Conformal Infinity and Asymptotic Structure

"Infinity" is where the hard questions of gravitation get answered — how much energy a radiating system has left, whether a signal escapes a black hole, the total mass of an isolated body — and yet in the physical metric infinity is exactly where distances and times blow up and geometry fails. Roger Penrose's remedy, from the early 1960s, was audacious in its simplicity: multiply the physical metric $\tilde g_{ab}$ by the square of a smooth factor $\Omega$ that shrinks to zero where physical distances diverge, $g_{ab}=\Omega^2\tilde g_{ab}$ . The rescaled ("unphysical") metric stays perfectly regular up to the surface $\Omega=0$ , and that surface — conformal infinity, written $\mathscr{I}$ and pronounced "scri" — becomes an honest boundary you can attach to the manifold and probe with ordinary differential geometry. The construction works because a conformal rescaling leaves null cones untouched: the causal skeleton that carries radiation survives intact even as the divergent distance is squeezed into a finite range. The single number classifying the boundary is the norm of its conformal normal $n_a=\nabla_a\Omega$ , fixed by the cosmological constant alone — null for asymptotically flat spacetimes, spacelike for de Sitter, timelike for anti-de Sitter (the stage of AdS/CFT holography). From the demand that the rescaled geometry be smooth at $\Omega=0$ flow the peeling of the Weyl tensor, the Bondi mass-loss formula, and the BMS symmetry group — invariant statements about how fields behave far from an isolated source. This article is the machine room beneath every Penrose diagram.

Key takeaways

Conformal infinity is built by rescaling the physical metric, $g_{ab}=\Omega^2\tilde g_{ab}$ , with $\Omega\to 0$ (but $\mathrm{d}\Omega\neq0$ ) far out; the locus $\Omega=0$ is a finite, regular boundary $\mathscr{I}$ .
Rescaling preserves null cones, so light rays still end at definite points of $\mathscr{I}$ even as their physical affine parameter runs to infinity.
The causal type of $\mathscr{I}$ is fixed by $\Lambda$ through $g^{ab}n_an_b\big|_{\mathscr{I}}=-\Lambda/3$ : null for $\Lambda=0$ , spacelike for $\Lambda>0$ (de Sitter), timelike for $\Lambda<0$ (anti-de Sitter).
For asymptotically flat spacetimes the boundary has five pieces — null infinity $\mathscr{I}^{\pm}$ (topology $\mathbb{R}\times S^2$ ), timelike infinity $i^{\pm}$ , spatial infinity $i^0$ .
Smoothness of the rescaled fields forces the peeling rates $\tilde\Psi_n=O(r^{-(5-n)})$ , the Bondi mass-loss law $\dot M_B\le 0$ , and the BMS symmetry group.
Key idea: multiply the physical metric by the square of a factor $\Omega$ that vanishes with non-vanishing gradient where distances diverge; its zero set is a finite regular boundary whose causal type — fixed by $-\Lambda/3$ — encodes the asymptotic structure and carries clean radiation data.

The core insight

Multiply the physical metric by the square of a conformal factor $\Omega$ that vanishes (with non-vanishing gradient) where physical distances diverge; the zero set of $\Omega$ becomes a finite, regular boundary $\mathscr{I}$ whose causal type — null, spacelike or timelike according to the sign of $-\Lambda/3$ — encodes the asymptotic structure, and on which fields acquire well-defined boundary data.

Where this lives today

Conformal infinity has moved from foundational geometry to working tool. In gravitational-wave astronomy the waveforms LIGO, Virgo and KAGRA compare against data are increasingly extracted genuinely at $\mathscr{I}^+$ , through Cauchy-characteristic extraction that carries numerical-relativity data to null infinity where the Bondi news is unambiguous — clean $1/r$ radiation, not gauge contamination. The memory effect — a permanent displacement of test masses after a burst — is now understood as a BMS supertranslation between two vacua on $\mathscr{I}^+$ , tied by Ward identities to the soft-graviton theorem in Strominger's "infrared triangle." At the opposite sign of $\Lambda$ , the timelike $\mathscr{I}$ of anti-de Sitter is the home of holography: AdS/CFT puts a conformal field theory on exactly this boundary. Even our de Sitter-like universe is a frontier — defining energy and observables on a spacelike $\mathscr{I}^+$ with horizons drives celestial and de Sitter holography.

