Asymptotic Flatness at Spatial Infinity

Asymptotic flatness at spatial infinity controls the geometry seen along spacelike directions and supplies the setting for ADM energy–momentum and Lorentz charges. It requires more than $h_{ij}\to\delta_{ij}$: one must identify each asymptotic end, control derivatives and constraints, restrict the transformations between asymptotic charts, and impose parity or an alternative boundary condition strong enough to make the symplectic structure and charges finite. A four-dimensional conformal description packages the same regime as direction-dependent data at a point $i^0$, whose regularity differs sharply from that of null infinity.

Required background. The Riemann Curvature Tensor supplies the curvature and derivative estimates used at the asymptotic end.

Helpful background. Pseudo-Riemannian Metrics and Signature fixes spacelike-normal conventions; Affine and Variational Geodesics distinguishes geodesic limits; Causality and Global Structure supplies conformal infinity; Formulations of General Relativity provides ADM constraints; and Exact Solutions of Einstein’s Equations supplies Schwarzschild and Kerr checks.

Asymptotically Euclidean ends and ADM data

Let $\Sigma$ be a spacelike hypersurface with induced Riemannian metric $h_{ij}$ and future-directed unit normal $u^a$, with $u^a u_a=-1$. We use

$$K_{ij}=-h_i{}^a h_j{}^b\nabla_a u_b =-\frac12\mathcal L_u h_{ij}.$$

An asymptotically Euclidean end is a component of $\Sigma\setminus C$, for some compact set $C$, equipped with a diffeomorphism to $\mathbb R^3\setminus B_R$. In the resulting asymptotically Cartesian coordinates $x^i$, with $r=(\delta_{ij}x^ix^j)^{1/2}$ and Euclidean radial normal $n^i=x^i/r$ distinct from $u^a$, a common leading class is

$$h_{ij}=\delta_{ij}+\frac{a_{ij}(n)}{r}+O_2(r^{-2}), \qquad K_{ij}=\frac{k_{ij}(n)}{r^2}+O_1(r^{-3}).$$

Here $O_s(r^{-p})$ controls up to $s$ Cartesian derivatives, with each derivative gaining one inverse power of $r$. A Hamiltonian momentum density with the sign matched to this convention is

$$\pi^{ij}=\sqrt h\,(Kh^{ij}-K^{ij})=\frac{b^{ij}(n)}{r^2}+O_1(r^{-3}),$$

possibly with an overall factor of $(16\pi G)^{-1}$. A source using $K^+_{ij}=+\tfrac12\mathcal L_u h_{ij}=-K_{ij}$ writes the same canonical density as $\sqrt h\,(K_+^{ij}-K_+h^{ij})$. Both the definition of extrinsic curvature and its algebraic appearance must be translated; changing only one reverses the physical momentum. The overall normalization affects charge formulas but not the falloff or parity discussion.

These estimates must accompany the Hamiltonian and momentum constraints and decay conditions for matter. A solution of the vacuum constraints satisfies them identically; a phase-space definition often states stronger asymptotic orders so the bulk generators converge. The basic falloffs $h_{ij}=\delta_{ij}+O(r^{-1})$ and $\pi^{ij}=O(r^{-2})$, and the need to strengthen them, are analyzed in Henneaux and Troessaert 2019, §III, pp. 5–9.

Topology is part of the data. A hypersurface may have two asymptotically Euclidean ends, as a complete Einstein–Rosen bridge does, and each end carries its own surface integrals. An end modeled on a quotient at infinity is asymptotically locally Euclidean rather than the standard ADM end. Neither small curvature nor the symbol $r\to\infty$ chooses among these possibilities.

ADM charges and allowed coordinates

For the stated end, the ADM energy is

$$E_{\mathrm{ADM}} =\frac{1}{16\pi G} \lim_{r\to\infty} \int_{S_r} (\partial_jh_{ij}-\partial_i h_{jj})n^i\,dS.$$

With the same convention, the ADM momentum is

$$P_i=\frac{1}{8\pi G}\lim_{r\to\infty} \int_{S_r}(K_{ij}-Kh_{ij})n^j\,dS.$$

Rotations and boosts carry an additional factor growing like $r$. Consequently, the Lorentz charges are more sensitive to subleading terms and parity than energy–momentum. Coordinate invariance is not granted under every diffeomorphism that tends vaguely to the identity. One declares an asymptotic chart class whose transition maps preserve the derivative estimates and boundary conditions. Rotations rotate the spatial momentum and leave the energy fixed; boosts mix energy and momentum. Translations leave the ADM four-momentum unchanged, although changing the origin shifts angular momentum by an orbital term. Transformations judged pure gauge must have vanishing generators, whereas “large” transformations with nonzero surface charges are asymptotic symmetries. A symmetry can therefore carry a nonzero generator without changing the value of every charge.

This is why a coordinate-free slogan and a coordinate calculation must meet in the middle. The chart fixes $r$, the antipodal map, and the leading sphere. The constraints and boundary conditions then ensure that changing to another allowed chart gives the appropriate Poincaré—or enlarged BMS—transformation law rather than an arbitrary shift in the charge.

Strict and twisted parity

Under the antipodal map $n^i\mapsto-n^i$, the classic Regge–Teitelboim conditions take

$$a_{ij}(-n)=a_{ij}(n), \qquad b^{ij}(-n)=-b^{ij}(n).$$

Thus the leading metric perturbation is even and the leading canonical momentum is odd in asymptotically Cartesian components. The potentially logarithmic part of the kinetic term has angular coefficient $b^{ij}\dot a_{ij}$, which is odd. Its integral over $S^2$ vanishes. The same parity structure removes dangerous leading terms in boosts and rotations. With the usual quotient by zero-charge transformations, the resulting asymptotic symmetry group reduces to Poincaré rather than retaining arbitrary supertranslations.

