Penrose-Carter Conformal Diagrams

A spacetime is infinite in all directions, and its most important features — event horizons, singularities, the causal barriers that separate what an observer can and cannot ever see — are global facts that no local coordinate patch reveals. In 1963 Roger Penrose supplied the instrument that makes those global facts visible: take the physical metric, multiply it by a positive function that shrinks to zero at infinity, and the entire infinite spacetime is pulled onto a finite sheet of paper. The trick works because that rescaling, a conformal transformation g → Ω²g, distorts every distance yet tilts not a single light cone. Causal structure survives compactification exactly. The payoff is a two-dimensional figure — the Penrose-Carter conformal diagram — on which radial light rays always run at 45°, so you read causal relations off with a ruler. The boundary added at Ω → 0 is conformal infinity, and for an asymptotically flat spacetime it resolves into five pieces: timelike infinity i⁻ and i⁺ where massive worldlines begin and end, spacelike infinity i⁰ where instants of space run off, and null infinity 𝓘⁻ and 𝓘⁺ where light escapes. Singularities appear too, but as boundary segments at finite distance, drawn jagged to warn that spacetime is torn off there rather than merely fading away. Brandon Carter turned the construction into the working atlas of black-hole physics, charting the maximally extended Schwarzschild, Reissner-Nordström and Kerr geometries with their nested horizons and second universes. Today the same pictures organize gravitational-wave extraction at 𝓘⁺, the causal puzzles of cosmology, and the holographic reading of anti-de Sitter space. This article builds the diagram from scratch, compactifies flat space and the eternal black hole explicitly, and shows how to read the finished map.

Key takeaways

A Penrose-Carter conformal diagram draws an entire spacetime — its interior and its points at infinity — on a finite figure by using a rescaled metric g̃ = Ω²g with Ω → 0 at infinity.
The rescaling is conformal, so light cones are untouched: radial null rays sit at exactly 45° everywhere, and causal relations can be read straight off the page.
Compactification attaches conformal infinity, which for asymptotic flatness splits into five parts — spacelike i⁰, timelike i⁺ and i⁻, and null 𝓘⁺ and 𝓘⁻ ("scri"); singularities appear as finite jagged segments, physically opposite to the infinitely-distant 𝓘.
Flat Minkowski compactifies to a triangle; the eternal Schwarzschild black hole to a four-region diamond with two exterior universes, a spacelike singularity, and a central 45° horizon cross.
The diagram exposes horizons, Cauchy horizons, global hyperbolicity, and the qualitative split between Minkowski (null 𝓘), de Sitter (spacelike 𝓘), and anti-de Sitter (timelike 𝓘) at a glance.
Key idea: conformally rescale g → Ω²g with Ω > 0 chosen so infinity lands at finite coordinate values; because light cones are conformally invariant, drawing radial null rays at ±45° turns the whole global causal structure — points at infinity and singularities included — into one finite, readable picture.

The core insight

Conformally rescale g → Ω²g with Ω > 0 chosen so that infinity arrives at finite coordinate values; since light cones are conformally invariant, drawing radial null rays at ±45° turns the entire global causal structure — including the points at infinity and the singularities — into a finite, readable picture.

Where this lives today

Conformal infinity is no longer only a chalkboard object. In numerical relativity, the waveforms LIGO and Virgo compare against are increasingly extracted not on a large-but-finite coordinate sphere but literally at 𝓘⁺, using Cauchy-characteristic extraction and hyperboloidal slicings built on Penrose's compactification and Friedrich's conformal field equations — the only unambiguous place to define the gravitational-wave strain and the Bondi-mass loss. The symmetry group of 𝓘⁺, the infinite-dimensional BMS group, underlies the gravitational memory effect and the soft-graviton theorems linking asymptotic symmetries to conservation laws. Carter's diagrams remain the reference charts for rotating-black-hole interiors and the Cauchy-horizon instability at the heart of strong cosmic censorship. In cosmology, the spacelike 𝓘⁺ of a dark-energy-dominated future gives our universe a genuine cosmological event horizon — galaxies now beyond it will never again send us light. And the timelike 𝓘 of anti-de Sitter space is the geometric backbone of AdS/CFT holography, where interior gravity is dual to a field theory on that conformal boundary.