The physical picture

Imagine drawing the whole of infinite Minkowski spacetime on a finite sheet of paper. You can only do it by shrinking distances harder and harder the further out you go — that ever-more-aggressive shrinking is the conformal factor $\Omega$ falling to zero. The reward is enormous: the edge of the paper is now a real place you can analyse with calculus. Light rays make the edge meaningful — because null cones are conformally invariant, a ray that escapes to infinity physically arrives at a definite point of the boundary. Outgoing rays collect on one null surface, $\mathscr{I}^+$ ; incoming rays on another, $\mathscr{I}^-$ . The first figure is that compression made literal — flat spacetime as a finite triangle whose edges are infinity.

Figure

Penrose diagram of Minkowski spacetime: the infinite (T,R) plane squeezed into a finite triangle. Null infinity ℐ⁺ and ℐ⁻ are the 45° edges where outgoing and incoming light rays terminate; the corners i⁺, i⁻ and i⁰ are future, past and spatial infinity. The boundary is Ω = 2 cos p cos q = 0, reached at p = π/2 or q = −π/2 while the edge r = 0 stays finite.

Think of $\Omega$ as an altimeter: its gradient $n_a=\nabla_a\Omega$ points "toward the boundary," and whether that pointer is null, timelike or spacelike is decided by the cosmological constant alone. The boundary is not a wall fields bounce off but a screen onto which the interior's asymptotic news — gravitational waves, fluxes, the slow decay of mass — is projected as clean data. The second figure carries the relation that classifies everything: the norm of the conformal normal is $-\Lambda/3$ , and its sign flips the boundary between the three regimes.

Figure

The sign of Λ fixes the causal character of ℐ through gᵃᵇnₐnᵦ = −Λ/3. Λ = 0 gives a null boundary at 45° (asymptotically flat); Λ > 0 gives −Λ/3 < 0, a timelike normal and spacelike boundary (de Sitter); Λ < 0 gives −Λ/3 > 0, a spacelike normal and timelike boundary (anti-de Sitter, ℐ a vertical cylinder).

The core equations

The construction begins with one algebraic move: a positive, smooth rescaling of the physical metric. The physical $\tilde g_{ab}$ is what geodesics and clocks respond to; the unphysical $g_{ab}$ is chosen to stay regular where $\tilde g_{ab}$ cannot.

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

where: $\tilde g_{ab}$ — physical metric on the physical manifold $\tilde M$ ; $g_{ab}$ — unphysical metric, smooth up to and including the boundary; $\Omega$ — conformal factor, smooth, with $\Omega>0$ on $\tilde M$ and $\Omega=0$ , $\mathrm{d}\Omega\neq0$ on the boundary. Reading: the same spacetime remeasured so infinite distance becomes finite. The boundary is defined as the zero set of $\Omega$ , with the extra demand that the gradient not vanish there, so it is a smooth hypersurface with a well-defined normal.

\mathscr{I}\equiv\partial M=\{\,p:\ \Omega(p)=0,\ \nabla_a\Omega(p)\neq0\,\} \qquad (2)

where: $\mathscr{I}$ — conformal infinity ("scri"); $\nabla_a$ — covariant derivative of the unphysical metric $g_{ab}$ . Reading: $\mathscr{I}$ is where $\Omega$ reads zero but is still changing, so it attaches smoothly to $\tilde M$ . The gradient of $\Omega$ is the normal covector to $\mathscr{I}$ , and its norm classifies the boundary. Evaluating the field equations on $\mathscr{I}$ (Derivation 2) collapses the story to one number, set by $\Lambda$ .

n_a=\nabla_a\Omega,\qquad g^{ab}n_an_b\big|_{\mathscr{I}}=-\frac{\Lambda}{3} \qquad (3)

where: $n_a$ — conformal normal covector to $\mathscr{I}$ ; $\Lambda$ — cosmological constant. Reading (mostly-plus): $\Lambda=0$ makes $n_a$ null (null $\mathscr{I}$ ); $\Lambda>0$ timelike (spacelike $\mathscr{I}$ ); $\Lambda<0$ spacelike (timelike $\mathscr{I}$ ). The causal skeleton survives because the Weyl tensor — the trace-free, tidal curvature — is conformally invariant with this index placement.

