Hawking Singularity Theorem

The Hawking singularity theorem is the moment cosmology stopped needing exact solutions to say something certain about the beginning of the Universe. Before it, the Big Bang was known only from the perfectly symmetric Friedmann–Lemaître–Robertson–Walker models, and the working assumption was that a realistic, lumpy universe would sidestep the singularity — the collapsing matter would miss itself and re-expand. Penrose had just shattered that prejudice for gravitational collapse (1965) using trapped surfaces and null geodesics. Hawking's contribution was to run the identical logic in the timelike direction and on cosmological scales: if the Universe is expanding everywhere on one spacelike slice and gravity is attractive, then reversing the comoving observers into the past forces them to converge — and convergence, made rigorous, means a geodesic that simply stops. The argument is a proof by contradiction built from three ingredients and nothing else: an energy condition (the strong energy condition, which through Einstein's equations makes gravity focus), a global-causality condition (global hyperbolicity, a Cauchy surface), and one initial datum (the expansion is bounded below by a positive number on that surface). No symmetry, no exact metric, no assumption about what matter is beyond one inequality. The engine underneath is the Raychaudhuri equation, the covariant statement that freely falling worldlines focus. What it delivers is precise and surprisingly modest: at least one timelike geodesic is incomplete — finite proper length to the past, inextendible. That, and not "infinite curvature," is the technical meaning of a singularity here.

Key takeaways

If a spacetime is globally hyperbolic, satisfies the strong energy condition, and is expanding everywhere on a Cauchy surface with θ ≥ b > 0, then it is past timelike geodesically incomplete — comoving observers cannot be extended past a finite proper time into the past.
The engine is the Raychaudhuri equation: positive expansion plus R_{μν}u^μu^ν ≥ 0 forces the expansion of the orthogonal geodesic congruence to have diverged from +∞ within a bounded past proper time τ ≤ 3/b, producing a focal (conjugate) point.
A focal point is fatal because a maximal (longest) timelike geodesic can contain no interior conjugate point; global hyperbolicity guarantees such a maximizer exists, so completeness and focusing cannot both hold.
The conclusion is geodesic incompleteness, NOT a curvature singularity: the theorem is silent about whether curvature invariants blow up or what the boundary is made of.
The strong energy condition R_{μν}u^μu^ν ≥ 0 (equivalently ρ + 3p ≥ 0 and ρ + p ≥ 0 for a perfect fluid) is obeyed by dust and radiation but violated by a positive cosmological constant and by inflation — exactly how modern cosmology evades the theorem.
Key idea: expansion on a Cauchy surface plus attractive gravity forces the orthogonal timelike geodesic congruence to have had infinite expansion a finite proper time in the past; that focal point cannot sit on a maximal geodesic in a complete globally hyperbolic spacetime — contradiction — so the spacetime is past timelike geodesically incomplete.

The core insight

Expansion (θ₀ > 0 on a Cauchy surface) plus attractive gravity (the strong energy condition, which gives R_{μν}u^μu^ν ≥ 0) forces the orthogonal timelike geodesic congruence to have had infinite expansion a finite proper time in the past. That focal point cannot lie on a maximal geodesic in a complete globally hyperbolic spacetime — a contradiction — so the spacetime is past timelike geodesically incomplete.

Where this lives today

The theorem's modern life is a story of loopholes and their closure. Inflation drives the early universe with a scalar field of equation of state w ≈ −1, violating the strong energy condition — so Hawking's theorem does not apply during inflation, and eternal-inflation models exploit exactly this. The sharpest response is the kinematic Borde–Guth–Vilenkin theorem (2003), which discards every energy condition and assumes only that the average Hubble expansion along a past-directed geodesic is positive; it still forces past-incompleteness, so even inflating spacetimes have a past boundary. Meanwhile the theorem is the north star for singularity-resolution programs: loop quantum cosmology replaces the Big Bang with a bounce by engineering an effective SEC violation at Planck density, and ekpyrotic and bouncing cosmologies do the same with negative-pressure components — all attacking the one inequality Hawking needs. And the subject was crowned in 2020, when Roger Penrose received the Nobel Prize for the singularity theorems — the collapse counterpart of the cosmological result proved here.