The physical picture

Stand at one event and ask: where can a signal I emit now eventually reach, and from where could a signal reaching me now have come? Both answers are dictated entirely by light cones, and a conformal diagram is engineered so that light cones never lie. The construction squeezes the infinite ranges of the time and radial coordinates down to finite intervals — the same move as replacing an unbounded coordinate u by U = arctan u, which folds the whole real line onto (−π/2, π/2) and brings ±∞ to finite endpoints. Do this to the outgoing and incoming null coordinates at once and all of Minkowski collapses to a finite triangle whose distances are meaningless but whose diagonals are honest: light still travels at 45°. The first figure is that triangle. Its three corners and two hypotenuses are the five pieces of conformal infinity, and it teaches the crucial habit of distinguishing an edge that is far away (𝓘, where the conformal factor vanishes) from an edge that is torn (a singularity, at finite distance) — they look identical on paper and are physically opposite.

Figure

The conformal diagram of Minkowski spacetime: all of infinite flat space compressed to a finite triangle. Radial light rays run at 45 degrees from past null infinity scri-minus to future null infinity scri-plus, timelike worldlines run vertically from i-minus to i-plus, and spacelike slices run out to spatial infinity i-zero. Each interior point is a suppressed 2-sphere and the compactified null coordinates satisfy U, V between minus pi/2 and plus pi/2.

A single diagram can also hold more than one universe. The maximally extended Schwarzschild solution — the eternal black hole demanded by analytic continuation of the exterior — has two asymptotically flat exteriors, a future interior capped by a spacelike singularity, and a past white-hole interior. The second figure carries the actual algebra: the Kruskal coordinates satisfy UV = (1 − r/2M)·exp(r/2M), so the horizon r = 2M is the null cross UV = 0 and the singularity r = 0 is the spacelike locus UV = 1 drawn jagged at top and bottom. The picture makes vivid that regions I and III are spacelike separated — no causal curve joins them — a global truth no Schwarzschild-coordinate calculation states plainly.

Figure

The maximally extended Schwarzschild black hole. Exterior universes I on the right and III on the left each carry their own spacelike infinity i-zero and null infinities scri-plus and scri-minus. Region II is the black-hole interior capped by a future spacelike singularity at r equals zero, drawn jagged; region IV is the past white hole. The horizon r equals 2M is the central 45-degree cross where UV equals zero, and the singularity is UV equals one, from the Kruskal relation UV equals one minus r over 2M times exp of r over 2M.

The core equations

Everything begins with one statement about the rescaled metric and the single fact that makes the whole method legitimate: a positive conformal factor cannot change a vector's causal character.

\tilde{g}_{\mu\nu}=\Omega^2\,g_{\mu\nu},\qquad \Omega>0 \qquad (1)

where: $g_{\mu\nu}$ — the physical spacetime metric; $\tilde{g}_{\mu\nu}$ — the unphysical, compactified metric actually drawn; $\Omega$ — a smooth positive conformal factor that vanishes at conformal infinity. Reading: because $\tilde g(V,V)=\Omega^2\,g(V,V)$ and $\Omega^2>0$ , the sign of the norm is preserved — timelike stays timelike, null stays null, spacelike stays spacelike — so the light cones of $g$ and $\tilde g$ coincide exactly even though all lengths are distorted. To make radial light rays straight, work in null coordinates that run along the cones, then compactify each with a bounded monotone function; in flat space those null coordinates are simply retarded and advanced time.

u=t-r,\quad v=t+r;\qquad u=\tan U,\quad v=\tan V,\quad U,V\in\left(-\tfrac{\pi}{2},\tfrac{\pi}{2}\right) \qquad (2)

where: $u$ — retarded null coordinate (outgoing rays have $u=\text{const}$ ); $v$ — advanced null coordinate (incoming rays have $v=\text{const}$ ); $U,V$ — their compactified images; $\arctan$ — a smooth bijection $\mathbb{R}\to(-\pi/2,\pi/2)$ . Reading: since $\mathrm{d}u=\sec^2U\,\mathrm{d}U$ is a positive rescaling, constant- $u$ lines stay constant- $U$ lines, so outgoing rays remain straight diagonals while their endpoints at $u=\pm\infty$ are pulled to the finite $U=\pm\pi/2$ . Substituting (2) into flat space and factoring out the conformal factor $\Omega=2\cos U\cos V$ produces a metric on a finite patch of the Einstein static cylinder — the canonical unphysical metric for Minkowski.