\tilde C^a{}_{bcd}=C^a{}_{bcd} \qquad (4)

where: $C^a{}_{bcd}$ — Weyl tensor with the first index raised; the tilde marks the physical version. Reading: the tensor carrying gravitational radiation is unchanged by $g_{ab}\to\Omega^2 g_{ab}$ , so radiative information stays finite at $\mathscr{I}$ — the seed of peeling. Smoothness of the rescaled Weyl field at $\mathscr{I}$ , expanded in $\Omega\sim 1/r$ , forces a staggered "peeling" falloff of the outgoing curvature.

\tilde\Psi_n=O\!\left(r^{-(5-n)}\right),\quad n=0,\dots,4 \qquad (5)

where: $\tilde\Psi_n$ — the five complex Weyl scalars in a Newman–Penrose null tetrad, $\tilde\Psi_4$ the transverse outgoing radiation, $\tilde\Psi_0$ the incoming part; $r$ — affine (luminosity) distance. Reading: $\tilde\Psi_4\sim r^{-1}$ dominates far out, $\tilde\Psi_0\sim r^{-5}$ (near-zone tidal field) dies fastest. On $\mathscr{I}^+$ that radiation drains energy at a rate fixed by the square of the asymptotic shear (the "news function") — the Bondi mass-loss formula.

\frac{\mathrm{d}M_B}{\mathrm{d}u}=-\frac{1}{4\pi}\oint_{S^2}|\sigma^0|^2\,\mathrm{d}^2\!S\ \le 0 \qquad (6)

where: $M_B(u)$ — Bondi mass at retarded time $u$ on $\mathscr{I}^+$ ; $u$ — retarded (Bondi) time along the generators of $\mathscr{I}^+$ ; $\sigma^0$ — asymptotic shear / news function encoding the flux. Reading: whenever news flows the Bondi mass strictly decreases — a system only loses energy to radiation. Integrating from the infinite past ties the Bondi mass on $\mathscr{I}^+$ to the ADM mass at spatial infinity $i^0$ — the total energy including all radiation the system will ever emit.

M_{\mathrm{ADM}}=\lim_{u\to-\infty}M_B(u),\qquad M_{\mathrm{ADM}}\ge M_B(u) \qquad (7)

where: $M_{\mathrm{ADM}}$ — total mass at $i^0$ ; $M_B(u)$ — Bondi mass after the radiation before $u$ has escaped. Reading: the two standard masses are one quantity read at two pieces of infinity; their difference is radiated energy. Finally, the concrete model that makes all this visible: flat spacetime itself, mapped into a finite patch of the Einstein static cylinder by compactified null coordinates (Derivation 1).

\mathrm{d}\tilde s^2=\Omega^{-2}\,\mathrm{d}s^2,\qquad \Omega=2\cos p\,\cos q,\qquad V=T+R=\tan p,\ \ U=T-R=\tan q \qquad (8)

where: $(T,R)$ — Minkowski time and radius; $(p,q)$ — compactified null coordinates of finite range, $-\tfrac{\pi}{2}<q\le p<\tfrac{\pi}{2}$ . Reading: the tangent map drags $T,R\to\pm\infty$ into finite corners, and $\Omega$ vanishes on the boundary triangle.

Derivation AConformal compactification of Minkowski spacetime

The goal is to show explicitly how the entire infinite flat spacetime maps into a finite triangle, so that the five boundary pieces $\mathscr{I}^{\pm}$ , $i^{\pm}$ and $i^0$ appear not as definitions imposed from outside but as the places where an honestly-built conformal factor $\Omega$ hits zero. The route: null coordinates, tangent-compression to finite range, strip the divergent prefactor, read off the zero set.

Step 1. Write the flat metric $\mathrm{d}\tilde s^2=-\mathrm{d}T^2+\mathrm{d}R^2+R^2\,\mathrm{d}\Omega_2^2$ and pass to null coordinates $V=T+R$ , $U=T-R$ ( $V\ge U$ ), giving $\mathrm{d}\tilde s^2=-\mathrm{d}U\,\mathrm{d}V+\tfrac14(V-U)^2\,\mathrm{d}\Omega_2^2$ . Why: null coordinates are the natural language of a boundary made of light rays — constant $U$ outgoing cones, constant $V$ incoming, and $\mathscr{I}$ is where they run to infinity.

