Penrose Singularity Theorem

For half a century after Karl Schwarzschild wrote down his 1916 solution, the singularity at its centre was regarded with suspicion. It sat at the heart of a perfectly spherical, static, idealized geometry, and the exact dust-collapse model of Oppenheimer and Snyder in 1939 inherited the same flawless symmetry. Lev Landau's school — Lifshitz, Khalatnikov, and later Belinski — argued that the singularity was a measure-zero accident: perturb the symmetry, add a little rotation or lumpiness, and the collapsing matter would swirl, miss the centre, and bounce. On that reading the infinities of Schwarzschild were an artefact of a special solution, not a prediction of the theory. Roger Penrose overturned this picture in a single four-page paper in 1965, assuming no symmetry whatsoever. His weapon fused two ideas from opposite ends of mathematics. The local half is the Raychaudhuri equation: under the mildest energy assumption gravity focuses light, so any bundle of rays that has begun to converge is doomed to pinch to a caustic within a bounded affine distance. The global half is topology: the causal-future boundary of a trapped surface — a compact surface from which even the outgoing light is converging — would, if every ray ran to its caustic, be a compact object that cannot be fitted into an open universe. The collision of these two facts is a contradiction, and the only escape is that some light ray simply ends: null geodesic incompleteness, Penrose's rigorous operational meaning of a singularity. This is the theorem that made singularities a robust prediction of general relativity, seeded the cosmic censorship programme, and earned the 2020 Nobel Prize in Physics.

Key takeaways

The Penrose singularity theorem (1965) proves that gravitational collapse generically produces a singularity, using no assumption of symmetry — killing the pre-1965 hope that asymmetry would let collapsing matter bounce.
Its three hypotheses are sharp: the null energy condition $R_{ab}k^ak^b\ge 0$ , the existence of a closed trapped surface $T$ , and global hyperbolicity with a non-compact Cauchy surface $\Sigma$ .
The conclusion is null geodesic incompleteness — at least one inextendible null geodesic runs out of affine parameter after a finite length — which Penrose adopted as the definition of a singularity.
The analytic engine is the Raychaudhuri focusing theorem: a converging null congruence reaches a caustic within $\Delta\lambda\le 2/|\theta_0|$ , capping the causal-future boundary $\dot J^+(T)$ .
The global engine is topological: focusing makes $\dot J^+(T)$ compact, but a compact, open, closed image inside a connected non-compact $\Sigma$ is impossible — so a generator must be incomplete.
It proves incompleteness, NOT that curvature diverges; the two coincide for Schwarzschild but are logically independent.
Key idea: a trapped surface makes both null normal congruences converge, and the focusing theorem then caps its future boundary into a compact achronal surface; that compact cap cannot be projected into a non-compact Cauchy surface without one light ray ending first — and that ending is the singularity.

The core insight

A trapped surface makes both null normal congruences converge ( $\theta<0$ ); the Raychaudhuri focusing theorem then forces each generator to reach a caustic within finite affine parameter, capping the future boundary $\dot J^+(T)$ into a compact achronal surface. In a globally hyperbolic spacetime that cap projects injectively and openly onto a Cauchy surface $\Sigma$ , giving a compact-open-closed subset — impossible when $\Sigma$ is connected and non-compact. The only escape is that some generator is incomplete: a singularity.

Where this lives today

The theorem's fingerprints are all over modern gravitational physics. In numerical relativity, the codes that model the binary black-hole mergers seen by LIGO and Virgo never wait for a global event horizon; they run apparent-horizon finders that solve $\theta_+=0$ on each time slice, detecting exactly the marginally trapped surfaces whose existence Penrose showed guarantees a singularity behind them. The trapped surface is also the object at the centre of the weak cosmic censorship conjecture — Penrose's proposal that the singularities his theorem produces are generically clothed by event horizons — and of the Penrose inequality $M\ge\sqrt{A/16\pi}$ (in $G=c=1$ ) bounding the mass by the area of the outermost trapped surface, proved in the time-symmetric case by Huisken–Ilmanen and Bray and still open in general. Trapped-surface estimates also bound black-hole production in trans-Planckian particle collisions. And because the theorem isolates the null energy condition as its sole physical assumption, it defines the precise target for every singularity-resolution programme — loop quantum cosmology bounces, string-inspired non-singular models, and effective-field-theory scenarios all arrange a controlled violation of that one condition.