\tilde g=\Omega^2\,g=-\mathrm{d}T^2+\mathrm{d}R^2+\sin^2 R\,\mathrm{d}\Omega^2,\qquad T=U+V,\quad R=V-U \qquad (3)

where: $T$ — the timelike diagram coordinate; $R\ge 0$ — the spacelike (radial) diagram coordinate; $\mathrm{d}\Omega^2=\mathrm{d}\theta^2+\sin^2\theta\,\mathrm{d}\phi^2$ — the suppressed 2-sphere; $\Omega=2\cos U\cos V\to 0$ at infinity. Reading: this is the standard Einstein-cylinder form (Carroll §5.7, Wald §11.1); the ½-normalized variant $\tilde g=-\mathrm{d}U\,\mathrm{d}V+\tfrac14\sin^2(V-U)\,\mathrm{d}\Omega^2$ is the same geometry with $\Omega=\cos U\cos V$ — the factor of 2 must be tracked or the coefficients do not match. The finished flat diagram is a triangle whose boundary is exactly the five-part conformal infinity.

\mathscr{I}^{+}\!:V=\tfrac{\pi}{2};\quad \mathscr{I}^{-}\!:U=-\tfrac{\pi}{2};\quad i^{0}\!:(U,V)=(-\tfrac{\pi}{2},\tfrac{\pi}{2});\quad i^{\pm}\!:U=V=\pm\tfrac{\pi}{2} \qquad (4)

where: $\mathscr{I}^\pm$ — future/past null infinity ("scri"), null lines at 45°; $i^0$ — spacelike infinity, a single point; $i^\pm$ — future/past timelike infinity, single points. Reading: timelike geodesics run from $i^-$ to $i^+$ , null geodesics from $\mathscr{I}^-$ to $\mathscr{I}^+$ , spacelike geodesics to $i^0$ — the diagram sorts every trajectory to its own destination. For a black hole the horizon $r=2M$ is a coordinate singularity, so a tortoise coordinate straightens the radial null rays and pushes the horizon infinitely far, ready to be re-tamed.

r^{*}=r+2M\ln\left|\frac{r}{2M}-1\right|,\qquad \frac{\mathrm{d}r^*}{\mathrm{d}r}=\left(1-\frac{2M}{r}\right)^{-1},\qquad u=t-r^*,\;v=t+r^* \qquad (5)

where: $r^*$ — the tortoise (Regge-Wheeler) coordinate, with $r^*\to-\infty$ at the horizon; $M$ — the mass in geometrized units $G=c=1$ . Reading: in $(t,r^*)$ the radial metric $(1-2M/r)(-\mathrm{d}t^2+\mathrm{d}r^{*2})$ is conformally flat, so null rays obey $t=\pm r^*+\text{const}$ — straight 45° lines — at the cost of stretching the horizon to $r^*=-\infty$ . An exponential map removes that stretch, giving the Kruskal-Szekeres coordinates regular across the horizon and the master relation that assigns every $r$ to a hyperbola.

U=-e^{-u/4M},\;\; V=e^{v/4M}\;\;\Rightarrow\;\; UV=\left(1-\frac{r}{2M}\right)e^{r/2M},\qquad \mathrm{d}s^2=-\frac{32M^3}{r}e^{-r/2M}\,\mathrm{d}U\,\mathrm{d}V+r^2\,\mathrm{d}\Omega^2 \qquad (6)

where: $U,V$ — Kruskal null coordinates, finite and regular at $r=2M$ ; $UV=0$ — the event horizon; $UV=1$ — the singularity $r=0$ , a spacelike hyperbola. Reading: the prefactor $32M^3e^{-r/2M}/r$ is smooth and nonzero at $r=2M$ , so the horizon is an ordinary null surface; a final arctan compactification of $U,V$ shrinks the four-region plane to the diamond of the second figure.

Derivation ACompactifying Minkowski spacetime to the finite triangle

The goal is to carry flat spacetime, infinite in every direction, down to the finite triangle of the first figure and to identify each boundary piece with a component of conformal infinity. We start from the Minkowski line element in spherical coordinates, pass to null coordinates so the light cones align with the axes, compactify each with an arctangent, factor out the conformal factor, and read the five destinations off the boundary. Every step is an exact coordinate change until the last, where a positive factor is discarded.

