Asymptotic Flatness

In Newtonian gravity an isolated body is one whose potential dies as $1/r$ , so infinitely far away there is simply no field. General relativity turns that innocent sentence into a research programme. "Far away" must be defined without a preferred coordinate system, and the approach to Minkowski must be threaded through a needle: sharp enough that a total energy and momentum exist and are conserved, yet loose enough that gravitational waves can still stream out to infinity and carry energy with them. Asymptotic flatness is the machinery that walks this line. The decisive move, due to Penrose, is to stop imposing falloff rates on metric components and instead attach a boundary by a conformal rescaling $\tilde g_{ab}=\Omega^2 g_{ab}$ , with $\Omega$ positive inside, vanishing at infinity, and having non-zero gradient there. Because null cones and causal structure are conformally invariant, that boundary is no artifact: outgoing light rays end on future null infinity $\mathscr{I}^+$ , incoming rays begin on past null infinity $\mathscr{I}^-$ , and the "edge of the universe" becomes an ordinary hypersurface one can do local geometry on. Two asymptotic regions then separate cleanly. Spatial infinity $i^0$ , reached along spacelike directions, holds the total conserved mass-energy — the ADM mass — and never feels the radiation. Null infinity $\mathscr{I}^{\pm}$ , reached along null directions, is where radiation registers, and the Bondi mass read on successive cuts of $\mathscr{I}^+$ literally drains as waves depart. Their symmetries are not the expected Poincaré group but the infinite-dimensional Bondi–Metzner–Sachs group, whose supertranslations tie to gravitational memory and the soft-graviton theorems of modern infrared physics.

Key takeaways

Asymptotic flatness makes precise the idea that an isolated system's spacetime approaches Minkowski far from its sources — both along light rays (null infinity) and spacelike slices (spatial infinity).
The definition is conformal, not coordinate-based: rescale by $\Omega^2$ so infinity ( $\Omega=0$ ) is dragged to a finite location and $\tilde g_{ab}=\Omega^2 g_{ab}$ extends smoothly across a genuine boundary.
That boundary is future and past null infinity $\mathscr{I}^{\pm}$ (each null, topology $S^2\times\mathbb{R}$ ), spatial infinity $i^0$ , and the timelike points $i^{\pm}$ .
With $\Lambda=0$ Einstein's equations force $\mathscr{I}$ null; the sign of $\Lambda$ otherwise makes it spacelike (de Sitter) or timelike (anti–de Sitter).
Null infinity carries the Bondi mass and the Bondi–Sachs mass-loss formula, making radiated energy rigorous and manifestly non-increasing.
The asymptotic symmetry group is the infinite-dimensional BMS group, enlarging Poincaré by angle-dependent supertranslations.
Key idea: replace falloff-in-coordinates with conformal compactification — $\tilde g_{ab}=\Omega^2 g_{ab}$ with $\Omega\to0$ smoothly at infinity attaches a finite boundary $\mathscr{I}\cup\{i^0\}$ on which global structure (causal boundary, mass, symmetries, radiation) becomes locally accessible.

The core insight

Replace falloff-in-coordinates with conformal compactification: $\tilde g_{ab}=\Omega^2 g_{ab}$ with $\Omega\to0$ smoothly at infinity attaches a finite boundary $\mathscr{I}\cup\{i^0\}$ on which the global structure — causal boundary, mass, symmetries, and radiation — becomes locally accessible.

Where this lives today

Asymptotic flatness stopped being purely formal the day gravitational waves were detected. Numerical-relativity codes cannot reach $\mathscr{I}^+$ on a finite grid, so they extrapolate waveforms to infinity or evolve on compactified hyperboloidal slices that place $\mathscr{I}^+$ on the computational boundary. The Bondi mass-loss formula then serves as an internal check: the energy the code says left through $\mathscr{I}^+$ must match the drop in Bondi mass, flagging error before the waveform enters a LIGO–Virgo–KAGRA template bank. The GW150914 numbers above — $3\,M_\odot c^2$ radiated, peak $3.6\times10^{49}$ W — are defined at null infinity and nowhere else. The most active frontier is the BMS group itself. Strominger and collaborators showed around 2014 that supertranslation symmetry, Weinberg's soft-graviton theorem, and the gravitational memory effect are three faces of one structure at $\mathscr{I}^+$ — an "infrared triangle" that has reshaped thinking on black-hole information, scattering amplitudes, and proposals to measure permanent detector displacements. Together with the positive energy theorem and celestial holography, it makes asymptotic flatness the place where the cleanest theorems of classical relativity meet quantum gravity.