The physical picture

The theorem has an unusual shape for a physics result: it is a proof by contradiction that plays a local focusing fact against a global topological fact, and neither half means much without the other. The first figure lays out that architecture as a single spacetime picture. A trapped surface $T$ sits at the bottom; the outgoing rays leaving it do not fan out as they would from an ordinary sphere but converge, and the null energy condition keeps them converging until they cross at a caustic. Those rays trace $\dot J^+(T)$ , the boundary of everything $T$ can causally influence, and because every ray caps off at a caustic within a bounded affine length, that boundary closes into a compact lid over $T$ .

Figure

The two-pronged proof: outgoing null generators leave a trapped surface T and focus to a caustic within a bounded affine length set by the initial expansion, capping the causal-future boundary ∂J⁺(T) into a compact achronal surface — but its projection down a timelike flow onto a non-compact Cauchy surface Σ would be a compact, open and closed piece of Σ, which is impossible, so one generator must be incomplete

Now the topology bites. In a globally hyperbolic spacetime every event is threaded by exactly one integral curve of a timelike flow crossing the Cauchy surface $\Sigma$ once. Flowing the compact lid $\dot J^+(T)$ down that flow projects it onto $\Sigma$ as a compact patch — and because the lid is an edgeless three-dimensional achronal surface mapped injectively into another three-dimensional surface, the projection is also open. A non-empty subset of a connected space that is at once open, closed, and compact must be the whole space; but $\Sigma$ is non-compact, so it cannot equal a compact patch. Something is false, and the only assumption that can give way is that every generator ran all the way to its caustic. One did not: it is incomplete, and that incompleteness is the singularity. The second figure quantifies the local half — the universal focusing curve behind the caustic.

Figure

The focusing blow-up: the curve plotted is the equality case θ(λ) = θ₀ / (1 + ½θ₀λ), the shear-free solution of dθ/dλ = −½θ², which plunges to −∞ at an affine length of 2 divided by the initial expansion. A general congruence with negative initial expansion obeying the null energy condition satisfies dθ/dλ ≤ −½θ², so it focuses at or before that curve does, provided the geodesic extends that far. Plotted in universal units, θ scaled by its initial value against affine parameter scaled by the same, the bounding caustic sits at scaled length 2 — for a solar-mass horizon a caustic within 2.95 km

The core equations

The theorem is a logical implication, so the cleanest first equation is its statement: a conjunction of three hypotheses on the left, one conclusion on the right. Everything below unpacks the two hypotheses that do the work and the machinery that connects them.

\big[\text{NEC}\big] + \big[\text{trapped surface } T\big] + \big[\text{glob. hyp., non-compact } \Sigma\big] \;\Rightarrow\; \exists\ \text{incomplete null geodesic} \qquad (1)

where: NEC — the null energy condition; $T$ — a closed trapped surface; $\Sigma$ — a Cauchy surface, assumed non-compact; the conclusion is null geodesic incompleteness, Penrose's operational definition of a singularity. Read it as: mild positive energy plus one strongly self-gravitating surface plus an open universe forces a light ray to end. The single physical assumption is the null energy condition — the weakest of the classical energy conditions. It is all the theorem asks of the matter, and identifying it as the only physical input is what later pinpointed how singularities might be avoided.

R_{ab}\,k^a k^b \ge 0 \quad \text{for all null } k^a \qquad (2)

where: $R_{ab}$ — Ricci tensor; $k^a$ — any future-directed null vector, $k_ak^a=0$ . Through Einstein's equations this is equivalent to $T_{ab}k^ak^b\ge 0$ — the energy density is non-negative in every null frame — because the trace term $\tfrac12 R g_{ab}k^ak^b$ vanishes for null $k^a$ . The trigger of the theorem is the closed trapped surface: a compact spacelike 2-surface both of whose future-directed orthogonal null congruences are converging. In flat space the outgoing rays from a sphere always expand; a trapped surface is where gravity has become strong enough to turn even the outgoing rays inward.