Step 2. Compress the infinite ranges with $V=\tan p$ , $U=\tan q$ , $-\tfrac{\pi}{2}<q\le p<\tfrac{\pi}{2}$ , so $\mathrm{d}V=\sec^2p\,\mathrm{d}p$ , $\mathrm{d}U=\sec^2q\,\mathrm{d}q$ and $V-U=\dfrac{\sin(p-q)}{\cos p\,\cos q}$ . Why: $\tan$ maps the finite interval onto the whole real line, so its inverse pulls infinity back to a finite coordinate edge without touching any interior point.

Step 3. Substituting gives $\mathrm{d}\tilde s^2=\dfrac{1}{4\cos^2p\,\cos^2q}\big(-4\,\mathrm{d}p\,\mathrm{d}q+\sin^2(p-q)\,\mathrm{d}\Omega_2^2\big)$ . Why: every divergence is now packaged into the single scalar prefactor $\big(4\cos^2p\cos^2q\big)^{-1}$ , which blows up exactly where $\cos p$ or $\cos q$ vanishes — the candidate boundary.

Step 4. Cancel the prefactor with $\Omega=2\cos p\,\cos q$ : the unphysical metric $\mathrm{d}s^2=\Omega^2\,\mathrm{d}\tilde s^2=-4\,\mathrm{d}p\,\mathrm{d}q+\sin^2(p-q)\,\mathrm{d}\Omega_2^2$ is smooth on a finite patch of the Einstein static cylinder $\mathbb{R}\times S^3$ . Why: the whole trick — divergent physical and regular unphysical geometries differ only by $\Omega^2$ , and the boundary lives where $\Omega=0$ .

Step 5. The boundary is $\Omega=0$ : $p=\tfrac{\pi}{2}$ ( $V\to+\infty$ ) is $\mathscr{I}^+$ , $q=-\tfrac{\pi}{2}$ ( $U\to-\infty$ ) is $\mathscr{I}^-$ , and their meeting points are $i^+$ ( $p=q=\tfrac{\pi}{2}$ ), $i^-$ ( $p=q=-\tfrac{\pi}{2}$ ), $i^0$ ( $p=\tfrac{\pi}{2},q=-\tfrac{\pi}{2}$ ). Why: the "five pieces of infinity" are now quantitative loci — the exact edges of the triangle in the first figure. *The null edges appear at 45° because the $(p,q)$ metric is off-diagonal; the intermediate result $V-U=\sin(p-q)/(\cos p\cos q)$ and the $\Omega^2$ cancellation are exact.*

Go deeper · Derivation B · 1 minThe causal character of infinity from the field equations

Derivation BThe causal character of infinity from the field equations

The goal is to prove that the one number classifying $\mathscr{I}$ — the norm of its normal — is fixed by the cosmological constant alone, $g^{ab}n_an_b\big|_{\mathscr{I}}=-\Lambda/3$ , covering the flat, de Sitter and anti-de Sitter cases at once. The strategy: feed the vacuum-with- $\Lambda$ Einstein equation through the conformal law of the Ricci tensor, clear the negative powers of $\Omega$ , and evaluate on $\mathscr{I}$ where $\Omega=0$ kills every term but one.

Step 1. Vacuum with cosmological constant gives $\tilde R_{ab}=\Lambda\,\tilde g_{ab}=\Lambda\,\Omega^{-2}g_{ab}$ . Why: this is the field content defining "asymptotically flat / de Sitter / anti-de Sitter," and the classification hinges entirely on $\Lambda$ .

Step 2. Insert the 4D conformal Ricci law under $g_{ab}=\Omega^2\tilde g_{ab}$ , $\tilde R_{ab}=R_{ab}+2\Omega^{-1}\nabla_a\nabla_b\Omega+g_{ab}\big(\Omega^{-1}\Box\Omega-3\Omega^{-2}\nabla_c\Omega\,\nabla^c\Omega\big)$ . Why: this bridges the geometry the equations control ( $\tilde R_{ab}$ ) and the geometry regular at the boundary; its most singular term is the one that survives.