The physical picture

Imagine squeezing the whole infinite extent of Minkowski space onto a finite sheet of paper, the way a wide-angle lens packs the horizon into one frame. The conformal factor $\Omega$ is that lens: it shrinks infinite proper distances to finite conformal ones, so the "edge at infinity" becomes an ordinary, drawable boundary. Crucially the lens preserves angles — hence light cones — so causal relations survive intact: null rays stay at 45°, and the boundary inherits an unambiguous causal meaning. Different ways of running off to infinity land on different pieces. Sit still forever and you approach future timelike infinity $i^+$ ; send a flash outward and it lands on future null infinity $\mathscr{I}^+$ ; sight along a spatial slice and you approach spatial infinity $i^0$ . Asymptotic flatness demands that all these pieces exist and look exactly like Minkowski's. The first figure is the prototype every definition abstracts: the triangular Penrose diagram of flat spacetime, with its five distinct infinities.

Figure

The conformal (Penrose) diagram of Minkowski space — every point is a 2-sphere, all light rays run at 45°, and the infinite spacetime is compressed to a finite triangle by Ω = cos p cos q with tan p = t+r, tan q = t−r. Outgoing rays end on ℐ⁺ at p = π/2; incoming rays begin on ℐ⁻ at q = −π/2; spatial infinity i⁰ carries the ADM mass

Why null infinity is special is a statement about how gravity radiates. Gravitational waves travel at the speed of light, so radiated energy leaks out along null directions and registers on $\mathscr{I}^+$ — never on $i^0$ , which the waves outrun. The Bondi mass, read on one cut of $\mathscr{I}^+$ after another, watches the energy drain away; the mass-loss formula makes this an exact, sign-definite theorem. The second figure shows the effect at its most dramatic, with numbers from the first detection.

Figure

Bondi mass of the GW150914 spacetime read along ℐ⁺: it holds at the ADM value M ≈ 65 M⊙ during inspiral, then drops through merger to the final Kerr value 62 M⊙. The ΔM = 3.0 M⊙c² ≈ 5.3×10⁴⁷ J deficit is exactly the integrated news flux, and dM_B/du ≤ 0 at every instant

Finally, the symmetry group is larger than intuition suggests. You would guess the ten-parameter Poincaré group. Instead the four translations swell into an infinite tower of angle-dependent "supertranslations," independent time shifts $f(\theta,\phi)$ at each direction on the sphere. This BMS enlargement is no blemish; it stores physical information — the gravitational memory left by a passing wave — and sits at one vertex of the memory / soft-theorem / asymptotic-symmetry triangle of modern infrared gravity.

The core equations

Everything begins with one line. The physical spacetime $(M,g_{ab})$ is untouched; we build a companion $(\tilde M,\tilde g_{ab})$ whose points include a boundary the physical metric could never reach.

\tilde g_{ab}=\Omega^2\,g_{ab} \qquad (1)

where: $g_{ab}$ — physical metric on $M$ ; $\tilde g_{ab}$ — unphysical metric, smooth on $\tilde M=M\cup\mathscr{I}$ ; $\Omega$ — conformal factor, $\Omega>0$ on $M$ . Reading: a rescaling changes lengths but not angles, fixing the light cones in place while freeing us to bring infinity to a finite location. For this to define a boundary, $\Omega$ must vanish at the surface we call infinity while its gradient must not — so the surface is a clean level set.

\Omega\big|_{\mathscr{I}}=0,\qquad n_a\equiv\nabla_a\Omega\big|_{\mathscr{I}}\neq0 \qquad (2)

where: $\mathscr{I}$ — conformal boundary at infinity; $n_a$ — normal covector, the gradient of $\Omega$ . Reading: $\Omega=0$ locates the boundary and $\mathrm{d}\Omega\neq0$ certifies it is a smooth hypersurface, not a pinched level set. The character of that hypersurface is set by the norm of $n_a$ — and this is where the field equations enter. In vacuum with vanishing $\Lambda$ the answer is forced.