\theta_{(\ell)} < 0 \quad\text{and}\quad \theta_{(n)} < 0 \ \text{ everywhere on } T \qquad (3)

where: $T$ — compact, orientable, spacelike 2-surface; $\ell^a$ — future-directed outgoing null normal; $n^a$ — future-directed ingoing null normal; $\theta_{(\ell)},\theta_{(n)}$ — the expansions of the two null normal congruences. Reading: both wavefronts leaving $T$ lose area, so no signal born on $T$ can enlarge the region it bounds. The dynamics of any expansion is governed by the null Raychaudhuri equation. For a hypersurface-orthogonal (twist-free) congruence the crucial feature is that every term on the right is non-positive once the null energy condition holds.

\frac{\mathrm{d}\theta}{\mathrm{d}\lambda} = -\tfrac12\theta^2 - \sigma_{ab}\sigma^{ab} - R_{ab}k^a k^b \qquad (4)

where: $\theta$ — expansion of the null congruence; $\lambda$ — affine parameter along the generators; $\sigma_{ab}$ — shear, giving $-\sigma_{ab}\sigma^{ab}\le 0$ ; $R_{ab}k^ak^b$ — the Ricci focusing term, $\ge 0$ under the NEC. Reading: gravity acts as a lens that can only converge light — a congruence that starts contracting is driven to contract faster. Integrating that inequality gives the focusing theorem, the analytic heart. A converging null congruence must reach a caustic within a finite affine parameter set entirely by its initial expansion.

\frac{1}{\theta(\lambda)} \ge \frac{1}{\theta_0} + \frac{\lambda}{2} \;\Rightarrow\; \theta \to -\infty \ \text{within}\ \Delta\lambda \le \frac{2}{|\theta_0|} \qquad (5)

where: $\theta_0<0$ — initial expansion on the trapped surface; $\Delta\lambda$ — affine parameter to the conjugate point (caustic). Reading: because $\theta_0<0$ , the reciprocal $1/\theta$ climbs from a negative value through zero by $\lambda=2/|\theta_0|$ , forcing $\theta\to-\infty$ — provided the geodesic extends that far. Compactness of $T$ upgrades the pointwise focusing bound to a single uniform one, which makes $\dot J^+(T)$ compact and delivers the final contradiction with the non-compact Cauchy surface.

f:\dot J^+(T)\hookrightarrow\Sigma \ \text{continuous, injective, open};\qquad f\big(\dot J^+(T)\big)\ \text{compact}\subsetneq\Sigma\ \text{non-compact} \qquad (6)

where: $\dot J^+(T)$ — topological boundary of the causal future of $T$ ; $f$ — projection of $\dot J^+(T)$ along a timelike congruence onto $\Sigma$ . Reading: a compact set that is also open and closed inside a connected non-compact $\Sigma$ cannot exist, so the completeness assumption behind " $\dot J^+(T)$ is compact" must fail.

Derivation AFocusing theorem — a converging null congruence must reach a caustic

The goal is to prove the analytic engine of the theorem: a null congruence whose expansion is negative at some point, obeying only the null energy condition, is forced to a caustic within an affine length no greater than $2/|\theta_0|$ — unless the geodesic is incomplete and simply ends first. We start from the null Raychaudhuri equation for a hypersurface-orthogonal congruence, exploit the sign of every term, linearize the resulting Riccati inequality, and read off the caustic distance.

Step 1. Write the null Raychaudhuri equation with the twist restored, $\dfrac{\mathrm{d}\theta}{\mathrm{d}\lambda}=-\tfrac12\theta^2-\sigma_{ab}\sigma^{ab}+\omega_{ab}\omega^{ab}-R_{ab}k^ak^b$ , and specialize to a congruence orthogonal to the surface $T$ . By Frobenius' theorem such a congruence is twist-free, $\omega_{ab}=0$ . Why: the twist is the only term that could appear with a defocusing sign, and hypersurface-orthogonality — automatic for the normals to a surface — removes it.