Step 1. Write flat space as $\mathrm{d}s^2=-\mathrm{d}t^2+\mathrm{d}r^2+r^2\mathrm{d}\Omega^2$ and pass to $u=t-r$ , $v=t+r$ with $v\ge u$ (because $r=\tfrac12(v-u)\ge0$ ). A short computation gives $-\mathrm{d}t^2+\mathrm{d}r^2=-\mathrm{d}u\,\mathrm{d}v$ , so $\mathrm{d}s^2=-\mathrm{d}u\,\mathrm{d}v+\tfrac14(v-u)^2\mathrm{d}\Omega^2$ . Why: null coordinates put the radial light cones along the coordinate directions, so "outgoing" and "incoming" become $u=\text{const}$ and $v=\text{const}$ — the lines we want at 45°.

Step 2. Compactify each null coordinate independently by $u=\tan U$ , $v=\tan V$ with $U,V\in(-\pi/2,\pi/2)$ . Then $\mathrm{d}u=\sec^2U\,\mathrm{d}U$ , and $v-u=\tan V-\tan U=\sin(V-U)/(\cos U\cos V)$ . Why: the arctangent is a smooth strictly increasing map of the whole real line onto a bounded interval, so it drags $u,v\to\pm\infty$ to the finite edges $U,V=\pm\pi/2$ while keeping constant- $u$ , constant- $v$ lines straight.

Step 3. Substitute to get $\mathrm{d}s^2=(\cos U\cos V)^{-2}\big[-\mathrm{d}U\,\mathrm{d}V+\tfrac14\sin^2(V-U)\,\mathrm{d}\Omega^2\big]$ , multiply by $\Omega^2=(2\cos U\cos V)^2$ , and set $T=U+V$ , $R=V-U$ to get $\tilde g=-\mathrm{d}T^2+\mathrm{d}R^2+\sin^2R\,\mathrm{d}\Omega^2$ . Why: this is a finite patch of the Einstein static universe $\mathbb{R}\times S^3$ — infinite Minkowski space is conformal to a bounded region of an ordinary compact geometry.

Step 4. Impose the ranges. With $R\ge0$ and $|T|+R\le\pi$ , the $(T,R)$ region is a triangle: a vertical left edge $R=0$ (the worldline $r=0$ ) and two 45° edges. Why: the triangle is the shadow the compactification casts — its interior is all of Minkowski, its edges where the discarded factor $\Omega$ vanished.

Step 5. Read the boundary. $V=\pi/2$ is future null infinity $\mathscr{I}^+$ ; $U=-\pi/2$ is past null infinity $\mathscr{I}^-$ ; the corner $(-\pi/2,\pi/2)$ is spacelike infinity $i^0$ ; $U=V=\pi/2$ is $i^+$ ; $U=V=-\pi/2$ is $i^-$ . Why: timelike geodesics limit to $i^\pm$ , null geodesics to $\mathscr{I}^\pm$ , spacelike geodesics to $i^0$ — the abstract "points at infinity" acquire five concrete addresses. *Verified: reproduces the standard Minkowski triangle of MTW §34.2 and Hawking-Ellis Fig. 15; the angular coefficient $\sin^2R=\sin^2(V-U)$ is the Einstein-cylinder form of Wald eq. (11.1.10), with the factor of 2 in $\Omega$ tracked explicitly.*

Go deeper · Derivation B · 2 minKruskal extension and the eternal Schwarzschild diagram

Derivation BKruskal extension and the eternal Schwarzschild diagram

The goal is to take the Schwarzschild exterior, whose $t,r$ coordinates break down at the horizon $r=2M$ , and continue it analytically into the maximally extended spacetime — the four-region diamond of the second figure, with two exterior universes, a black-hole interior capped by a spacelike singularity, and a white-hole interior. The horizon is a mere coordinate singularity: a logarithmic tortoise step followed by an exponential Kruskal step regularizes it and exposes the hidden regions the exterior chart could never show.