\tilde n^a\tilde n_a\big|_{\mathscr{I}}=0\;\Longrightarrow\;\mathscr{I}\ \text{is a null hypersurface}\qquad(\Lambda=0) \qquad (3)

where: $\tilde n^a=\tilde g^{ab}n_b$ — normal raised with the unphysical metric. Reading: a null normal means $\mathscr{I}$ is woven from null rays, which is why light reaches it. This follows from the conformal transformation of the Ricci scalar: at $\mathscr{I}$ one finds $\tilde n^a\tilde n_a=\tfrac{1}{12}\Omega^2 R=$ (with $R=4\Lambda$ ) $\propto-\Lambda/3$ , zero for $\Lambda=0$ and pushing $\mathscr{I}$ spacelike ( $\Lambda>0$ ) or timelike ( $\Lambda<0$ ) otherwise. The concrete model is flat spacetime, which in null coordinates already shows the radial structure compactification will tame.

\mathrm{d}s^2=-\mathrm{d}t^2+\mathrm{d}r^2+r^2\,\mathrm{d}\Omega^2\;\xrightarrow{\,u=t-r,\;v=t+r\,}\;-\mathrm{d}u\,\mathrm{d}v+\tfrac14(v-u)^2\,\mathrm{d}\Omega^2 \qquad (4)

where: $u$ — retarded time $t-r$ ; $v$ — advanced time $t+r$ ; $\mathrm{d}\Omega^2$ — round unit-2-sphere metric. Reading: radial light rays are the lines $u=\text{const}$ or $v=\text{const}$ , adapting coordinates to the directions along which we reach $\mathscr{I}^{\pm}$ . Compactifying is an arctangent away: map the infinite ranges into a finite interval and pick the matching conformal factor.

v=\tan p,\quad u=\tan q,\qquad \Omega=\cos p\,\cos q \qquad (5)

where: $p,q\in(-\tfrac{\pi}{2},\tfrac{\pi}{2})$ with $p\ge q$ — compactified null coordinates. Reading: the rescaled metric $\tilde g=-\mathrm{d}p\,\mathrm{d}q+\tfrac14\sin^2(p-q)\,\mathrm{d}\Omega^2$ is regular at the edges $p=\pi/2$ ( $\mathscr{I}^+$ ) and $q=-\pi/2$ ( $\mathscr{I}^-$ ). (Any constant multiple works; $\cos p\cos q$ clears the $\sec^2$ factors cleanly.) At spatial infinity $i^0$ sits the total energy, a surface integral over a sphere of divergent radius — the ADM mass.

M_{\rm ADM}=\frac{1}{16\pi G}\oint_{S^2_\infty}\left(\partial_j h_{ij}-\partial_i h_{jj}\right)\mathrm{d}S^i \qquad (6)

where: $h_{ij}$ — deviation of the spatial metric from flat, $\gamma_{ij}=\delta_{ij}+h_{ij}$ ; $S^2_\infty$ — sphere at $r\to\infty$ ; $G$ — Newton's constant. Reading: only the falloff matters, so this energy is a boundary quantity — there is no local gravitational energy density. At null infinity the counterpart is the Bondi mass, a function of retarded time whose derivative is a negative-definite flux built from the news.

M_B(u)=\frac{1}{4\pi G}\oint_{S^2}m_B(u,x^A)\,\mathrm{d}^2S,\qquad \frac{\mathrm{d}M_B}{\mathrm{d}u}=-\frac{1}{32\pi G}\oint_{S^2}N_{AB}\,N^{AB}\,\mathrm{d}^2S\;\le\;0 \qquad (7)

where: $m_B$ — Bondi mass aspect (the $1/r$ coefficient of $g_{uu}$ ); $N_{AB}=\partial_u C_{AB}$ — news tensor; $A,B$ — sphere indices. Reading: since $N_{AB}N^{AB}\ge0$ pointwise, the Bondi mass can only fall or hold steady — radiation carries away strictly positive energy. The transformations preserving this structure make up the BMS group, whose most surprising members are supertranslations.

\xi^a=f(\theta,\phi)\,\partial_u+\dots\quad(\text{supertranslations})\;\subset\;\text{BMS}=\text{Lorentz}\ltimes\text{(supertranslations)} \qquad (8)

where: $\xi^a$ — asymptotic symmetry generator preserving $\mathscr{I}$ ; $f(\theta,\phi)$ — arbitrary smooth function on the sphere. Reading: the four lowest harmonics of $f$ recover ordinary translations, but every higher harmonic is an independent, physical symmetry — an angle-dependent time shift Poincaré never knew about.