Step 2. Impose the null energy condition. Through Einstein's equations $R_{ab}k^ak^b=8\pi G\,T_{ab}k^ak^b\ge 0$ , so the focusing term contributes $-R_{ab}k^ak^b\le 0$ ; the shear term $-\sigma_{ab}\sigma^{ab}\le 0$ is automatically non-positive since the shear is spacelike. Why: this is the sole physical assumption in the entire theorem, and it enters here and only here — everything downstream is geometry and topology.

Step 3. Discard the two non-positive terms to obtain the differential inequality $\dfrac{\mathrm{d}\theta}{\mathrm{d}\lambda}\le -\tfrac12\theta^2$ . Why: dropping non-positive terms only makes the right-hand side larger, so any bound proved for the simplified inequality holds a fortiori for the true evolution — the real congruence focuses at least this fast.

Step 4. Linearize by substituting $u=1/\theta$ , giving $\dfrac{\mathrm{d}u}{\mathrm{d}\lambda}=-\dfrac{1}{\theta^2}\dfrac{\mathrm{d}\theta}{\mathrm{d}\lambda}\ge\tfrac12$ , hence $u(\lambda)\ge u_0+\tfrac12\lambda$ with $u_0=1/\theta_0$ . Why: the substitution converts the nonlinear Riccati inequality into a harmless linear one — the quadratic blow-up of $\theta$ becomes the linear growth of its reciprocal.

Step 5. Because $\theta_0<0$ we have $u_0<0$ , and the increasing reciprocal must reach zero no later than $\lambda_\star=-2u_0=2/|\theta_0|$ . As $u\to 0^-$ , $\theta=1/u\to-\infty$ : a conjugate point, where neighbouring generators cross and the transverse area collapses to zero. Why: a divergence of $\theta$ to minus infinity is exactly a caustic — the light rays have been focused onto one another.

Step 6. State the dichotomy. Either the generator extends to $\lambda_\star$ and forms its caustic, or it is inextendible before $\lambda_\star$ — that is, incomplete. This fork is the seed from which the global argument grows a contradiction. Why: the theorem never shows a caustic actually forms; it shows a caustic is forced unless a geodesic ends, and it is the ending that becomes the singularity. *Note: integrating the equality $\dot\theta=-\tfrac12\theta^2$ exactly gives $\theta(\lambda)=\theta_0/(1+\tfrac12\theta_0\lambda)$ , which diverges precisely at $\lambda=2/|\theta_0|$ ; the inequality only shortens the distance.*

Go deeper · Derivation B · 3 minFrom focusing to incompleteness — Penrose's global topological argument

Derivation BFrom focusing to incompleteness — Penrose's global topological argument

The goal is to convert the local caustic bound into a global theorem by assuming the opposite of the conclusion and deriving a contradiction. Suppose the spacetime is null geodesically complete, globally hyperbolic with a non-compact Cauchy surface $\Sigma$ , obeys the null energy condition, and contains a trapped surface $T$ . We follow the causal-future boundary of $T$ , show completeness plus focusing would make it compact, then show no such compact object embeds in an open universe — so completeness fails.

Step 1. Characterize the boundary. The set $\dot J^+(T)=\partial J^+(T)$ is achronal, and away from $T$ it is generated by null geodesics leaving $T$ orthogonally, each remaining on the boundary only up to its first conjugate point. Why: past a caustic, neighbouring generators have crossed, so a point further along can be reached by a strictly timelike curve — it lies in the interior $I^+(T)$ , not on the boundary.