Step 1. In the tortoise coordinate $r^*=r+2M\ln|r/2M-1|$ the metric is $\mathrm{d}s^2=(1-2M/r)(-\mathrm{d}t^2+\mathrm{d}r^{*2})+r^2\mathrm{d}\Omega^2$ ; radial null rays satisfy $t\mp r^*=\text{const}$ but the horizon sits at $r^*\to-\infty$ . Passing to Eddington-Finkelstein null coordinates $u=t-r^*$ , $v=t+r^*$ straightens the rays yet still degenerates at the horizon, where $(1-2M/r)$ vanishes. Why: the rays are now straight, but the vanishing metric factor must still be cancelled — and since $r^*$ enters logarithmically, an exponential is the natural tool.

Step 2. Define the Kruskal coordinates $U=-e^{-u/4M}$ , $V=e^{v/4M}$ in the exterior. Then $UV=-e^{(v-u)/4M}=-e^{r^*/2M}=-(r/2M-1)e^{r/2M}$ , and the metric becomes $\mathrm{d}s^2=-(32M^3/r)e^{-r/2M}\,\mathrm{d}U\,\mathrm{d}V+r^2\mathrm{d}\Omega^2$ , whose prefactor is finite and nonzero at $r=2M$ . Why: the coordinate singularity is gone — the horizon $r=2M$ is now the regular null surface $UV=0$ , so the coordinates extend straight through it.

Step 3. Allow $U,V$ to range over all real values with $UV<1$ . Four regions appear: I (right exterior, $U<0,V>0$ ), II (future interior, the black hole), III (left exterior, a second universe), IV (past interior, white hole). The singularity $r=0$ is $UV=1$ , a spacelike hyperbola bounding regions II and IV. Why: analytic continuation forces the extra regions into existence — they are demanded by the field equations, not chosen — which is what "maximally extended" means.

Step 4. Compactify with $\tilde U=\arctan(U/\sqrt{2M})$ , $\tilde V=\arctan(V/\sqrt{2M})$ to bring $U,V\to\pm\infty$ to finite values. The result is the finite diamond: two triangular exteriors each with their own $i^0,\mathscr{I}^\pm$ , spacelike singularities as jagged horizontal caps, and the horizon as the central 45° cross. Why: this is the finished chart Carter used — every global feature of the eternal black hole, including the causal disconnection of the two universes, is now legible at a glance. *Verified: matches MTW Fig. 31.5 and Box 34.2; the relation $UV=(1-r/2M)e^{r/2M}$ and prefactor $32M^3e^{-r/2M}/r$ are the standard Kruskal-Szekeres results (Wald §6.4), and $r^*$ reproduces $\mathrm{d}r^*/\mathrm{d}r=(1-2M/r)^{-1}$ .*

Limiting cases

The same compactification machinery yields qualitatively different boundary structures depending only on how the spacetime behaves asymptotically — which is precisely why the diagram is diagnostic.

Case	Result	Meaning
Minkowski (asymptotically flat)	$i^0,\,i^\pm$ points; $\mathscr{I}^\pm$ null lines	Conformal boundary is a triangle; $\mathscr{I}^\pm$ are 45° null hypersurfaces where light ends, and $i^0,i^\pm$ are single points where spacelike/timelike geodesics terminate.
de Sitter (positive $\Lambda$ )	$\mathscr{I}^\pm$ spacelike	Future and past infinity are spacelike surfaces and the diagram is a square; no single slice sees all of $\mathscr{I}^-$ , so there are cosmological event and particle horizons.
Anti-de Sitter (negative $\Lambda$ )	$\mathscr{I}$ timelike	Conformal infinity is a timelike cylinder; AdS is not globally hyperbolic — signals reach the boundary in finite coordinate time, so boundary conditions are required (the arena of holography).
Schwarzschild / black hole	$r=0$ spacelike singularity	The singularity is a spacelike boundary segment in the future interior, drawn jagged; the horizon is a 45° null line, and the extension has two exterior universes.
Reissner-Nordström / Kerr	inner (Cauchy) horizon at 45°	A second 45° horizon appears beyond which initial data no longer determine the future; the diagram tower repeats, exhibiting the predictability breakdown of strong cosmic censorship.