Derivation AConformal completion of Minkowski space

The goal is to build the conformal boundary of flat spacetime explicitly, so that $\mathscr{I}^{\pm}$ , $i^0$ , and $i^{\pm}$ appear as concrete edges and corners of a finite triangle rather than definitions taken on faith. We pass to null coordinates, compress their infinite ranges with an arctangent, and read off the conformal factor that renders the rescaled metric smooth to the boundary — the prototype every asymptotically flat spacetime must resemble.

Step 1. Take $\mathrm{d}s^2=-\mathrm{d}t^2+\mathrm{d}r^2+r^2\,\mathrm{d}\Omega^2$ and set $u=t-r$ , $v=t+r$ ( $v\ge u$ ). With $t=\tfrac12(u+v)$ , $r=\tfrac12(v-u)$ one gets $-\mathrm{d}t^2+\mathrm{d}r^2=-\mathrm{d}u\,\mathrm{d}v$ , so $\mathrm{d}s^2=-\mathrm{d}u\,\mathrm{d}v+\tfrac14(v-u)^2\,\mathrm{d}\Omega^2$ . Why: radial light rays are now $u=\text{const}$ or $v=\text{const}$ , aligning coordinates with the null directions to infinity.

Step 2. Compress with $v=\tan p$ , $u=\tan q$ , mapping $u,v\in(-\infty,\infty)$ onto $p,q\in(-\tfrac{\pi}{2},\tfrac{\pi}{2})$ with $p\ge q$ ; the images of $u,v=\pm\infty$ are the finite lines $p,q=\pm\tfrac{\pi}{2}$ . Why: the arctangent drags infinity to finite coordinate values while staying smooth inside.

Step 3. Transform. With $\mathrm{d}v=\sec^2 p\,\mathrm{d}p$ , $\mathrm{d}u=\sec^2 q\,\mathrm{d}q$ , the null part is $-\sec^2 p\,\sec^2 q\,\mathrm{d}p\,\mathrm{d}q$ , and $v-u=\sin(p-q)/(\cos p\cos q)$ makes the angular part $\tfrac14\sin^2(p-q)/(\cos^2 p\cos^2 q)\,\mathrm{d}\Omega^2$ . Why: both divergences sit in one factor $1/(\cos^2 p\cos^2 q)$ a rescaling can absorb.

Step 4. Choose $\Omega=\cos p\cos q$ and form $\tilde g=\Omega^2 g$ : the $\sec^2$ factors cancel, leaving $\tilde g=-\mathrm{d}p\,\mathrm{d}q+\tfrac14\sin^2(p-q)\,\mathrm{d}\Omega^2$ , smooth up to $p=\tfrac{\pi}{2}$ , $q=-\tfrac{\pi}{2}$ where $\Omega=0$ . Why: a metric that could never reach infinity is replaced by one that extends across it, light cones untouched.

Step 5. Read off the boundary: $\mathscr{I}^+$ at $p=\tfrac{\pi}{2}$ (outgoing rays end), $\mathscr{I}^-$ at $q=-\tfrac{\pi}{2}$ (incoming begin), $i^0$ at $(\tfrac{\pi}{2},-\tfrac{\pi}{2})$ , $i^{\pm}$ at $(\pm\tfrac{\pi}{2},\pm\tfrac{\pi}{2})$ . Why: this is the standard Penrose diagram, the template for the definition. *Verified independently: $-\mathrm{d}t^2+\mathrm{d}r^2=-\mathrm{d}u\,\mathrm{d}v$ and the cancellation with $\Omega=\cos p\cos q$ recomputed by hand; boundary locations match Wald §11.1. Note the consistent factor is $\cos p\cos q$ , not $2\cos p\cos q$ , for the stated $\tilde g$ .*

Go deeper · Derivation B · 1 minThe Bondi mass-loss formula from the news

Derivation BThe Bondi mass-loss formula from the news

The goal is to show that the Bondi mass on cuts of $\mathscr{I}^+$ decreases monotonically with retarded time, with the flux given by the squared news tensor — the exact, gauge-invariant statement that gravitational radiation carries positive energy away. We install Bondi coordinates, expand in $1/r$ , define the news as the rate of change of the shear, and feed the leading Einstein constraint into a sphere integral where the divergence drops out, leaving a non-positive result.

