Asymptotic Flatness at Null Infinity

Asymptotic flatness at future null infinity replaces the statement “go to large radius along outgoing light rays” by a regular conformal boundary $\mathscr I^+$. The boundary must be null, have the topology appropriate to angular directions and retarded time, obey sufficiently regular asymptotic field equations, and—when black holes or a global symmetry action are under discussion—have a separately stated completeness property. This definition is local to the asymptotic region: it does not require every null geodesic in the interior to escape to infinity.

Required background. The Riemann Curvature Tensor supplies the Weyl and Ricci tensors used in the conformal field-equation argument.

Helpful background. Pseudo-Riemannian Metrics and Signature fixes null-normal conventions; Affine and Variational Geodesics explains affine completeness; Causality and Global Structure supplies conformal boundaries; Formulations of General Relativity provides characteristic evolution; and Exact Solutions of Einstein’s Equations supplies Minkowski and Schwarzschild tests.

A conformal definition of future null infinity

Let $(\widehat M,\widehat g_{ab})$ be a smooth, time-oriented physical spacetime. A useful four-dimensional vacuum definition says that it is asymptotically flat at $\mathscr I^+$ if there are a manifold $M$ with boundary $\mathscr I^+$, a diffeomorphism identifying $\widehat M$ with the interior $M\setminus\mathscr I^+$ near that boundary, a defining function $\Omega$, and an unphysical metric $g_{ab}$ such that

$$g_{ab}=\Omega^2\widehat g_{ab}, \qquad \Omega=0, \qquad n_a:=\nabla_a\Omega\ne0 \quad\text{on }\mathscr I^+.$$

One then specifies all of the following rather than treating them as tacit consequences:

1. $g_{ab}$ extends with a stated regularity—commonly $C^4$ for curvature balance laws—and $\Omega$ is smooth enough for the required derivatives.
2. $\mathscr I^+$ has topology $S^2\times\mathbb R$. Its $S^2$ factor labels generators and its $\mathbb R$ factor labels position along them.
3. The physical vacuum Einstein equation holds in a neighborhood of $\mathscr I^+$. A standard matter extension instead requires $\Omega^{-2}\widehat T_{ab}$ to have a $C^2$ limit there.

This particular definition, including the $C^4$ metric, topology, vacuum condition, and admissible matter weakening, is given in Ashtekar and Khera 2024, §2.1, pp. 4–5. Other applications use fewer derivatives, weighted Sobolev conditions, or polyhomogeneous expansions. Such variants can be perfectly meaningful, but their conclusions about peeling or asymptotic charges must be weakened accordingly.

Choose $\Omega\gt 0$ on the physical side. The nonvanishing-gradient condition is essential: it says that $\Omega$ is a genuine transverse coordinate near the boundary and has a simple zero, corresponding to $\Omega\sim r^{-1}$ along outgoing null directions. Merely finding a function that vanishes at large radius would not establish this regularity.

Why the boundary is null

Nullness is not an additional drawing convention. It follows from the conformal transformation of the vacuum equation. For $\widehat g_{ab}=\Omega^{-2}g_{ab}$ in four dimensions, the transformed Ricci tensor contains the leading term

$$\widehat R_{ab} =R_{ab}+2\Omega^{-1}\nabla_a\nabla_b\Omega +g_{ab}\Omega^{-1}\Box\Omega -3g_{ab}\Omega^{-2}(\nabla\Omega)^2.$$

Multiplying the vacuum equation by $\Omega^2$ and taking the regular limit $\Omega\to0$ gives $(\nabla\Omega)^2=0$ at $\mathscr I^+$. Hence $n_a$ is a null normal. A null hypersurface’s normal is also tangent to it, so

$$n^a=g^{ab}\nabla_b\Omega$$

generates the null curves ruling $\mathscr I^+$. The pullback $q_{ab}$ of $g_{ab}$ to the boundary is therefore degenerate, with signature $(0,+,+)$ and kernel generated by $n^a$. Field equations also imply that the generators preserve the conformal two-geometry in an appropriate conformal frame. These consequences are developed in Frauendiener 2004, §2.3, pp. 15–19.

The conformal completion is not unique. For any smooth positive $\omega$ at the boundary,

$$\Omega'=\omega\Omega, \qquad g'_{ab}=\omega^2g_{ab},$$

describes the same physical metric. On $\mathscr I^+$ this gives $q'_{ab}=\omega^2q_{ab}$ and $n'^a=\omega^{-1}n^a$. One can choose a divergence-free conformal frame in which $\nabla_a n^a=0$ on $\mathscr I^+$; in vacuum, a standard refinement has $\nabla_a n_b=0$ there. The remaining rescalings are constant along each generator. The universal null-infinity structure is consequently a conformal equivalence class of pairs $(q_{ab},n^a)$, not one preferred round metric together with one preferred retarded-time normalization.