Step 3. Multiply by $\Omega^2$ to clear negative powers: $\Lambda\,g_{ab}=\Omega^2 R_{ab}+2\Omega\,\nabla_a\nabla_b\Omega+g_{ab}\big(\Omega\,\Box\Omega-3\,\nabla_c\Omega\,\nabla^c\Omega\big)$ . Why: every term now carries an explicit power of $\Omega$ except the last, so the boundary limit is unambiguous — provided $R_{ab}$ and $\nabla_a\nabla_b\Omega$ are finite there, exactly the smoothness defining a conformal completion.

Step 4. On $\mathscr{I}$ ( $\Omega=0$ ) every explicit- $\Omega$ term drops, leaving $\Lambda\,g_{ab}=-3\,g_{ab}\,\nabla_c\Omega\,\nabla^c\Omega$ ; contracting gives $g^{ab}n_an_b\big|_{\mathscr{I}}=-\Lambda/3$ . Why: this line classifies everything — $\Lambda=0$ null $\mathscr{I}$ ; $\Lambda>0$ timelike normal, spacelike $\mathscr{I}$ ; $\Lambda<0$ spacelike normal, timelike $\mathscr{I}$ . *Note on signs: the expression " $-\tfrac13\Lambda\Omega^2$ " cannot be the boundary value — it vanishes on $\mathscr{I}$ ; the correct finite result is $-\Lambda/3$ , which fixes the table sign entries (de Sitter $<0$ , AdS $>0$ ).*

Go deeper · Derivation C · 1 minWeyl invariance and the peeling of gravitational radiation

Derivation CWeyl invariance and the peeling of gravitational radiation

The goal is to show why demanding that curvature be regular at $\mathscr{I}$ forces the outgoing gravitational field into the staggered falloff $\tilde\Psi_n\sim r^{-(5-n)}$ , so that far out only the transverse radiative part $\tilde\Psi_4\sim 1/r$ survives — the peeling theorem. The route: use the conformal invariance of the mixed-index Weyl tensor to make the radiative field boundary-smooth, then Taylor-expand it in $\Omega\sim1/r$ .

Step 1. The Riemann tensor splits into a Ricci part and the trace-free Weyl part, and the Weyl tensor with one index raised is conformally invariant, $\tilde C^a{}_{bcd}=C^a{}_{bcd}$ . Why: the Ricci part vanishes in vacuum, so all free radiative data live in Weyl — and its invariance carries that data intact to the boundary.

Step 2. Since $C^a{}_{bcd}$ (regular at $\mathscr{I}$ ) equals $\tilde C^a{}_{bcd}$ , and lowering an index costs a factor $\Omega^2$ , the combination $\Omega^{-1}\tilde C_{abcd}$ extends smoothly to $\mathscr{I}$ ; the physical Weyl tensor vanishes there like $\Omega\sim1/r$ . Why: the analytic core — a wildly-varying physical object becomes a finite boundary field once the right power of $\Omega$ is stripped, legalising Taylor expansion.

Step 3. Project onto a Newman–Penrose null tetrad aligned with the outgoing geodesics and expand in $\Omega\sim1/r$ ; the components $\tilde\Psi_0,\dots,\tilde\Psi_4$ appear at successive orders, giving $\tilde\Psi_n=O(r^{-(5-n)})$ . Why: the tetrad sorts the field by orientation to the outgoing ray, and smoothness forces these projections on at staggered rates — $\tilde\Psi_4\sim r^{-1}$ dies last, $\tilde\Psi_0\sim r^{-5}$ first.

Step 4. So at large $r$ the field "peels," dominated by the $r^{-1}$ radiative term $\tilde\Psi_4$ that defines outgoing gravitational waves. Why: this makes $\mathscr{I}^+$ the unambiguous home of radiation — only the slowest-falling part reaches the boundary, exactly the news driving the Bondi mass loss (6).