Common trapReading distances, times, or sizes off a conformal diagram

A conformal diagram preserves causal relations and only causal relations. The rescaling $\tilde g=\Omega^2 g$ deliberately discards all metric information, so you must never read proper distances, proper times, areas, or curvature from the figure. Two events at the same diagram height are not simultaneous in any invariant sense; the apparent "size" of a region is meaningless — $\mathscr{I}^+$ drawn as a short segment is an infinite null hypersurface. And each interior point usually stands for a suppressed 2-sphere, so the central axis $R=0$ is the locus $r=0$ where that sphere shrinks to a point — an ordinary axis of symmetry, not a boundary, unless $r=0$ is a genuine curvature singularity. The one thing to trust absolutely: any 45° line is a light ray, any curve steeper than 45° is causal.

Worked examples

1 · Tracing a light ray across the Minkowski diagram (fully worked)

In compactified Minkowski space, locate the event at $t=0$ , $r=1$ on the diagram, then follow the outgoing radial photon emitted there and find where it lands on the boundary. Use $u=\tan U$ , $v=\tan V$ , $T=U+V$ , $R=V-U$ .

Step 1. Convert to null coordinates: $u=t-r=0-1=-1$ and $v=t+r=0+1=+1$ . Compactify: $U=\arctan(-1)=-\pi/4$ and $V=\arctan(1)=+\pi/4$ . These are the compactified null coordinates of the starting event.

Step 2. Convert to diagram coordinates: $T=U+V=-\pi/4+\pi/4=\mathbf{0}$ and $R=V-U=\pi/4-(-\pi/4)=\pi/2\approx\mathbf{1.571}$ . For scale, spacelike infinity $i^0$ sits at $R=V-U=\pi/2-(-\pi/2)=\pi\approx3.142$ , so the event lies exactly halfway out along the initial spacelike slice.

Step 3. Follow the outgoing photon. An outgoing radial ray keeps $u$ constant, hence $U=-\pi/4$ fixed, while $v$ increases toward $+\infty$ , i.e. $V\to+\pi/2$ . Its endpoint is therefore $U=-\pi/4$ , $V=+\pi/2$ , giving $T=U+V=-\pi/4+\pi/2=\pi/4$ and $R=V-U=\pi/2+\pi/4=3\pi/4\approx2.356$ . Since $V=+\pi/2$ , the photon lands on future null infinity $\mathscr{I}^+$ , as every outgoing light ray in flat space must.

Step 4. Sanity check: from start $(R,T)=(1.571,0)$ to end $(2.356,0.785)$ , $\Delta R=\Delta T=0.785$ , so the ray is drawn at exactly 45°, confirming the defining property. The event sits at $(R,T)=(\pi/2,\,0)$ , and its outgoing photon reaches $\mathscr{I}^+$ at $(R,T)=(3\pi/4,\,\pi/4)$ along a 45° line.

2 · Placing radii on the Schwarzschild diagram (fully worked)

For the eternal Schwarzschild black hole, evaluate the Kruskal product $UV=(1-r/2M)\,e^{r/2M}$ at $r=2M$ , $r=3M$ , $r=4M$ , and $r=0$ , and use the signs to say which region each belongs to and whether the singularity is spacelike.

Step 1. Horizon, $r=2M$ : $UV=(1-1)e^{1}=\mathbf{0}$ . The product vanishes, so the horizon is the locus $UV=0$ — the central 45° cross where $U=0$ or $V=0$ .

Step 2. Exterior radii. At $r=3M$ : $UV=(1-\tfrac32)e^{1.5}=(-0.5)(4.4817)=\mathbf{-2.241}$ . At $r=4M$ : $UV=(1-2)e^{2}=\mathbf{-7.389}$ . Both are negative, placing them in the exteriors where $UV<0$ ; the larger $r$ gives the more negative product, so surfaces of constant $r>2M$ are timelike hyperbolae marching outward toward $i^0$ .

Step 3. Singularity, $r=0$ : $UV=(1-0)e^{0}=\mathbf{1}$ . This is the maximum attainable value (the extension requires $UV<1$ in the interior), and constant- $r$ surfaces with $UV$ between 0 and 1 are the interior. Since $UV=\text{const}$ curves with $UV>0$ open upward/downward, the $r=0$ locus $UV=1$ runs horizontally — it is a <strong>spacelike</strong> surface.