Strict parity is sufficient but not unique. Henneaux and Troessaert allow a twisted parity class: the opposite-parity part of the leading metric and momentum is permitted only when it has the specific form of an improper asymptotic diffeomorphism. They supplement this with stronger asymptotic constraints and a radial gauge condition. The symplectic form remains finite, while the nontrivial angle-dependent translations realize a BMS group at spatial infinity Henneaux and Troessaert 2019, §§III–VI, pp. 5–24.

The distinction is physical, not cosmetic. If one calls every supertranslation “gauge,” its charge is removed by definition. If one allows unrestricted opposite-parity coefficients, the symplectic form or Lorentz generators can diverge. Twisted parity occupies a controlled middle ground, but it is one boundary-condition choice among several; results proved in that phase space should not be attributed automatically to all asymptotically flat data.

Spatial infinity as a conformal point

The Ashtekar–Hansen spacetime description attaches a preferred point $i^0$ to the physical spacetime $(\widehat M,\widehat g_{ab})$. With $g_{ab}=\Omega^2\widehat g_{ab}$, its normalized conditions include

$$\Omega\big|_{i^0}=0, \qquad \nabla_a\Omega\big|_{i^0}=0, \qquad \nabla_a\nabla_b\Omega\big|_{i^0}=2g_{ab}\big|_{i^0},$$

together with the causal placement of $i^0$, vacuum behavior near it, $C^3$ regularity away from the point, and a special $C^{\gt 0}$ regularity at the point. This notation does not mean an ordinary Hölder exponent. The metric is continuous, while its connection can have finite, regular direction-dependent limits whose angular derivatives commute with taking the limit. The precise definition and its implications are summarized in Ashtekar and Khera 2024, §2.3, pp. 10–13.

The weakened regularity is necessary rather than a defect to be ignored: in this completion the ADM mass is stored in a radial discontinuity of the connection. Requiring $g_{ab}$ to be $C^1$ at $i^0$ would force that mass to vanish, while mere continuity would not control it. Spacelike approach directions form the unit hyperboloid $\mathcal H$ in $T_{i^0}M$, and regular direction-dependent curvature limits become smooth tensor fields on $\mathcal H$. This “blowup” of one point into directions is the bridge used in matching spatial and null infinity.

The conformal-point and Cauchy-end descriptions are related but serve different theorems. A suitable $C^{\gt 0}$ completion induces standard ADM falloffs on sufficiently regular hypersurfaces through $i^0$. Conversely, coordinate falloffs alone do not guarantee the full spacetime completion, its causal condition, or regular limits in every spacelike direction.

Scope and limitations

Spatial asymptotic flatness does not imply a smooth $\mathscr I^+$, absence of gravitational radiation, or completeness of null generators. It also does not require the physical spacetime to be geodesically complete: Kerr satisfies standard spatial-infinity conditions despite its black-hole region. The claim is about an asymptotic end, not the whole interior.

Parity conditions do not follow from $1/r$ decay. They must be imposed, derived from a chosen gluing class, or replaced by another mechanism that proves finiteness and integrability. Matter introduces another boundary: its energy and momentum densities need decay compatible with the constraints and the intended surface charges. Finally, matching a spatial BMS or Poincaré subgroup to symmetries on $\mathscr I^\pm$ requires additional regularity and corner assumptions; it is not established by an ADM phase space alone.

Common pitfalls

One end is not the entire hypersurface. ADM data are assigned end by end. A compact interior can connect several ends without making their charges identical.

Parity refers to asymptotically Cartesian components. Applying “even” or “odd” independently to spherical-coordinate components introduces basis signs and can reverse the condition. Define the antipodal action before classifying coefficients.

Finiteness is not integrability. A surface variation may be finite yet fail to be the variation of a function on phase space. A valid Hamiltonian generator must satisfy both requirements.

Exercises

On a time-symmetric Schwarzschild slice in isotropic radius, the leading metric is

$$h_{ij}=\left(1+\frac{GM}{2r}\right)^4\delta_{ij} =\left(1+\frac{2GM}{r}+O(r^{-2})\right)\delta_{ij}, \qquad K_{ij}=0.$$

Use the ADM surface integral to recover the mass $M$, and verify the strict leading parity conditions.

Solution

Write $h_{ij}=(1+f)\delta_{ij}$ with $f=2GM/r+O(r^{-2})$. Then

$$\partial_jh_{ij}=\partial_i f=-\frac{2GM}{r^2}n_i+O(r^{-3}),$$

while

$$\partial_i h_{jj}=3\partial_i f=-\frac{6GM}{r^2}n_i+O(r^{-3}).$$

Their difference contracted with $n^i$ is $4GM/r^2+O(r^{-3})$. Since $\int_{S_r}dS=4\pi r^2$, the limit gives

$$E_{\mathrm{ADM}}=\frac{1}{16\pi G}(4GM)(4\pi)=M.$$

The coefficient $a_{ij}=2GM\delta_{ij}$ is independent of $n$ and therefore even. Because $K_{ij}=0$, the leading canonical momentum coefficient vanishes and is simultaneously odd and even; in particular it satisfies the required odd condition. The second asymptotic end of the maximally extended time-symmetric slice has its own identical calculation.

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.
• Henneaux, Marc, and Cédric Troessaert. “The Asymptotic Structure of Gravity at Spatial Infinity in Four Spacetime Dimensions.” arXiv:1904.04495 [hep-th] (2019). arXiv.
• Regge, Tullio, and Claudio Teitelboim. “Role of Surface Integrals in the Hamiltonian Formulation of General Relativity.” Annals of Physics 88, no. 1 (1974): 286–318. DOI.