Limiting cases

The causal character of conformal infinity is fixed by the cosmological constant through the boundary norm of $n_a=\nabla_a\Omega$ (mostly-plus signature, corrected sign convention as derived above):

Case	Result	Meaning
Asymptotically flat ( $\Lambda=0$ )	$g^{ab}n_an_b\big\vert_{\mathscr{I}}=0$	$\mathscr{I}$ is null, topology $\mathbb{R}\times S^2$ , split into $\mathscr{I}^{\pm}$ — the home of gravitational radiation, the Bondi mass and the BMS group.
Asymptotically de Sitter ( $\Lambda>0$ )	$g^{ab}n_an_b\big\vert_{\mathscr{I}}=-\Lambda/3<0$	$n_a$ timelike, so $\mathscr{I}$ is spacelike ( $\mathscr{I}^+$ future, $\mathscr{I}^-$ past); cosmological horizons appear and there is no unambiguous total radiated energy.
Asymptotically anti-de Sitter ( $\Lambda<0$ )	$g^{ab}n_an_b\big\vert_{\mathscr{I}}=-\Lambda/3>0$	$n_a$ spacelike, so $\mathscr{I}$ is timelike — a genuine conformal boundary $\mathbb{R}\times S^{d-1}$ ; not globally hyperbolic, boundary conditions required, the setting of AdS/CFT.

Common trapReaching scri at finite conformal parameter does not mean reaching it in finite physical time

A radial null ray hits $\mathscr{I}^+$ at a finite value of the conformal coordinate ( $p\to\pi/2$ ), and it is tempting to conclude infinity has become "close." It has not. Conformal rescaling preserves null geodesics as point sets but rescales their affine parameter: the same curve reaching the boundary at finite conformal parameter runs to infinite physical affine parameter and proper distance. That is the whole design — the divergent "infinity" is compressed into a finite locus you can analyse, without any physical distance shrinking. A related trap: $\Omega$ is not unique ( $\Omega\to\omega\Omega$ for positive smooth $\omega$ is equally valid), so any statement depending on the particular $\Omega$ — rather than the conformal class and normalised quantities on $\mathscr{I}$ — is gauge, not physics.

Worked examples

1 · The Bondi mass budget of GW150914 (fully worked)

The first binary-black-hole merger detected by LIGO, GW150914, had initial component masses of about $36\,M_\odot$ and $29\,M_\odot$ and a final black hole of about $62\,M_\odot$ . Treating the system as isolated and asymptotically flat, find the energy radiated to $\mathscr{I}^+$ and check it against the mass-loss inequality (7).

Step 1. Match the masses to the pieces of infinity. The ADM mass at $i^0$ counts everything — the inspiralling pair before appreciable radiation, $M_{\mathrm{ADM}}\approx 36+29=65\,M_\odot$ . The final Bondi mass on $\mathscr{I}^+$ , after the waves escape, is the remnant $M_B(+\infty)\approx 62\,M_\odot$ .

Step 2. By the integrated form of (6)–(7) the radiated energy is the drop in Bondi mass, $\Delta M = M_{\mathrm{ADM}}-M_B(+\infty)=65-62=\mathbf{3\,M_\odot}$ — positive, as $\dot M_B\le 0$ requires.

Step 3. In joules, with $1\,M_\odot=1.989\times10^{30}\,\mathrm{kg}$ and $c=2.998\times10^{8}\,\mathrm{m/s}$ : $\Delta E=\Delta M\,c^2=(5.97\times10^{30})(8.99\times10^{16})=\mathbf{5.4\times10^{47}\,J}$ .

Step 4. That $\approx3\,M_\odot c^2$ was emitted in $\approx0.2\,\mathrm{s}$ , a peak luminosity around $3.6\times10^{56}\,\mathrm{W}\approx 200\,M_\odot c^2\,\mathrm{s^{-1}}$ — for an instant, more power in gravitational waves than the light of every star in the observable universe. The $3\,M_\odot$ missing from GW150914 is exactly the energy the news $\sigma^0$ delivered to $\mathscr{I}^+$ , obeying $M_{\mathrm{ADM}}\ge M_B$ to the letter.

2 · Where a photon lands on scri, and the causal type of our own universe

Two short computations. (a) An outgoing radial light ray leaves the origin at $T=0$ and races outward forever; using (8), find the finite boundary coordinate where it terminates. (b) Our universe has $\Lambda\approx1.1\times10^{-52}\,\mathrm{m^{-2}}$ ; use (3) to classify its future infinity.

