark construction whose normalization must be translated before comparison.

Failure signatures differ. A frequency-domain root may solve the wrong boundary branch; a time waveform may be clean until an outer-boundary reflection arrives; a matrix eigenvalue may be a stable-looking pseudospectral artifact; a branch-cut quadrature may carry an erratum-level sign error; and a smeared point source may converge to the wrong jump. No single residual excludes all of these possibilities.

What the chapter establishes

After completing the sequence, a reader should be able to design a perturbative computation as a tuple of equation, conventions, domain, boundary or source data, algorithm, refinement sequence, independent check, and rejection criteria. The chapter does not derive the Regge–Wheeler or Teukolsky equations, prove resonance completeness, model nonlinear merger ringdown, or perform observational inference. It establishes the numerical bridge needed to make a bounded perturbative claim reproducible.

Review the chapter

Diagnostic. A spectral solver, continued-fraction solver, and time-domain fit agree on a Schwarzschild frequency to six digits. All three use an angular eigenvalue file produced by one symbolic script. Does three-way agreement provide three independent checks?

Solution

Not fully. The algorithms test different radial and temporal representations, but the shared angular file creates one common failure channel. Verify the angular eigenvalue independently—for Schwarzschild, recover the analytic $a\omega=0$ value—and then check each method’s own residuals, resolution sequence, precision sweep, and boundary behavior. Agreement becomes substantially stronger once the shared input is tested rather than merely repeated.

References

• Berti, E., Cardoso, V., and Starinets, A. O. (2009). “Quasinormal modes of black holes and black branes.” Classical and Quantum Gravity 26, 163001. arXiv.
• Leaver, E. W. (1985). “An analytic representation for the quasi-normal modes of Kerr black holes.” Proceedings of the Royal Society A 402, 285–298. DOI.