Step 2. Get a uniform focusing bound. Both expansions are negative on $T$ and continuous on the compact surface, so $\theta_0=\max(\sup_T\theta_{(\ell)},\ \sup_T\theta_{(n)})<0$ is a single finite negative number, and every generator reaches its caustic by $\lambda_\star=2/|\theta_0|$ . Why: compactness turns the pointwise "each ray eventually focuses" into "all rays have focused by one common affine parameter", which is what makes the boundary bounded rather than merely eventually-capping.

Step 3. Conclude the boundary is compact. Under the completeness assumption every generator runs all the way to its caustic, so $\dot J^+(T)$ is the image of the compact set $\{(p,\lambda):p\in T,\ 0\le\lambda\le\lambda_\star(p)\}$ under the continuous null exponential map — hence compact. It is also an edgeless achronal three-manifold without boundary. Why: this is the step that spends the completeness hypothesis; drop it and a generator could simply end before its caustic, leaving the boundary open.

Step 4. Project onto the Cauchy surface. Choose a timelike vector field $V^a$ whose integral curves each meet $\Sigma$ exactly once — global hyperbolicity guarantees one exists. Flowing along $V^a$ defines $f:\dot J^+(T)\to\Sigma$ , injective because $\dot J^+(T)$ is achronal, so no two of its points share an integral curve. Why: this transplants the compact lid into the Cauchy surface, where its compactness can be pitted against the non-compactness of an open universe.

Step 5. Show the image is open, hence clopen. The map $f$ sends an edgeless three-manifold injectively and continuously into the three-manifold $\Sigma$ , so by invariance of domain it is an open map: $f(\dot J^+(T))$ is open in $\Sigma$ , and being the continuous image of a compact set it is also compact and therefore closed. Why: invariance of domain is the modern rendering of Penrose's original "an open map of a compact manifold" reasoning, and it is exactly why equal dimensions matter.

Step 6. Spring the contradiction. A non-empty subset of a connected space that is simultaneously open and closed must be the whole space, so $f(\dot J^+(T))=\Sigma$ . But $f(\dot J^+(T))$ is compact while $\Sigma$ is non-compact — impossible. Why: the open universe has no proper clopen subset and cannot be exhausted by a compact patch, so a premise is false.

Step 7. Identify which premise fails. Every hypothesis except completeness is assumed as data; the contradiction therefore refutes completeness. Some null generator of $\dot J^+(T)$ never reached its caustic — it is inextendible after finite affine length. That incomplete geodesic is the singularity, forced with no appeal to symmetry. Why: geodesic incompleteness is the coordinate-free meaning of a singularity — a photon whose history simply ends — and it is the theorem's true conclusion.

Limiting cases

The theorem's hypotheses draw the exact boundary between collapse that must be singular and collapse that need not be; relaxing each entry changes the verdict.

Case	Result	Meaning
Spherical collapse (Oppenheimer–Snyder)	$\theta_{\text{out}}<0 \iff r<2GM$	Inside the Schwarzschild radius every constant- $r$ sphere is trapped; Penrose recovers the known central singularity but now with no symmetry assumed.
Weak field, no trapped surface	$\theta_{\text{out}}>0$	A diffuse cloud that never forms a trapped surface is unconstrained — it may bounce or virialize. The theorem only fires once $T$ exists.
Closed (compact) universe	$\Sigma$ compact	The final projection permits a compact image into a compact $\Sigma$ , so no contradiction; Penrose's theorem does not apply and one needs the Hawking–Penrose theorem.
Null energy condition violated	$R_{ab}k^ak^b<0$	Exotic or quantum negative energy can defocus the congruence, dissolve the caustic, and evade the theorem — the loophole every bounce scenario exploits.
Incompleteness vs curvature	geodesic ends, curvature bounded	The theorem yields incompleteness only; a quasi-regular (conical) singularity has finite curvature yet still terminates geodesics.