Step 1. Adopt Bondi gauge near $\mathscr{I}^+$ : $\mathrm{d}s^2=-\left(1-\tfrac{2m_B}{r}\right)\mathrm{d}u^2-2\,\mathrm{d}u\,\mathrm{d}r+r^2\!\left(\gamma_{AB}+\tfrac{1}{r}C_{AB}\right)\mathrm{d}x^A\mathrm{d}x^B+\dots$ , with mass aspect $m_B$ , shear $C_{AB}$ , round-sphere metric $\gamma_{AB}$ . Why: built from outgoing cones and areal radius, so the coefficients are the physical data on $\mathscr{I}^+$ .

Step 2. Define the news $N_{AB}=\partial_u C_{AB}$ . Nonzero news is the invariant marker that radiation crosses $\mathscr{I}^+$ ; a stationary configuration has $N_{AB}\approx0$ . Why: the shear shifts under supertranslations, but its $u$ -derivative is the honest radiative degree of freedom.

Step 3. The Bondi mass is the mass aspect averaged over a cut, $M_B(u)=\tfrac{1}{4\pi G}\oint_{S^2}m_B\,\mathrm{d}^2S$ . Why: this is the far-field energy at retarded time $u$ ; cuts are snapshots as the waves peel off.

Step 4. The leading $uu$ Einstein equation at $\mathscr{I}^+$ reads $\partial_u m_B=-\tfrac18 N_{AB}N^{AB}+\tfrac14 D^A D^B N_{AB}$ , with $D_A$ the sphere derivative. Why: this constraint is the asymptotic remnant of the field equations, tying local mass loss to the radiation plus a total divergence.

Step 5. Integrate over $S^2$ : the divergence vanishes on the closed sphere, leaving $\dfrac{\mathrm{d}M_B}{\mathrm{d}u}=-\dfrac{1}{32\pi G}\oint_{S^2}N_{AB}N^{AB}\,\mathrm{d}^2S\le0$ . The integrand is a square, so the Bondi mass never increases. Why: the monotonic decrease is derived, not assumed — a theorem, not an intuition. *Verified independently: the divergence vanishes on the closed sphere and the surviving square gives $\mathrm{d}M_B/\mathrm{d}u\le0$ ; the coefficient $1/32\pi G$ matches Bondi–van der Burg–Metzner (1962), Sachs (1962), Wald §11.2, MTW §35.*

Limiting cases

Asymptotic flatness sharpens or fails in recognizable regimes, and the way it does so is informative.

Case	Result	Meaning
Vanishing $\Lambda$	$\tilde n^a\tilde n_a\big\|_{\mathscr{I}}=0$	The defining case: $\mathscr{I}$ is null, radiation reaches it, the Bondi–Sachs apparatus applies.
Nonzero $\Lambda$	$\tilde n^a\tilde n_a\propto-\Lambda/3\neq0$	Asymptotic flatness lost — the boundary turns spacelike (de Sitter) or timelike (anti–de Sitter).
Weak-field tail	$g_{00}\simeq-\left(1-\tfrac{2GM}{r}\right),\;r\to\infty$	Minkowski plus a $1/r$ correction whose coefficient is the mass — the heuristic made rigorous.
Stationary at $i^0$	$g_{ab}=\eta_{ab}+O(1/r),\;\partial_c g_{ab}=O(1/r^2)$	No radiation, so news vanishes and $M_B=M_{\rm ADM}$ : one conserved energy.
Radiative cut of $\mathscr{I}^+$	$\dfrac{\mathrm{d}M_B}{\mathrm{d}u}=-\dfrac{1}{32\pi G}\oint_{S^2}N_{AB}N^{AB}\,\mathrm{d}^2S\le0$	Bondi mass strictly decreases at a rate set by the news flux.

Common trapSpatial infinity and null infinity are not the same condition

The phrase "asymptotically flat" hides two logically independent demands, and conflating them is the commonest error here. Flatness at spatial infinity $i^0$ (spacelike directions) defines and conserves the ADM mass and momentum; it is blind to radiation, since outgoing waves race off along null directions and never touch $i^0$ . Flatness at null infinity $\mathscr{I}^{\pm}$ (light rays) gives the Bondi mass and lets it decrease as waves depart. The deep bridge is that the Bondi mass at the far past of $\mathscr{I}^+$ equals the ADM mass — but that is a theorem, not a definition. A related trap: smoothness of $\mathscr{I}$ needs the stress-energy to fall off fast enough, so long-range fields — unscreened Maxwell, or any $\Lambda\neq0$ — generically spoil the standard construction.