The physical picture

Picture a swarm of free-falling observers — a congruence of timelike geodesics — launched perpendicular to a spacelike snapshot Σ of the Universe. "Expanding everywhere" means their worldlines are flying apart: the volume of any small comoving cell grows, so the expansion scalar θ = ∇_μ u^μ is positive across all of Σ. Gravity, attractive for ordinary matter, decelerates that expansion. Run the movie backwards: the expansion was larger in the past, attractive gravity makes it larger still further back, and the Raychaudhuri equation says it diverges — the worldlines crowd together and cross at a focal point within proper time no greater than 3/b. That crossing is where the geodesics stop being extendible — the past has a floor. The Raychaudhuri equation is just Newtonian tidal focusing written covariantly. The −θ²/3 term says a spreading cloud self-decelerates kinematically; the shear −2σ² always focuses; and the −R_{μν}u^μu^ν term is the gravitational tidal pull, which the strong energy condition guarantees never helps the cloud re-expand. With every term non-positive, θ can only fall — and once falling, it falls to −∞ in finite proper time, as the first figure shows.

Figure

A hypersurface-orthogonal timelike geodesic congruence launched from a Cauchy surface Σ with expansion θ₀ ≥ b > 0 focuses into the past: attractive gravity drives the expansion to θ → −∞ at a focal point reached within proper time τ ≤ 3/b, the past floor Hawking's theorem forces

Why is a focal point fatal? A timelike geodesic that maximizes proper time between a surface and a point is the Lorentzian analogue of a straight line in flat space. But once it passes a point conjugate to Σ — a focal point — a nearby curve of strictly greater proper time exists, so it is no longer the maximizer. Global hyperbolicity guarantees such a maximizer exists to any past point, carrying no interior focal point. Focusing forces one within proper time 3/b; the only escape is that the geodesic never got that far — too short, i.e. incomplete. The two facts collide, and the collision is the theorem. The second figure shows the payoff: whether the past has a floor is decided entirely by the strong energy condition.

Figure

Two expanding cosmologies compared. With the strong energy condition the scale factor a ∝ t^(2/3) reaches a = 0 at a finite past proper time — a Big Bang, the incomplete geodesic Hawking guarantees. A de Sitter universe a = e^(Ht) driven by Λ > 0 violates the condition, stays a > 0 for all past time, and extends to a geodesically complete de Sitter spacetime

The core equations

Everything in the theorem flows from a single evolution equation for the expansion of a geodesic congruence. The Raychaudhuri equation tracks how the fractional volume-growth rate θ of a bundle of freely falling worldlines changes along them, splitting the change into a kinematic self-focusing piece, a shear piece, a rotation piece, and a gravitational tidal piece.

\frac{\mathrm{d}\theta}{\mathrm{d}\tau} = -\frac{1}{3}\theta^2 - \sigma_{\mu\nu}\sigma^{\mu\nu} + \omega_{\mu\nu}\omega^{\mu\nu} - R_{\mu\nu}u^\mu u^\nu \qquad (1)

where: θ — expansion scalar, θ = ∇_μ u^μ, the fractional rate of change of the congruence's cross-sectional volume; τ — proper time along the geodesics; u^μ — unit timelike tangent, u^μ u_μ = −1; σ_{μν} — shear (symmetric, trace-free, spatial), with σ² = σ_{μν}σ^{μν} ≥ 0; ω_{μν} — vorticity, which vanishes for a hypersurface-orthogonal congruence; R_{μν}u^μu^ν — the tidal focusing term fed by the Einstein equations. For a congruence orthogonal to a spacelike surface the vorticity vanishes identically (Frobenius), the shear term is non-positive, and the strong energy condition pins the Ricci term non-negative — so every term works the same way and the equation collapses to a one-line inequality.

\frac{\mathrm{d}\theta}{\mathrm{d}\tau} \le -\frac{1}{3}\theta^2 \qquad (2)

where: σ² ≥ 0 makes −σ² ≤ 0; ω_{μν} = 0 by hypersurface-orthogonality; R_{μν}u^μu^ν ≥ 0 is the strong energy condition. This Riccati inequality has one inevitable consequence: the solution runs to −∞ in finite time. Substituting θ⁻¹ turns (2) into a linear bound and exposes the blow-up — the focal-point clock at the heart of the theorem.

