Curvature-Based Wave Extraction

Curvature extraction converts four-dimensional numerical geometry on finite coordinate spheres into radiative modes. The conversion is not automatic: the Newman–Penrose tetrad, sphere geometry, harmonic basis, time coordinate, and finite-radius regime all enter before a waveform exists. A trustworthy result therefore couples the computed Weyl scalar to explicit convention tests, resolution studies, and an asymptotic interpretation.

Helpful background. Formulations of General Relativity supplies the hypersurface and tetrad decompositions; Differential Equations for Relativity supplies interpolation and time-integration tools; Black-Hole Geometry and Mechanics supplies the strong-field background; Gravitational Radiation and Waveforms defines strain, news, modes, and fluxes; and Computational Relativity supplies the shared numerical-evidence framework.

From numerical curvature to Ψ₄

Use a null tetrad $(\ell^a,n^a,m^a,\bar m^a)$ with $\ell^a$ outward directed, $n^a$ inward directed,

$$\ell\mathbin{\cdot}n=-1, \qquad m\mathbin{\cdot}\bar m=1,$$

and all other inner products zero. These conditions leave a boost freedom. Fix it asymptotically using a future timelike unit vector $v^a$, outward radial unit vector $s^a$, and oriented orthonormal tangents $e_\theta^a,e_\phi^a$:

$$\ell^a=\frac{v^a+s^a}{\sqrt2},\qquad n^a=\frac{v^a-s^a}{\sqrt2},\qquad m^a=\frac{e_\theta^a+i e_\phi^a}{\sqrt2}.$$

The tangents agree with the usual increasing spherical angles in the asymptotic frame. For the site curvature convention, define

$$\Psi_4=-C_{abcd}\,n^a\bar m^b n^c\bar m^d.$$

Let $u$ be retarded time and $r_{\mathrm A}$ the areal radius. Define the finite leading curvature and rescaled complex strain by

$$\Psi_4^0(u,\Omega):=\lim_{r_{\mathrm A}\to\infty} r_{\mathrm A}\Psi_4, \qquad H(u,\Omega):=\lim_{r_{\mathrm A}\to\infty} r_{\mathrm A}(h_+-ih_\times).$$

With the same curvature, tetrad, and polarization package,

$$\Psi_4^0=\partial_u^2H, \qquad \Psi_4=\partial_u^2(h_+-ih_\times)+O(r_{\mathrm A}^{-2})$$

in the asymptotic wave zone. These asymptotic equalities do not identify finite-radius numerical $\Psi_4$ with detector strain: one must approach an asymptotically inertial frame, choose retarded time, and integrate twice with controlled constants or low-frequency treatment. A data product declaring $r\Psi_4=-r\ddot h$ belongs to a different complete convention package and must be translated as a package before it is compared with the quantities defined here.

On a $3+1$ slice, implementations usually construct a spatial orthonormal triad from the extraction-sphere normal and two tangents, combine it with the slice normal, and contract the electric and magnetic parts of the Weyl tensor. The ordering and orientation of the tangents fix the sign of cross polarization. Gram–Schmidt orthonormalization in the coordinate basis can become ill-conditioned near angular patch boundaries, so the implementation should record its seed vectors and handedness check. Bishop and Rezzolla use the same signature, null normalization, Weyl-scalar definition, and plane-wave relation in Bishop and Rezzolla 2016, §3.3, Eqs. (95)–(103), printed pp. 22–23, PDF.

A monochromatic check

Take one asymptotic mode $H=Ae^{-i\omega u}$ with real $\omega\gt0$. Direct differentiation gives

$$N=-i\omega H,\qquad \Psi_4^0=-\omega^2H.$$

Thus curvature has two extra inverse-time factors and a minus sign relative to a sinusoidal strain. This minus sign comes from differentiation, even though the relation $\Psi_4^0=\ddot H$ itself has a plus sign. Recovering $H=-\Psi_4^0/\omega^2$ tests both sign and normalization.

At $\omega=0$ that inversion is singular. In the time domain, integrating twice determines $H$ only up to $C_0+C_1u$. The constant $C_0$ fixes a strain offset; $C_1$ shifts the news and therefore the energy flux. Low-frequency filtering requires separate checks of memory and long-duration drift.

Spheres, modes, and frame dependence

At coordinate radius $r_{\mathrm c}$, compute the physical area $A$ and define the areal radius $r_{\mathrm A}=\sqrt{A/(4\pi)}$. A mode convention must say whether the integration uses coordinate angles or angles mapped to an asymptotic frame. With spin-weight $-2$ harmonics,

$$\Psi_4^{\ell m}(t,r)= \int_{S_r}\Psi_4(t,\theta,\phi)\, {}_{-2}\bar Y_{\ell m}(\theta,\phi)\,d\Omega.$$

Here $d\Omega=\sin\theta\,d\theta\,d\phi$ is the measure of the chosen unit-sphere angular chart, not automatically $dA/r_{\mathrm A}^2$ on a distorted physical sphere. The harmonic normalization, orientation, and mode ordering belong in the data. A distorted coordinate sphere can mix modes even when its areal radius is large. Interpolate the metric and derivatives onto the sphere at a declared order, compute the sphere geometry rather than assuming $dA=r_{\mathrm c}^2d\Omega$, and test whether changing the angular chart changes the final modes beyond truncation error.