Worked examples

1 · The ADM mass of a Schwarzschild black hole (fully worked)

In isotropic coordinates the Schwarzschild spatial metric is $\gamma_{ij}=\left(1+\tfrac{M}{2\rho}\right)^4\delta_{ij}$ ( $G=c=1$ ). To leading order in $M/\rho$ , evaluate the ADM integral (6) and confirm it returns $M$ .

Step 1. Expand: $\left(1+\tfrac{M}{2\rho}\right)^4=1+\tfrac{2M}{\rho}+O(M^2/\rho^2)$ , so to leading order $h_{ij}=\dfrac{2M}{\rho}\,\delta_{ij}$ — the perturbation whose falloff the ADM integral reads. Data: $h_{ij}=(2M/\rho)\delta_{ij}$ .

Step 2. Differentiate. With $h_{jj}=\tfrac{6M}{\rho}$ , $\partial_i h_{jj}=-\dfrac{6M x_i}{\rho^3}$ and $\partial_j h_{ij}=-\dfrac{2M x_i}{\rho^3}$ , so $\partial_j h_{ij}-\partial_i h_{jj}=\dfrac{4M x_i}{\rho^3}$ .

Step 3. Contract with $\mathrm{d}S^i=x_i\,\rho\,\mathrm{d}\Omega$ : the integrand is $\dfrac{4M x_i}{\rho^3}\cdot x_i\,\rho\,\mathrm{d}\Omega=4M\,\mathrm{d}\Omega$ , since $x_i x_i=\rho^2$ . A constant $4M\,\mathrm{d}\Omega$ .

Step 4. Integrate and divide: $\oint 4M\,\mathrm{d}\Omega=16\pi M$ , so $M_{\rm ADM}=\dfrac{1}{16\pi}\cdot16\pi M=M$ . $M_{\rm ADM}=M$ — the surface integral at infinity recovers the mass with no reference to the interior.

2 · Energy radiated to $\mathscr{I}^+$ by GW150914 (fully worked)

The first detected merger began as holes of $36\,M_\odot$ and $29\,M_\odot$ and left a Kerr hole of $62\,M_\odot$ . Using the Bondi mass-loss formula as bookkeeping, find the energy that crossed $\mathscr{I}^+$ and compare its peak rate to the starlight of the observable universe (about $10^{49}$ W).

Step 1. The early-time Bondi mass equals the binary's ADM mass, $M_B(u\to-\infty)\approx36+29=65\,M_\odot$ ; the late-time value is the final hole, $62\,M_\odot$ . The mass-loss formula guarantees the difference was radiated: $\Delta M=3.0\,M_\odot$ . $\Delta M=3.0\,M_\odot c^2$ went out through $\mathscr{I}^+$ .

Step 2. In joules, with $M_\odot=1.989\times10^{30}\ \mathrm{kg}$ and $c^2=8.988\times10^{16}\ \mathrm{m^2/s^2}$ : $E=3.0\times1.989\times10^{30}\times8.988\times10^{16}=5.4\times10^{47}\ \mathrm{J}$ . $E\approx5.3\times10^{47}$ J radiated.

Step 3. The flux $\mathrm{d}M_B/\mathrm{d}u=-\tfrac{1}{32\pi G}\oint N_{AB}N^{AB}\,\mathrm{d}^2S$ peaks near merger; the observed peak luminosity was $L_{\rm peak}\approx3.6\times10^{49}\ \mathrm{W}$ , roughly $200\,M_\odot$ per second. $L_{\rm peak}\approx3.6\times10^{49}$ W.

Step 4. That exceeds the combined starlight of the observable universe (about $10^{49}$ W): for a few milliseconds this binary outshone the visible cosmos in gravitational radiation, all of it a drop in the Bondi mass on $\mathscr{I}^+$ . The formula turns "waves carry energy" into $5\times10^{47}$ J of sign-definite bookkeeping.

Physics of the portrait

Asymptotic flatness had to be invented twice: as a physical intuition (a body far from everything feels no gravity) and as a rigorous structure surviving the coordinate freedom of general relativity. The second is largely the work of Roger Penrose, who saw that the way to control infinity is to bring it to a finite place with a conformal rescaling.