Step 1. For the ray $R=T$ , so $U=T-R=0$ while $V=T+R\to+\infty$ . In compactified coordinates $q=\arctan U=0$ (fixed) and $p=\arctan V\to\tfrac{\pi}{2}$ .

Step 2. So the ray ends at $p=\tfrac{\pi}{2}$ , $q=0$ — a single finite point on $\mathscr{I}^+$ , though it travelled infinitely far physically. An infinite journey ends at a finite address — the first figure made arithmetic.

Step 3. For (b), equation (3) gives $g^{ab}n_an_b\big|_{\mathscr{I}}=-\Lambda/3\approx\mathbf{-3.7\times10^{-53}\ m^{-2}}<0$ .

Step 4. A negative norm means a timelike normal, so future infinity is spacelike — the de Sitter case. An accelerating universe like ours has a spacelike $\mathscr{I}^+$ guarded by cosmological horizons: no observer sees all of it, and there is no globally defined total energy radiated to $\mathscr{I}^+$ — unlike the flat merger above. The same tiny number that drives cosmic acceleration fixes the shape of infinity.

Physics of the portrait

Conformal infinity is, more than almost any idea in relativity, the creation of one person's geometric imagination — turning "a system very far away" into a boundary you can draw, differentiate, and prove theorems about.

Origins · 1 minRoger Penrose and how this was discovered

Roger Penrose

1931– · conformal methods and the global structure of spacetime

Roger Penrose reshaped general relativity by importing the geometer's habit of thinking about boundaries and global structure into a subject dominated by local coordinate calculations. Between 1962 and 1965, using the spin-coefficient formalism he developed with Ezra Newman, he introduced the conformal treatment of infinity: rescale the metric by $\Omega^2$ , let $\Omega\to0$ far out, and attach the zero set as a boundary $\mathscr{I}$ . The move was quintessentially Penrose — trade the physical metric, singular at infinity, for a conformal class, which is not, keeping exactly the causal information that matters. From it he extracted the peeling theorem, the first rigorous definition of asymptotic flatness, and the conformal diagrams that now bear his name. The same way of seeing led him in 1965 to the singularity theorem that reopened gravitational collapse and, decades later, to the 2020 Nobel Prize. Conformal infinity is the quiet foundation beneath much of that work — the arena where gravitational radiation, the Bondi and ADM masses, the BMS group and the holographic boundary of AdS are made precise.

History

1962

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

1963

— Penrose introduces the conformal treatment of infinity, bringing infinity to a finite boundary

1965

— Penrose establishes the peeling property of the Weyl tensor for zero-rest-mass fields

1978

— Ashtekar and Hansen give a unified geometric treatment of spatial and null infinity

2014

— Strominger links BMS supertranslations to soft graviton theorems and gravitational-wave memory

Going further

Texts. Penrose & Rindler, Spinors and Space-Time, Vol. 2, ch. 9 · Wald, General Relativity, ch. 11 · Hawking & Ellis, Large Scale Structure, ch. 5 · Misner, Thorne & Wheeler, Gravitation, ch. 34 · Carroll, Spacetime and Geometry, app. H · Geroch, "Asymptotic Structure of Space-Time" (1977). Two points of detail are worth noting. In the compactified-Minkowski construction, $V-U=\sin(p-q)/\cos p\cos q$ and $\Omega=2\cos p\cos q$ cancels the prefactor exactly. On the causal classification: the boundary value is the finite $g^{ab}n_an_b|_{\mathscr{I}}=-\Lambda/3$ (from the $\Omega^2$ -weighted conformal Ricci equation evaluated on $\mathscr{I}$ ), not the expression " $-\tfrac13\Lambda\Omega^2$ " which vanishes there; accordingly the de Sitter and AdS sign entries in the limiting-cases table reflect the standard result (de Sitter spacelike, AdS timelike). GW150914 masses are $\approx36+29\to62\,M_\odot$ , giving $\approx3\,M_\odot c^2$ radiated and a peak luminosity $\sim3.6\times10^{56}\,\mathrm{W}$ . The cosmological constant used is $\Lambda\approx1.1\times10^{-52}\,\mathrm{m^{-2}}$ , giving $-\Lambda/3<0$ and hence a spacelike $\mathscr{I}^+$ .

