Curvature Scattering and Tails

A massless gravitational disturbance has a null wavefront, yet in curved spacetime its retarded field need not be confined to that wavefront. Curvature acts as a scattering potential: one part of the signal follows the direct null cone, while another samples the geometry and reaches the observation event from inside its past light cone. That second contribution is a propagation tail. This page explains its causal origin and its limits; detailed black-hole tail exponents and Green-function spectra belong to Black-Hole Perturbation Theory, while source-generated hereditary multipoles are treated separately in wave generation.

Required background. Waves, Green Functions, and Geometric Optics supplies retarded fundamental solutions and characteristic propagation; Curvature supplies the curvature couplings that scatter the field.

Helpful background. Einstein Equations supplies the linearized gravitational operator; Relativistic Cosmology supplies an important family of time-dependent curved backgrounds.

Direct and tail parts of a retarded field

Let $P$ be a normally hyperbolic wave operator acting on one component of a field, or schematically on a tensor with the necessary parallel propagators understood. Its retarded Green function satisfies

$$P_xG_{\mathrm{ret}}(x,x')=-4\pi\delta_4(x,x')$$

and vanishes unless $x$ lies in the causal future of $x'$. Within a convex normal neighborhood, where the two events are joined by a unique geodesic, introduce Synge’s world function $\sigma(x,x')$, one half the squared geodesic distance. With signature $(-,+,+,+)$, timelike separation has $\sigma<0$. The local Hadamard form is

$$G_{\mathrm{ret}}(x,x') =\Theta_+(x,x') \left[U(x,x')\delta(\sigma) +V(x,x')\Theta(-\sigma)\right].$$

The smooth coefficient $U$ multiplies the direct term on the future null cone. The smooth coefficient $V$ multiplies the tail term at timelike-related events inside that cone. The precise delta normalization can be moved between $G$, $U$, and the field equation; the support distinction cannot. The construction and the transport equations determining $U$ and $V$ are given in Poisson, Pound, and Vega 2011, §§14.1–14.5, eqs. (14.1)–(14.33).

For linearized gravity on a vacuum background, the Lorenz-gauge equation contains a Riemann-tensor potential in addition to the principal wave operator. The principal part fixes the null characteristics, while curvature enters the lower-order transport and scattering. Flanagan and Hughes derive the curved-background short-wavelength equation and explain its scale separation in Flanagan and Hughes 2005, §5, pp. 23–29, PDF. A tail therefore does not make the leading wave covector timelike: with $k_\mu=\nabla_\mu\theta$, the high-frequency wavefront still obeys $k_\mu k^\mu=0$ and follows future-directed affine null rays.

Huygens propagation and its failure

The retarded Green function of the massless wave equation in four-dimensional Minkowski spacetime has support only on the light cone. In the notation above, $U$ reduces to the flat direct coefficient and $V=0$. A sharply emitted pulse is received sharply after the null travel time. This is Huygens propagation; it is dimension dependent and is not a generic property of all wave equations.

Curvature can make $V$ nonzero. A transparent perturbative check begins with a flat operator $P_0$ and a weak, geometry-induced potential $\mathcal V$,

$$P=P_0-\mathcal V, \qquad G_{\mathrm{ret}} =G_0-\frac{1}{4\pi}G_0\mathcal V G_0+O(\mathcal V^2),$$

where the product denotes a spacetime convolution over an intermediate event $y$. The factor $-1/(4\pi)$ follows from the normalization $P_0G_0=-4\pi\delta_4$: applying $P_0$ to the correction gives $+\mathcal V G_0$, canceling the potential term in $PG_0$. If instead one defines the inverse kernel $E_0=-G_0/(4\pi)$, the familiar inverse-operator series is $E=E_0+E_0\mathcal V E_0+\cdots$.

Each factor $G_0$ is supported on a null segment, but a broken path $x'\to y\to x$ generally takes longer than the direct null path $x'\to x$. Integrating over all scattering events $y$ fills timelike-separated arrival times. Removing $\mathcal V$ removes this first scattered term and restores the flat four-dimensional result.

For perturbations of a stationary compact-object exterior, angular decomposition produces radial equations of the schematic form

$$\left(-\partial_t^2+\partial_{r_*}^2-\mathcal V_\ell(r)\right)\psi_{\ell m}=0.$$

The potential contains angular and curvature barriers. An outgoing pulse can scatter from them and return later, producing ringdown and late-time tails through different parts of the Green function. Their detailed spectral decomposition, decay powers, and horizon boundary conditions are outside this page; the universal point here is the causal support created by propagation through a nontrivial exterior.

Three effects that must not be conflated

Geometric transport follows one null ray and determines leading phase, focusing, and parallel polarization transport. It is local in the ray bundle and is accurate when wavelength is short compared with every background scale.

Propagation tails are nonlocal Green-function contributions. They superpose scattered paths and can persist after the direct pulse has passed. A local observer may receive them at timelike separation from the emission event even though no massive gravitational mode has been introduced.

Source hereditary terms arise when a radiative multipole at retarded time depends on the source’s earlier history, often because radiation scatters from the source’s own long-range field during generation. Their final waveform integral can look tail-like, but its inputs and approximation scheme belong to multipolar wave generation. The present page instead holds the emitted disturbance fixed and asks how a prescribed background propagates it.

Multiple-image lensing is different again: in geometric optics, several stationary null paths can connect source and observer, each with its own delay and Morse phase. Diffraction replaces that discrete sum by a coherent wave integral. Neither phenomenon should be labeled an interior-cone Green-function tail merely because it produces delayed structure.

Causal and approximation checks

The Hadamard expression is local. Caustics, several connecting geodesics, horizons, boundaries, and topology require a global retarded construction; a single $U\delta(\sigma)+V\Theta(-\sigma)$ chart does not encode all of them. Likewise, the Born series is explanatory only when the effective potential is weak enough for the truncation to be controlled.

Three checks remain useful in every regime. First, the retarded field must vanish outside the causal future. Second, setting curvature and other lower-order potentials to zero in four-dimensional Minkowski spacetime must remove $V$. Third, the direct singularity must remain on the null cone even when a smooth interior tail is present. A calculation that shifts the characteristic cone without changing the principal operator is internally inconsistent.

Exercise

Suppose a compact source emits a delta-like pulse at event $x'$. Compare the received support in four-dimensional flat spacetime for $V=0$ with the support when a smooth $V\neq0$ is present. Does the latter imply a subluminal graviton?

Solution

When $V=0$, convolution with the Green function samples only $U\delta(\sigma)$, so the ideal pulse reaches an observation worldline when its event is null related to $x'$. When $V\neq0$, the term $V\Theta(-\sigma)$ contributes at later events timelike related to $x'$; the observer sees a direct arrival followed by a tail. This does not imply a subluminal graviton. The singular wavefront is still governed by the principal operator and travels on the null cone. The interior signal is a coherent sum of fields scattered by the background, analogous to delayed propagation through many broken null paths.

References

• Éanna É. Flanagan and Scott A. Hughes, “The Basics of Gravitational Wave Theory,” New Journal of Physics 7 (2005) 204, arXiv:gr-qc/0501041v3, DOI, Open PDF.
• Eric Poisson, Adam Pound, and Ian Vega, “The Motion of Point Particles in Curved Spacetime,” Living Reviews in Relativity 14 (2011) 7, arXiv:1102.0529v3, DOI, Open PDF.