Origins · 1 minRoger Penrose and how this was discovered

Roger Penrose

1931– · conformal treatment of infinity in general relativity

Roger Penrose turned the study of infinity from a problem in coordinate falloff into one in local differential geometry. In papers from 1963 on, he introduced the conformal rescaling $\tilde g_{ab}=\Omega^2 g_{ab}$ that drags "infinitely far away" to a finite boundary — future and past null infinity $\mathscr{I}^{\pm}$ — where smoothness of the rescaled metric replaces vague $1/r$ falloff with sharp, checkable conditions. Because the factor preserves angles and hence null cones, causal structure survives compression, which is why light rays end on $\mathscr{I}$ . The construction gave a rigorous home to results growing up alongside it: the ADM mass of Arnowitt, Deser and Misner at spatial infinity, and the mass-loss analysis of Bondi, van der Burg, Metzner and Sachs at null infinity, whose symmetry group proved to be not Poincaré but the infinite-dimensional BMS group. Geroch, Ashtekar and Hansen then made the picture invariant. Penrose's diagrams — the finite triangles now drawn in every relativity course — are the visible face of this machinery, which also underlies his singularity theorems and the modern soft-theorem revival that reads gravitational memory off $\mathscr{I}^+$ . He shared the 2020 Nobel Prize for showing that black-hole formation is a robust prediction of the theory, built on exactly this control of global causal structure.

History

1959

— Arnowitt, Deser and Misner give the Hamiltonian formulation and define the ADM mass at spatial infinity

1962

— Bondi, van der Burg, Metzner and Sachs derive the mass-loss formula and discover the BMS group at null infinity

1963

— Penrose introduces the conformal treatment of infinity and the compactified spacetime diagram

1965

— Penrose formalizes asymptotic simplicity and the null boundary 𝓘

1981

— Schoen and Yau, and independently Witten, prove the positive energy theorem for the ADM mass

Going further

Texts. Wald, General Relativity (1984), Ch. 11 · Misner, Thorne & Wheeler, Gravitation (1973), Ch. 34–35 · Hawking & Ellis, The Large Scale Structure of Space-Time (1973), §5.1 · Penrose & Rindler, Spinors and Space-Time, Vol. 2 (1986) · Carroll, Spacetime and Geometry (2004), Ch. 5 and App. H · Schutz, A First Course in General Relativity, 2nd ed. (2009).

Provenance. The conformal completion follows Penrose (1963, 1965), Wald Ch. 11, and Geroch (1977); ADM mass from Arnowitt–Deser–Misner (1959) and MTW §20; mass loss from Bondi–van der Burg–Metzner (1962) and Sachs (1962); intrinsic spatial infinity from Ashtekar–Hansen (1978). A verifier should recompute: (1) the conformal Minkowski metric from $v=\tan p,\,u=\tan q$ with $\Omega=\cos p\cos q$ — this factor, not $2\cos p\cos q$ , is consistent with the stated $\tilde g$ (Wald §11.1); (2) the null character of $\mathscr{I}$ from the conformal Ricci-scalar law at $\Lambda=0$ , sign $\tilde n^a\tilde n_a\propto-\Lambda/3$ ; (3) the coefficient $1/32\pi G$ ; (4) the Schwarzschild ADM integral $=M$ from $h_{ij}=(2M/\rho)\delta_{ij}$ ; (5) the GW150914 figures $\Delta M=3.0\,M_\odot$ , $E\approx5.3\times10^{47}$ J, $L_{\rm peak}\approx3.6\times10^{49}$ W. Conventions: signature $(-,+,+,+)$ , $G$ Newton's constant.

Prerequisites. Conformal Transformations and Causal Invariance · Penrose–Carter Conformal Diagrams · Conformal Infinity and Asymptotic Structure · Null Geodesics and Light Cones · Causal Structure of Spacetime. Leads to. Null Infinity (Scri) · Global Hyperbolicity · Conformal Infinity and Asymptotic Structure.

Ask the tutor: Ask me to walk through why the Bondi mass at the far past of $\mathscr{I}^+$ equals the ADM mass at $i^0$ — the theorem tying null and spatial infinity together.

Is asymptotic flatness just the statement that the metric goes to Minkowski at large r?