Common trapThe theorem proves infinite curvature

It does not. The conclusion is null geodesic incompleteness — at least one inextendible null geodesic of finite affine length — and nothing more. Whether the curvature actually diverges (a scalar curvature singularity, as in Schwarzschild) or stays bounded (a quasi-regular or conical singularity) is a logically separate question the theorem never touches; the two happen to coincide for Schwarzschild but need not in general. Two further precision traps: the non-compact Cauchy surface is essential — closed universes evade this theorem entirely, which is why the Hawking–Penrose theorem with its stronger generic and energy conditions was later needed for cosmology — and the operative hypothesis is the null energy condition, the weakest of the classical conditions, not the dominant or strong condition, which is exactly why the result survives into so many quantum-corrected settings.

Worked examples

1 · The affine leash — how soon must light focus? (fully worked with numbers)

A supermassive black hole with the mass of Sagittarius A, $M=4.3\times10^{6}\,M_\odot$ , has just formed an apparent horizon. Take the ingoing null congruence on that horizon, with expansion $\theta_-=-2/r$ . Using the focusing bound, find the maximum affine distance to the caustic, and compare it with the horizon radius.*

Step 1. Fix the scale. In geometrized units $GM_\odot/c^2=1.477$ km, so the horizon radius is $r_s=2GM/c^2=2(1.477\ \text{km})(4.3\times10^{6})=\mathbf{1.27\times10^{7}\ km}$ , about $0.085$ AU — roughly a fifth of Mercury's orbit.

Step 2. Evaluate the initial expansion. On the horizon $\theta_0=\theta_-=-2/r_s=-2/(1.27\times10^{7}\ \text{km})=-1.58\times10^{-7}\ \text{km}^{-1}$ , a tiny convergence rate because the horizon is enormous.

Step 3. Apply the focusing bound $\Delta\lambda\le 2/|\theta_0|=2/(2/r_s)=r_s$ . The caustic distance is at most one horizon radius, a clean, mass-independent statement: $\Delta\lambda\le r_s=\mathbf{1.27\times10^{7}\ km}$ .

Step 4. Interpret. Even for a hole millions of times the mass of the Sun, an ingoing light ray that reaches the horizon is guaranteed to focus to a caustic within one horizon radius of affine length — the local seed that, run through the topological argument, becomes the inevitability of the interior singularity. The caustic bound equals $r_s\approx 1.27\times10^{7}$ km for Sagittarius A*, and $\Delta\lambda\le r_s$ for any mass.

2 · When does collapse pull the trigger? (fully worked with numbers)

A $10\,M_\odot$ stellar core is collapsing. Using the Schwarzschild outgoing expansion $\theta_{\text{out}}=\tfrac{1}{r}(1-2GM/r)$ , decide which constant- $r$ 2-spheres are trapped, and locate the marginally trapped surface that first triggers the theorem.

Step 1. Compute the Schwarzschild radius: $r_s=2GM/c^2=2(1.477\ \text{km})(10)=\mathbf{29.5\ km}$ . This is where the outgoing expansion changes sign.

Step 2. Test a sphere at $r=40$ km (outside): $\theta_{\text{out}}=\tfrac{1}{40}(1-29.5/40)=\tfrac{1}{40}(0.263)=+6.6\times10^{-3}\ \text{km}^{-1}>0$ . The outgoing front still expands, so the sphere is untrapped — the theorem has no purchase yet.

Step 3. Test a sphere at $r=20$ km (inside): $\theta_{\text{out}}=\tfrac{1}{20}(1-29.5/20)=\tfrac{1}{20}(-0.475)=-2.4\times10^{-2}\ \text{km}^{-1}<0$ , while $\theta_{\text{in}}=-2/20=-0.10\ \text{km}^{-1}<0$ . Both negative — the sphere is trapped.

Step 4. Locate the trigger. The marginally trapped surface $\theta_{\text{out}}=0$ sits exactly at $r=r_s=29.5$ km; once the collapsing core shrinks within this radius a closed trapped surface exists, and Penrose's theorem immediately guarantees a singularity — with no assumption that the collapse stay spherical. Untrapped at 40 km, trapped at 20 km; the theorem fires the instant the core crosses $r_s=29.5$ km.

Physics of the portrait

The Penrose singularity theorem is the invention of one mathematician who saw that the deadlock over gravitational collapse was not a problem of solving harder differential equations but of asking a global, topological question.