Bondi coordinates and the Minkowski test

Near a sufficiently regular $\mathscr I^+$ one may introduce retarded Bondi coordinates $(u,r,x^A)$, but a Bondi expansion is a gauge realization of the boundary conditions, not their coordinate-free definition. Its determinant condition fixes the meaning of the areal radius, and the vacuum equations then organize the radial expansion and identify the shear, news, mass aspect, and angular-momentum aspect. A concrete hierarchy is presented in Compère and Fiorucci 2018, §3.1.2, pp. 72–74.

Minkowski spacetime makes the completion explicit. In outgoing coordinates,

$$d\widehat s^2=-du^2-2\,du\,dr+r^2d\omega_2^2.$$

Set $\Omega=r^{-1}$, so $dr=-\Omega^{-2}d\Omega$. The unphysical line element becomes

$$ds^2=-\Omega^2du^2+2\,du\,d\Omega+d\omega_2^2.$$

It extends smoothly to $\Omega=0$. There $n^a=\partial_u$, each cut has the unit-sphere metric, and $u\in\mathbb R$ makes every generator complete in both directions. Retarded Schwarzschild gives the equally smooth replacement $-\Omega^2(1-2GM\Omega)du^2$ in the first term. Thus a black-hole exterior can have a regular $\mathscr I^+$ even though the complete spacetime is not asymptotically simple.

Topology, regularity, and completeness are separate

The topology $S^2\times\mathbb R$ does not make $n^a$ complete. A nowhere-zero vector field on that manifold can traverse the entire coordinate range in finite parameter. In a divergence-free frame, completeness means that an affine parameter $u$ satisfying $n^aD_a u=1$ runs through an unbounded interval along every generator. For black-hole constructions, one commonly requires all of $\mathscr I^+$ to be complete. For matching with $i^0$, past completeness—$u\to-\infty$ toward spatial infinity—can suffice. For a global action of the full BMS group, both past and future completeness are normally imposed. These distinctions, including examples that are past complete but not future complete, are stated in Ashtekar and Khera 2024, §2.1, pp. 5–6.

Three other properties must not be smuggled into the definition:

• Smoothness: $C^4$ regularity supports a continuously differentiable rescaled Weyl tensor and classical peeling. Weaker differentiability may preserve only the leading part of the hierarchy. See peeling and polyhomogeneity.
• Asymptotic simplicity: regularity near $\mathscr I$ does not say that every inextendible physical null geodesic has an endpoint there. Captured rays in a black-hole spacetime do not. See black-hole-compatible variants.
• Matching: a well-defined $\mathscr I^+$ does not canonically identify its BMS group with that of $\mathscr I^-$. Such an identification requires control at spatial infinity.

Nor does a null completion control timelike infinity. Massive matter can approach $i^+$ rather than $\mathscr I^+$, and additional decay or scattering hypotheses are needed to close global balance laws. Finally, a cosmological constant changes the conformal-boundary character: for nonzero $\Lambda$, the above vacuum limiting argument does not produce the same null boundary.

Common pitfalls

Bondi gauge is not the boundary itself. Coordinate coefficients are useful only after the residual transformations and differentiability class are stated. A singular Bondi chart can fail even when another conformal chart is regular.

Boundary-generator completeness is not physical geodesic completeness. The former concerns $n^a$ in the unphysical boundary geometry. The latter concerns affinely parametrized geodesics of $\widehat g_{ab}$ in the physical interior.

A regular future boundary says nothing automatic about the past boundary. Time-asymmetric collapse can have different structures at $\mathscr I^-$ and $\mathscr I^+$. Each component and its completeness must be checked.

Exercises

For Minkowski spacetime in outgoing coordinates, use the unphysical metric

$$ds^2=-\Omega^2du^2+2\,du\,d\Omega+d\omega_2^2$$

to show directly that $\mathscr I^+=\{\Omega=0\}$ is null and find its generator. Is the generator complete?

Solution

In the $(u,\Omega)$ block, $g_{ab}$ has matrix

$$\begin{pmatrix}-\Omega^2&1\\[2pt]1&0\end{pmatrix},$$

whose inverse is

$$\begin{pmatrix}0&1\\[2pt]1&\Omega^2\end{pmatrix}.$$

The normal is $n_a=(d\Omega)_a$, so $n^an_a=g^{\Omega\Omega}=\Omega^2$, which vanishes at $\Omega=0$. Also $n^a=g^{a\Omega}=\partial_u+\Omega^2\partial_\Omega$, hence $n^a=\partial_u$ on $\mathscr I^+$. Its affine parameter is $u$ up to an affine transformation. Because $u$ ranges over all of $\mathbb R$, every generator is both past and future complete.

References

• Ashtekar, Abhay, and Neev Khera. “Unified Treatment of Null and Spatial Infinity III: Asymptotically Minkowski Space-times.” Journal of High Energy Physics 2024, no. 2 (2024): 210. DOI.
• Compère, Geoffrey, and Adrien Fiorucci. “Advanced Lectures on General Relativity.” arXiv:1801.07064 [hep-th] (2018). arXiv.
• Frauendiener, Jörg. “Conformal Infinity.” Living Reviews in Relativity 7 (2004): 1. DOI.