A tetrad spin rotation $m^a\rightarrow e^{i\chi}m^a$ rotates the complex phase of $\Psi_4$; a boost of $\ell^a$ and $n^a$ changes its normalization. More general finite-radius tetrad changes can mix Coulomb curvature into the nominal radiative scalar. An asymptotically correct tetrad suppresses these ambiguities with radius, but “orthonormal” alone does not guarantee the correct asymptotic normalization. Ruiz and collaborators give explicit mode and flux formulas, with their convention package, in Ruiz et al. 2008, §III.A, Eqs. (3.1)–(3.5), and §§III.B–C. Source formulas must be translated as a package rather than copying a sign in isolation.

Benchmarks that expose contamination

Begin with stationary Schwarzschild and Kerr data. In a tetrad approaching the appropriate principal tetrad, the radiative modes should vanish up to numerical and finite-radius contamination. A nonzero signal correlated with the background Coulomb scalar is a tetrad warning, not physical radiation. Rotate the angular basis by a known angle and verify the predicted spin-weight phase. Translate or rotate the coordinate grid and check invariant waveform content after applying the declared frame transformation.

The radiative benchmark is a linear Teukolsky wave or a perturbative Schwarzschild waveform with a known solution. Extract on several radii in the wave zone and at no fewer than three Cauchy and angular resolutions. For each $(\ell,m)$ mode, report amplitude and phase differences, the null or curvature constraint residual used by the evolution, and observed convergence. Vary the tetrad construction independently of the grid resolution. A convergent waveform in one tetrad does not estimate tetrad systematic error.

Finite-radius error and discretization error must be separated. At fixed radius, refine the Cauchy and extraction grids. At fixed resolved grid, vary radius and compare $r_{\mathrm A}\Psi_4^{\ell m}$ at a common retarded time. If the extraction spheres remain in the near zone, increasing polynomial fit order can produce a smooth but physically unjustified limit. Compare, where available, with characteristic extraction at future null infinity rather than treating extrapolation as truth.

Required data and failure checks

The environment record should pin the spacetime dataset and checksum, evolution and extraction revisions, metric-derivative method, interpolation order, angular grid, tetrad algorithm, harmonic library and normalization, floating-point precision, and time integrator. Inputs include coordinate and areal radii, retarded-time definition, mass scale, frame map, and the full $\Psi_4$ sign convention. Outputs should include complex modes, sphere geometry, tetrad inner-product residuals, refinement differences, and metadata sufficient to reconstruct strain later.

A clean future rerun would process one versioned perturbed-black-hole dataset over Cauchy resolutions $h$, $h/2$, $h/4$, extraction radii such as $50GM$ through $150GM$, and at least two tetrads. It would emit machine-readable phase, amplitude, null-background, and convergence comparisons. This specification is not evidence that an extraction program has run.

Reject the finite claim if tetrad norms or handedness fail, stationary Kerr produces unresolved radiation, mode phases change unexpectedly under a known basis rotation, interpolation and evolution refinement do not enter a common convergence regime, or radius dependence is comparable to the claimed physical effect. Double integration to strain is also a failure mode when arbitrary drift removal changes memory or low-frequency content; retain $\Psi_4$ and document every integration choice.

Check your understanding

A code changes only the complex tangent vector according to $m^a\rightarrow e^{i\chi}m^a$, with constant $\chi$. Its extracted $\Psi_4$ phase changes, although the spacetime and sphere are unchanged. Does this demonstrate a physical polarization rotation?

Solution

No. With the stated definition, two factors of $\bar m^a$ give $\Psi_4\rightarrow e^{-2i\chi}\Psi_4$. This is the expected spin-weight transformation under a basis rotation. Transform both waveform and polarization basis to the same frame before comparing them. A residual difference after that transformation, especially one that does not decrease with radius or resolution, would instead indicate an implementation, interpolation, or tetrad-normalization problem.

Continue with Finite-Radius Extrapolation.

References

• Bishop, Nigel T., and Luciano Rezzolla. 2016. “Extraction of Gravitational Waves in Numerical Relativity.” Living Reviews in Relativity 19, 2. arXiv.
• Ruiz, Milton, Ryoji Takahashi, Miguel Alcubierre, and Darío Núñez. 2008. “Multipole Expansions for Energy and Momenta Carried by Gravitational Waves.” General Relativity and Gravitation 40, 1705–1729. arXiv.