Origins · 1 minRoger Penrose and how this was discovered

Roger Penrose

b. 1931 · global methods and the singularity theorems

Roger Penrose came to relativity from pure mathematics — projective geometry, tessellations, and the impossible figures he devised with his father and shared with M. C. Escher — and it was that geometer's instinct for the global shape of things that broke the impasse over gravitational collapse. Through the early 1960s the field was stuck: Oppenheimer and Snyder had shown in 1939 that a spherical dust star collapses to a singularity, but the Landau school argued this was a symmetry artefact certain to vanish under realistic perturbations. Rather than integrate lumpy collapse equations, Penrose reframed the question. In a 1965 Physical Review Letters paper of barely three pages he introduced the trapped surface and imported global tools — causal structure, achronal boundaries, Cauchy surfaces, invariance of domain — to prove that a trapped surface, the null energy condition, and a non-compact Cauchy surface force geodesic incompleteness, with symmetry playing no role at all. The proof inaugurated the era of global techniques in general relativity; developed with Stephen Hawking over the following decade, the same machinery produced the cosmological singularity theorems and much of black-hole thermodynamics. Penrose went on to the cosmic censorship conjectures, the Penrose inequality, spin networks, and twistor theory, and shared the 2020 Nobel Prize in Physics for showing that black-hole formation is a robust prediction of general relativity.

History

1939

— Oppenheimer and Snyder model spherical dust collapse ending in a singularity

1955

— Raychaudhuri derives the focusing equation for a congruence of geodesics

1965

— Penrose introduces trapped surfaces and proves the first singularity theorem

1970

— Hawking and Penrose extend the theorem to cosmological (big-bang) singularities

2020

— Penrose receives the Nobel Prize for the black-hole singularity prediction

Going further

Texts. Penrose, "Gravitational Collapse and Space-Time Singularities", Phys. Rev. Lett. 14, 57 (1965) · Hawking & Ellis, The Large Scale Structure of Space-Time (1973), ch. 8 · Wald, General Relativity (1984), §9.5 · Misner, Thorne & Wheeler, Gravitation (1973), ch. 34 · Carroll, Spacetime and Geometry (2004), ch. 9 · Senovilla & Garfinkle, "The 1965 Penrose singularity theorem", Class. Quantum Grav. 32, 124008 (2015). The focusing bound $\Delta\lambda\le 2/|\theta_0|$ follows by comparison with the exact solution $\theta(\lambda)=\theta_0/(1+\tfrac12\theta_0\lambda)$ of the equality case $\dot\theta=-\tfrac12\theta^2$ ; shear and the Ricci focusing term only shorten the distance. The worked numbers use $GM_\odot/c^2=1.477$ km, giving $r_s=2.95$ km per solar mass, $29.5$ km for $10\,M_\odot$ , and $1.27\times10^{7}$ km for Sagittarius A* ( $4.3\times10^{6}\,M_\odot$ ); the Schwarzschild expansion $\theta_{\text{out}}=\tfrac1r(1-2GM/r)$ changes sign at $r=2GM$ . The Penrose inequality $M\ge\sqrt{A/16\pi}$ is quoted in $G=c=1$ units.

Builds on: Global Hyperbolicity · Cauchy Surfaces · Null Geodesics and Light Cones · Chronological and Causal Futures and Pasts · Achronal Sets and Edges · Causal Structure of Spacetime. Leads to: Hawking–Penrose Singularity Theorem · Cosmic Censorship Conjectures · Penrose–Carter Conformal Diagrams · Black Hole Event Horizons.

Ask the tutor: Ask me to walk through the focusing integral step by step, or to show exactly where the compact lid $\dot J^+(T)$ collides with a non-compact Cauchy surface in Penrose's contradiction.

Does the theorem prove that the curvature becomes infinite at the singularity?