\theta(\tau) \le \frac{1}{\theta_0^{-1} + \tfrac13\tau} \quad\Longrightarrow\quad \theta \to -\infty \ \text{within}\ \tau \le \frac{3}{|\theta_0|} \qquad (3)

where: θ₀ — the expansion at τ = 0 on the Cauchy surface; 3/|θ₀| — the upper bound on the proper time to a focal point (to the future if θ₀ < 0, to the past if θ₀ > 0), the entire quantitative content of the focusing lemma. The Ricci term is made non-negative by an assumption about matter, imported through the trace-reversed Einstein equations — the strong energy condition, or timelike convergence condition.

R_{\mu\nu}u^\mu u^\nu \ge 0 \iff \Big(T_{\mu\nu} - \tfrac12 T g_{\mu\nu}\Big)u^\mu u^\nu \ge 0 \qquad (4)

where: T_{μν} — stress-energy tensor; T = g^{μν}T_{μν} its trace; u^μ — any unit timelike vector, the condition holding for all of them. The left equality uses R_{μν} = 8π(T_{μν} − ½T g_{μν}) in units G = c = 1. For an isotropic perfect fluid the condition unpacks into two inequalities on rest-frame density and pressure — the form one checks against a cosmological model.

\rho + 3p \ge 0 \quad\text{and}\quad \rho + p \ge 0 \qquad(\text{perfect-fluid SEC}) \qquad (5)

where: ρ — energy density in the fluid rest frame; p — isotropic pressure. Dust (p = 0) and radiation (p = ρ/3) satisfy both; a positive cosmological constant (p = −ρ) fails the first. For the comoving observers of an FLRW universe — an irrotational, shear-free geodesic congruence — the abstract θ is simply three times the Hubble rate.

\theta_{\rm FLRW} = 3\frac{\dot a}{a} = 3H \qquad (6)

where: a(t) — scale factor; H = ȧ/a — Hubble parameter. An expanding universe has H > 0, hence θ > 0 — the expansion hypothesis is just "the Universe is expanding." The theorem needs that expansion bounded away from zero everywhere on one slice — a single, coordinate-free hypothesis.

\mathcal{H}(\Sigma):\ \theta\big|_\Sigma \ge b > 0 \ \text{everywhere} \qquad (7)

where: Σ — a smooth spacelike Cauchy surface; b — a positive lower bound on the expansion of the past-directed orthogonal congruence. It is this b that sets the proper-time depth 3/b of the past. Finally, the contradiction is phrased through proper-time length. In a globally hyperbolic spacetime the longest curve from Σ to a point is a geodesic — and that maximizer refuses to hold a focal point in its interior.

L[\gamma] = \int \sqrt{-g_{\mu\nu}\dot x^\mu \dot x^\nu}\,\mathrm{d}\lambda,\qquad L\ \text{maximal} \Rightarrow \text{no interior point conjugate to } \Sigma \qquad (8)

where: L[γ] — Lorentzian length (proper time) of a causal curve γ; a conjugate (focal) point is where neighbouring orthogonal geodesics from Σ reconverge, beyond which γ ceases to maximize L. Equations (3) and (8) pull in opposite directions — something must break.

Derivation ADeriving the Raychaudhuri focusing equation

The goal is to show that the evolution equation (1) is not an assumption but a geometric identity, following from the Ricci identity and the geodesic equation alone, with no field equations until the last step where R_{μν} is read physically. We build the congruence's deformation tensor, decompose it, propagate it, and take the trace — where curvature enters through the failure of derivatives to commute.

Step 1. Define the deformation tensor B_{μν} = ∇_ν u_μ, measuring how the tangent to one worldline changes as you step to a neighbour. For a geodesic congruence it is purely spatial (B_{μν}u^ν = 0 = u^μ B_{μν}), so it decomposes as B_{μν} = ⅓θ h_{μν} + σ_{μν} + ω_{μν}, with h_{μν} = g_{μν} + u_μ u_ν the spatial metric and θ = h^{μν}B_{μν} the trace. Why: this is the covariant version of splitting a fluid element's deformation into isotropic expansion, volume-preserving shear, and rigid rotation.

Step 2. Propagate B along the congruence. Using the geodesic equation u^ρ∇_ρ u^μ = 0 and the Ricci identity ∇_ρ∇_ν u_μ − ∇_ν∇_ρ u_μ = R_{σμρν}u^σ, one finds u^ρ∇_ρ B_{μν} = −B_{μρ}B^ρ{}_ν − R_{ρμνσ}u^ρ u^σ. Why: the curvature term is the geodesic-deviation content — the only place the geometry of spacetime, as opposed to the kinematics of the flow, enters.