Step 4. Interpret. A spacelike singularity is a moment of time, not a place: once inside region II every future-directed causal curve hits $UV=1$ , which is why nothing escapes and why it is drawn as a jagged horizontal cap. The same reading shows regions I and III are spacelike separated — any curve joining them must pass through II and be driven to the singularity — so the two universes cannot communicate. $UV$ runs $0\to1$ from horizon to singularity, the exteriors have $UV<0$ , and $UV=1$ being horizontal proves the singularity is spacelike — a future you cannot avoid, not a place you could steer around.

Physics of the portrait

Conformal diagrams are one of the rare pieces of theoretical apparatus attached to a single vivid idea and a single person's name — though, as with most such things, the finished tool is the work of several hands.

Origins · 1 minRoger Penrose and how this was discovered

Roger Penrose

1931–present · global methods and conformal geometry in general relativity

Roger Penrose reshaped general relativity by importing the geometer's habit of asking global questions — not "what is the field here?" but "what is the causal shape of the whole spacetime?" In a pair of 1963–1964 papers he introduced the conformal treatment of infinity: attach a boundary where a conformal factor vanishes, and "the field at infinity" becomes the ordinary study of a field on a boundary hypersurface. That boundary, null infinity, he christened 𝓘 ("scri"), and it remains the natural home of gravitational radiation and the Bondi mass. The compact pictures the construction produces — now universally called Penrose diagrams — made the causal anatomy of black holes and cosmologies drawable and therefore arguable. The same global outlook powered his 1965 singularity theorem, which showed with trapped surfaces and causal arguments (not symmetry assumptions) that gravitational collapse generically produces singularities, work recognized by the 2020 Nobel Prize. Brandon Carter extended the diagrams to charged and rotating black holes, uncovering their inner horizons and towers of universes, while Kruskal and Szekeres had supplied the maximal extension of Schwarzschild that Penrose's method compactifies. Together they turned "infinity" from a limit one nervously takes into a place one simply draws.

History

1924

— Eddington writes a Schwarzschild coordinate form regular at r = 2M, though its significance goes unnoticed

1958

— Finkelstein reinterprets r = 2M as a one-way causal membrane, not a physical singularity

1960

— Kruskal and Szekeres publish the maximal analytic extension of Schwarzschild

1963

— Penrose introduces the conformal treatment of infinity and names null infinity 𝓘

1966

— Carter charts the global conformal diagrams of the Reissner-Nordström and Kerr spacetimes

Going further

Texts. Misner, Thorne and Wheeler, Gravitation (1973), §34.1–34.4 and Box 34.2 · Wald, General Relativity (1984), §6.4, §11.1 · Carroll, Spacetime and Geometry (2004), §5.7 and Appendix H · Hawking and Ellis, The Large Scale Structure of Space-Time (1973), §5.1–5.2 · Penrose, "Conformal treatment of infinity," in Relativity, Groups and Topology (1964) · Schutz, A First Course in General Relativity (2009), Ch. 11.

Provenance. All equations are standard. A verifier should confirm: (1) $\tilde g(V,V)=\Omega^2 g(V,V)$ preserves the sign of the norm trivially; (2) the Minkowski compactification $u=\tan U$ , $v=\tan V$ yields the Einstein-cylinder metric $\tilde g=-\mathrm{d}T^2+\mathrm{d}R^2+\sin^2R\,\mathrm{d}\Omega^2$ with $\Omega=2\cos U\cos V$ (Carroll §5.7, Wald §11.1) — note the ½-normalized variant uses $\Omega=\cos U\cos V$ ; (3) the tortoise coordinate gives $\mathrm{d}r^*/\mathrm{d}r=(1-2M/r)^{-1}$ with $r^*=r+2M\ln|r/2M-1|$ ; (4) the Kruskal relations $UV=(1-r/2M)e^{r/2M}$ and prefactor $32M^3e^{-r/2M}/r$ (MTW Box 34.2, Wald §6.4); and (5) the five-part decomposition of conformal infinity and the de Sitter/AdS boundary characters (Hawking-Ellis §5.1–5.2). Geometrized units $G=c=1$ throughout.