That is the intuition, but the naive version is gauge-dependent and too weak to work with. "Goes to Minkowski at large $r$ " presupposes a coordinate $r$ and a rate of approach, and different coordinates disagree about whether a spacetime qualifies. Asymptotic flatness replaces this with a coordinate-free condition: there exists a factor $\Omega$ bringing infinity to a smooth boundary $\mathscr{I}$ where the rescaled metric is regular and $\mathrm{d}\Omega\neq0$ . That automatically encodes the right falloff, separates null from spatial infinity, and — being a statement about a genuine boundary — lets you prove theorems (mass positivity, mass loss) the coordinate version cannot even phrase.

What is the difference between the ADM mass and the Bondi mass?

Both are the total energy of an isolated system, but measured at different infinities. The ADM mass lives at spatial infinity $i^0$ and is a constant — it counts everything, including energy that will later be radiated, and never changes. The Bondi mass lives on cuts of null infinity $\mathscr{I}^+$ labeled by retarded time $u$ and decreases as waves cross $\mathscr{I}^+$ : it is the energy still in the system at that moment. A theorem ties them together — the Bondi mass in the far past of $\mathscr{I}^+$ equals the ADM mass — so their total difference over $\mathscr{I}^+$ is exactly the energy radiated.

Why must null infinity be a null surface — could a definition just declare it so?

It is a consequence of Einstein's equations, not a definition. Demand a smooth conformal completion of a vacuum spacetime with $\Lambda=0$ , and the conformal transformation of the curvature forces the norm of the boundary normal $n_a=\nabla_a\Omega$ to vanish at $\mathscr{I}$ — a vanishing norm means the normal, and the surface, is null. The cosmological constant would change this: $\tilde n^a\tilde n_a\propto-\Lambda/3$ at the boundary, so $\Lambda>0$ makes $\mathscr{I}$ spacelike (de Sitter) and $\Lambda<0$ timelike (anti–de Sitter). "Null infinity" is null precisely because $\Lambda$ is taken to be zero in this idealization.

How can gravitational waves carry energy if there is no local energy density in general relativity?

This is the puzzle asymptotic flatness resolves. There genuinely is no gauge-invariant local energy density for the gravitational field — the equivalence principle lets you transform it away at any point. But energy at infinity is well defined, because there the geometry settles down enough to compare against a fixed Minkowski background. The Bondi mass on a cut of $\mathscr{I}^+$ is such a quantity, and the mass-loss formula shows it dropping by the integrated news flux. So "waves carry energy" is a statement about a boundary quantity: the energy is unambiguous only once the waves reach $\mathscr{I}^+$ and register as a fall in the Bondi mass.

What are supertranslations, and why should I care about an infinite-dimensional symmetry group?

They are the surprise in the symmetry group of null infinity. You expect the transformations preserving the asymptotic structure to be the ten-parameter Poincaré group, but the four rigid translations turn out to be the lowest members of an infinite family: independent time shifts $f(\theta,\phi)$ at each point of the sphere at infinity, enlarging Poincaré to BMS. They matter because they are not empty gauge — they relate configurations differing by gravitational memory (a permanent displacement of test masses after a wave passes), and through Ward identities they are equivalent to Weinberg's soft-graviton theorem. That is why an old result in asymptotic geometry now sits at the center of work on scattering, black-hole information, and celestial holography.

Does a nonzero cosmological constant ruin all of this — and isn't our universe's Λ nonzero?

A nonzero $\Lambda$ does break strict asymptotic flatness: $\mathscr{I}$ is no longer null (it is spacelike for the observed $\Lambda>0$ ), the Bondi construction's light rays no longer end on it the same way, and even defining the total energy becomes subtle and partly open. Our $\Lambda$ is small but positive. The resolution is scale separation: for an isolated system like a binary black hole, the relevant "infinity" is far from the source yet deep inside the cosmological horizon, where the geometry is superbly approximated as asymptotically flat. So it is an idealization — but a controlled one, and the right tool for gravitational-wave astronomy even in a $\Lambda$ -dominated cosmos.

Asymptotic Flatness

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · The ADM mass of a Schwarzschild black hole (fully worked)

2 · Energy radiated to $\mathscr{I}^+$ by GW150914 (fully worked)

Physics of the portrait

Going further

Before this

Where this leads

Asymptotic Flatness

Where this lives today

The physical picture

The core equations

Limiting cases

Worked examples

1 · The ADM mass of a Schwarzschild black hole (fully worked)

2 · Energy radiated to I+\mathscr{I}^+I+ by GW150914 (fully worked)

Physics of the portrait

Going further

Before this

Where this leads

2 · Energy radiated to $\mathscr{I}^+$ by GW150914 (fully worked)