Step 3. Take the trace on μν. The left side becomes dθ/dτ; the quadratic term gives −B_{μν}B^{νμ}; the curvature term contracts to −R_{μν}u^μu^ν using R_{ρμν}{}^ρ = R_{μν}. Substituting B_{μν}B^{νμ} = ⅓θ² + σ² − ω² (cross terms vanish by orthogonality) yields dθ/dτ = −⅓θ² − σ² + ω² − R_{μν}u^μu^ν, which is equation (1). Why: the trace isolates the volume behaviour — whether the bundle's cross-section collapses to zero, which is what makes a focal point.

Go deeper · Derivation B · 1 minFocusing to a conjugate point in finite proper time

Derivation BFocusing to a conjugate point in finite proper time

The goal is to prove the quantitative heart of the theorem: under the strong energy condition, a hypersurface-orthogonal timelike geodesic congruence contracting at rate θ₀ < 0 reaches θ → −∞ — a focal point conjugate to the surface — within proper time no larger than 3/|θ₀|, and by time-reversal an expanding congruence must have emerged from such a focal point in the past. This clock puts a floor under the past; we start from equation (1) for an irrotational congruence and track a reciprocal.

Step 1. Drop the non-negative terms. For a hypersurface-orthogonal congruence ω_{μν} = 0; the shear obeys σ² ≥ 0; and the strong energy condition gives R_{μν}u^μu^ν ≥ 0. Hence dθ/dτ = −⅓θ² − σ² − R_{μν}u^μu^ν ≤ −⅓θ². Why: we throw away everything that could only make focusing stronger, so the bound is conservative — the real congruence focuses at least this fast.

Step 2. Change to the reciprocal. While θ < 0, d(θ⁻¹)/dτ = −θ⁻²(dθ/dτ) ≥ −θ⁻²(−⅓θ²) = ⅓. The reciprocal grows at least linearly, slope one-third. Why: the reciprocal is exactly what linearizes the −⅓θ² nonlinearity — the standard trick for a Riccati inequality.

Step 3. Integrate. From d(θ⁻¹)/dτ ≥ ⅓ we get 1/θ(τ) ≥ 1/θ₀ + τ/3. With θ₀ < 0 the right-hand side starts negative and increases toward zero, crossing zero at τ⋆ = −3/θ₀ = 3/|θ₀|. Why: the reciprocal reaching zero from below is precisely θ plunging to −∞ — a line in θ⁻¹ is a blow-up in θ.

Step 4. Conclude the blow-up. As τ → τ⋆⁻, 1/θ → 0⁻ and θ → −∞. Geometrically the neighbouring geodesics have crossed: a focal point conjugate to the surface has formed at or before proper time 3/|θ₀| — provided the geodesic extends that far. *Why: the equation says only that IF the geodesic reaches τ⋆ a focal point is unavoidable — the escape hatch is incompleteness, which the theorem then corners.*

Step 5. Orient it cosmologically. An expanding universe has θ₀ ≥ b > 0 to the future, so the past-directed congruence is contracting and (3) applies with |θ₀| ≥ b: a focal point lies in the past within proper time τ ≤ 3/b. Why: "expanding to the future" and "focusing to the past" are the same statement, and 3/b is the finite past depth the next derivation turns into a contradiction.

Go deeper · Derivation C · 2 minWhy a focal point contradicts completeness — the theorem's logic

Derivation CWhy a focal point contradicts completeness — the theorem's logic

The goal is to combine the focusing lemma with the spacetime's global structure to force past timelike geodesic incompleteness. We argue by contradiction: assume the spacetime is globally hyperbolic with Cauchy surface Σ, satisfies the strong energy condition, has θ ≥ b > 0 for the past-directed normal on Σ, AND — the hypothesis we will refute — is past timelike geodesically complete, so every past-directed geodesic extends to arbitrary length. From these we derive an impossibility.

Step 1. A maximal geodesic exists. Global hyperbolicity makes the space of causal curves between Σ and any past point p compact, so the length functional (8) attains its maximum; the maximizer is a timelike geodesic meeting Σ orthogonally (Avez–Seifert). Call its length d(Σ, p). Why: global hyperbolicity does one job — it upgrades "there are long curves" into "there is a longest curve," the only thing the no-conjugate-point rule can bite on.