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

Ask the tutor: Ask me to walk you through why de Sitter's future infinity is spacelike while a black hole's is null — and what that does to the very notion of "total radiated energy."

What is the difference between ℐ⁺ and i⁰?

Different pieces of infinity, reached by different curves. Future null infinity $\mathscr{I}^+$ is where outgoing light rays end — massless fields and gravitational radiation propagate there, hosting the Bondi mass and news function. Spatial infinity $i^0$ is a single point where spacelike geodesics end, where the conserved charges (ADM mass and momentum) are defined. Timelike geodesics — massive bodies at rest — end at $i^+$ . Loosely: massless things go to $\mathscr{I}^{\pm}$ , massive things to $i^{\pm}$ , and the whole spacetime at one instant taken to infinite radius lives at $i^0$ .

Is scri a part of the physical spacetime, or an add-on?

It is an idealised add-on — a boundary attached to the unphysical, rescaled manifold, not events any observer visits. In the physical metric $\mathscr{I}$ sits at infinite distance; no clock reaches it. Its value is as a limiting arena: outward-propagating fields have well-defined limits on $\mathscr{I}$ (the news function, peeling coefficients, asymptotic charges) capturing the invariant content of "what the system looks like far away." So $\mathscr{I}$ is real in that its boundary data are physical and measurable in principle, but it is not a place — it is the precise stand-in for "arbitrarily far, arbitrarily late."

Why is the Bondi mass smaller than the ADM mass?

Because they are one energy read at two pieces of infinity, separated by the radiation between them. The ADM mass at $i^0$ counts the total, including every graviton the system will ever emit. The Bondi mass $M_B(u)$ on $\mathscr{I}^+$ is measured after the radiation before $u$ has flowed off. Since $\dot M_B=-\tfrac{1}{4\pi}\oint|\sigma^0|^2\le0$ , $M_B$ only decreases, so $M_B(u)\le M_{\mathrm{ADM}}$ with equality as $u\to-\infty$ . The gap is exactly the radiated energy — the $3\,M_\odot$ of GW150914.

What is the BMS group, and why isn't it just the Poincaré group?

You might expect the transformations preserving the structure of $\mathscr{I}^+$ to form the Poincaré group — ten translations, rotations and boosts. Instead you get the Bondi–Metzner–Sachs group, which replaces the four rigid translations with an infinite-dimensional family of "supertranslations," angle-dependent shifts along the null generators of $\mathscr{I}^+$ . Null infinity is floppier than a point: a null hypersurface with $S^2$ cross-sections admits far more structure-preserving maps. The supertranslations are physical — they relabel gravitational-wave vacua and are the symmetry behind the memory effect and soft-graviton theorems.

How does conformal infinity connect to AdS/CFT and holography?

Through the anti-de Sitter case, $\Lambda<0$ , where equation (3) gives $g^{ab}n_an_b>0$ , a spacelike normal, hence a timelike boundary $\mathscr{I}$ . A timelike boundary is a lower-dimensional spacetime in its own right — signals reach it and return, so AdS is not globally hyperbolic and needs boundary conditions. The completion hands you that boundary with a conformal class of metrics, not a single metric. AdS/CFT lives on exactly this structure: bulk gravity is dual to a conformal field theory on the boundary — conformal precisely because the boundary geometry is defined only up to the rescaling $\Omega\to\omega\Omega$ .

Can every spacetime be conformally compactified?

No — a smooth conformal completion is a genuine restriction, essentially the technical content of "asymptotically flat" (or de Sitter / anti-de Sitter). One demands a factor $\Omega$ with $\Omega=0$ , $\mathrm{d}\Omega\neq0$ on a boundary attached smoothly to the manifold, with the physical metric solving the vacuum-with- $\Lambda$ equations near it. Spacetimes with the wrong falloff, or matter that does not die off fast enough, fail to admit one, and for them Bondi mass and peeling do not apply. Even in the flat case there are subtleties at $i^0$ , where the completion is generically only continuous — which is why $i^0$ needs its own treatment (Ashtekar–Hansen).