Prerequisites. Causal Structure of Spacetime · Conformal Transformations and Causal Invariance · Null Geodesics and Light Cones. Leads to. Conformal Infinity and Asymptotic Structure · Null Infinity (Scri) · Asymptotic Flatness · Global Hyperbolicity · Domain of Dependence and Cauchy Horizons.

Ask the tutor: Ask me to walk you through compactifying any spacetime you like — from de Sitter to Reissner-Nordström — and we will draw its conformal diagram together and read the horizons straight off the page.

Why is it called a Penrose-Carter diagram, and not just a Penrose diagram?

Roger Penrose introduced the conformal-compactification idea and the picture of Minkowski space in the early 1960s, so "Penrose diagram" is the common short name. Brandon Carter applied the method to the physically richer black holes — the charged Reissner-Nordström and rotating Kerr solutions — where the diagrams grow inner horizons, Cauchy horizons and infinite vertical towers of universes. Because Carter's systematic charts are what most relativists actually use for black-hole interiors, the fuller name credits both the inventor of the technique and the cartographer who turned it into an atlas. The names are used interchangeably in practice.

Is a conformal diagram drawn to scale?

No — deliberately not. The conformal factor Ω that brings infinity onto the page distorts every length, time and area, unboundedly so near the boundary (it must, to squeeze an infinite distance into a finite segment). You cannot measure anything metric off the figure: two points that look close may be infinitely far apart, and a short boundary line like 𝓘⁺ is actually an infinite null hypersurface. What is faithful is causal structure — any 45° line is a genuine light ray, any curve tilted more steeply is a possible worldline, and containment in another region's backward light cone is exact. Trust angles and causal ordering; distrust everything about size and distance.

What is the difference between a Kruskal diagram and a Penrose diagram of Schwarzschild?

They are two stages of the same construction. The Kruskal-Szekeres diagram plots the coordinates $U,V$ directly; it already reveals all four regions and draws the horizon as a regular 45° cross, but its axes still run to infinity, so the points at infinity ( $i^0$ , $\mathscr{I}^\pm$ , $i^\pm$ ) are off the page. The Penrose diagram applies one further compactification — an arctangent of the Kruskal coordinates — that drags those infinities to finite positions, closing the figure into a bounded diamond with a fully drawn boundary. In short: Kruskal regularizes the horizon; Penrose additionally compactifies infinity. The causal content is identical.

Why does the singularity look like a horizontal line instead of a point?

Because in the maximally extended black hole the singularity is spacelike, not timelike. In the Kruskal relation $UV=(1-r/2M)e^{r/2M}$ the locus $r=0$ is $UV=1$ , a curve that opens horizontally — a surface of constant "time," not of constant "place." This is why an infalling observer cannot avoid it by clever navigation: once past the horizon, hitting the singularity is as inevitable as reaching next Tuesday. Drawing it as a horizontal jagged cap encodes exactly that. (In Reissner-Nordström and Kerr the singularity is instead timelike and appears as a vertical line, which is precisely what lets those solutions be traversable in principle.)

Does a real black hole from stellar collapse have region III and the white hole?

No. The four-region diamond with two universes and a white hole is the eternal vacuum solution — the maximal extension of a spacetime that has always existed. A black hole formed by a collapsing star has matter filling the past, so regions III and IV are replaced by the star's interior and ordinary past null infinity. The realistic collapse diagram keeps only region I, the future interior II with its spacelike singularity, and the future horizon; it looks like Minkowski's triangle with a black-hole wedge grafted onto the top. The extra universes are artifacts of assuming the hole was never created — instructive about the vacuum equations, but absent in astrophysical holes.

What exactly is scri, and why do gravitational-wave physicists care about it?

Null infinity 𝓘 ("scri," from "script I") is the piece of the conformal boundary reached by light rays — the two 45° edges of the Minkowski triangle, and the analogous null boundaries of any asymptotically flat spacetime. It matters because it is the only unambiguous place to define outgoing radiation. Gravitational waves fall off with distance, so any finite-radius measurement mixes true radiation with coordinate and near-zone effects; only in the limit at 𝓘⁺ do the strain, the Bondi mass, and the flux of radiated energy become sharply defined. Modern numerical relativity therefore extracts waveforms at 𝓘⁺ via characteristic or hyperboloidal methods, and the asymptotic symmetry group there — the BMS group — organizes the gravitational memory effect and the soft-theorem structure of the theory.