No — this is the single most common misreading. Penrose's theorem proves null geodesic incompleteness: at least one inextendible null geodesic of finite affine length. It says nothing about whether the curvature actually diverges. For the Schwarzschild singularity the curvature does blow up, but that is a separate fact established by direct calculation, not by the theorem. There exist quasi-regular (conical) singularities where geodesics terminate while every curvature scalar stays finite; the theorem would flag those as singular too. Incompleteness is the coordinate-independent definition of a singularity precisely because "the point where curvature is infinite" is not well defined once that point has been excised from the manifold.

What exactly is a trapped surface, and why is it the right trigger?

A closed trapped surface is a compact, spacelike 2-surface — think of a distorted sphere — on which both families of orthogonal future-directed light rays are converging: both null expansions are negative everywhere on it. In flat space the outgoing rays from any sphere fan out, so a trapped surface is a local signal that gravity has become strong enough to turn even outgoing light inward. It is the right trigger because it is entirely local and requires no symmetry: you can test the sign of the two expansions from data in a neighbourhood of the surface, on a single slice of time, without knowing anything about the global future. That is what let Penrose sidestep the intractable problem of solving asymmetric collapse in detail.

Why is the non-compactness of the Cauchy surface essential?

It is what makes the final contradiction bite. Focusing forces the causal-future boundary $\dot J^+(T)$ to be compact, and projecting it onto the Cauchy surface $\Sigma$ gives an image that is compact, open, and closed. In a connected space, a non-empty subset that is both open and closed must be the entire space — so the image would have to be all of $\Sigma$ . If $\Sigma$ is non-compact this is impossible, because a compact set cannot equal a non-compact one, and the contradiction follows. If instead $\Sigma$ is compact, as in a closed universe, a compact image filling all of $\Sigma$ is perfectly consistent, no contradiction arises, and the theorem simply does not apply. That gap is exactly why the Hawking–Penrose theorem, with stronger energy and generic conditions, was later needed for cosmological singularities.

How can the theorem apply to realistic, lumpy stars when it assumes so little?

That economy is the whole point. The theorem assumes no symmetry, no stationarity, and no particular equation of state — only the null energy condition, which every known classical form of matter satisfies, plus one trapped surface, plus an open universe. A real collapsing star is turbulent, rotating, and magnetized, and no one can solve its evolution in closed form; but none of that matters. As soon as the collapse becomes compact enough to enclose a single trapped surface, a singularity is guaranteed. This is exactly what demolished the Landau-school hope that asymmetry or rotation would save collapsing matter — the conclusion is insensitive to precisely the complications they were counting on.

Where is the loophole for avoiding singularities?

Any one of the three hypotheses can be denied, but only one of them is physical rather than geometric, and that is where the work goes: violate the null energy condition. The other two escapes — never forming a trapped surface, or living in a closed universe with a compact Cauchy surface — are not open to a collapsing star in an open universe, so a theory that keeps those features and still removes the singularity must arrange $R_{ab}k^ak^b<0$ somewhere: a negative Ricci focusing term, which whenever Einstein's equations hold means a negative null energy density, defocusing the light rays before they reach a caustic. Classical matter cannot do this, but quantum fields can, in amounts limited by quantum energy inequalities, and effective descriptions of loop quantum cosmology, string-gas bounces, and certain modified-gravity models all exploit exactly such a controlled violation. The theorem thus does more than predict singularities: it converts the open problem of resolving them into the sharp technical question of how much, and over how large a region, the null energy condition can be broken.

How does this theorem relate to the Hawking–Penrose theorem and cosmic censorship?

They form a ladder. Penrose's 1965 theorem is about the future of a trapped surface in an open universe and uses only the null energy condition — it is the black-hole case. The Hawking–Penrose theorem (1970) trades the non-compact Cauchy surface for stronger generic and strong-energy conditions and so also covers the past singularity of cosmology, the big bang. Both prove geodesic incompleteness. Cosmic censorship is the separate, still-unproven conjecture that sits on top: having shown singularities are generic, Penrose proposed that they are generically hidden behind event horizons (weak censorship) and that they do not destroy predictability for outside observers (strong censorship). The singularity theorems are proven; censorship is the outstanding open problem they raised.