Step 2. A maximizer has no interior focal point. If a geodesic from Σ to p had a point conjugate to Σ strictly before p, one could deform it into a strictly longer neighbouring curve, contradicting maximality. So it carries no interior focal point, and by the focusing lemma its length obeys d(Σ, p) ≤ 3/b. Why: this is the linchpin — focusing (local) and maximality (global) are incompatible beyond proper time 3/b, forcing the maximizer to be short.

Step 3. The past is shallow. The bound holds for every p to the past of Σ, so d(Σ, p) ≤ 3/b for all p ∈ J⁻(Σ): the entire past of the Cauchy surface lies within proper time 3/b of it. Why: a point-by-point bound becomes a uniform statement about the whole causal past — a very strong constraint.

Step 4. Collide with completeness. Past completeness demands a past-directed geodesic from Σ of proper length exceeding 3/b — extend any orthogonal geodesic past that time and its endpoint p satisfies d(Σ, p) > 3/b, contradicting Step 3. The two cannot both hold, so completeness is false: at least one past-directed timelike geodesic orthogonal to Σ is inextendible past proper time 3/b. Why: this is the contradiction the edifice was built for — completeness is exactly what would let the geodesic reach its forbidden focal point, so it is what must give.

Step 5. State the theorem. Hawking (1967): if (i) R_{μν}u^μu^ν ≥ 0 for all timelike u, (ii) the spacetime is globally hyperbolic with a Cauchy surface Σ, and (iii) θ ≥ b > 0 everywhere on Σ for the past-directed normal, then every past-directed timelike geodesic from Σ has length at most 3/b — the spacetime is past timelike geodesically incomplete. Why: named precisely, the hypotheses are falsifiable — drop any one and the conclusion can fail, as inflation and de Sitter show.

Limiting cases

Each hypothesis can hold or fail; the theorem's reach is best seen at each boundary — where it bites, and where it says nothing.

Case	Result	Meaning
Strong energy condition holds	R_{μν}u^μu^ν = 8π(T_{μν} − ½T g_{μν})u^μu^ν ≥ 0	Holds for dust (ρ ≥ 0), radiation, and any fluid with ρ + 3p ≥ 0 and ρ + p ≥ 0. Focusing is guaranteed; the theorem applies and predicts a past singularity.
SEC violated by Λ > 0	ρ_Λ + 3p_Λ = ρ_Λ − 3ρ_Λ = −2ρ_Λ < 0	A positive cosmological constant has p = −ρ. De Sitter is complete and past-eternal — the hypothesis is unmet, so this is no contradiction, not a counterexample.
Inflation	w = p/ρ ≈ −1 ⟹ ρ + 3p < 0	The inflaton violates the SEC, so the theorem cannot force a singularity during inflation — the loophole inflation and bouncing models exploit.
Global hyperbolicity dropped	Hawking–Penrose (1970): strong causality + generic + energy + a trapped set	The unified theorem trades global hyperbolicity for strong causality plus a generic condition, covering cosmology and collapse at once.

Common trapThe theorem proves a curvature singularity at a single first moment of time

Three misreadings, all common. First, the conclusion is geodesic incompleteness, not a curvature singularity: at least one timelike geodesic has finite proper length and cannot be extended. Whether curvature invariants diverge there is a separate question — in FLRW they do (R ∼ t⁻²), but the theorem is silent, and the boundary could be a milder inextendibility. Second, it needs the STRONG energy condition R_{μν}u^μu^ν ≥ 0, not the weak or null condition Penrose uses for collapse; a positive cosmological constant and inflation violate it, so there the theorem simply makes no prediction — quoting it against an inflationary model is a category error. Third, it does not assert a synchronized cosmic "first instant": it guarantees one incomplete geodesic of bounded past length, not a single spacelike initial singularity on which all worldlines end at once.

Worked examples

1 · The Hawking bound on the age of the Universe (fully worked)

Our Universe expands with Hubble constant H₀ ≈ 70 km s⁻¹ Mpc⁻¹, and in the matter- and radiation-dominated eras it obeyed the strong energy condition. Treating "now" as the Cauchy surface, use the focusing bound to cap the proper time back to the singularity, and compare with the measured age.

Step 1. Identify the expansion bound b. The comoving congruence has θ = 3H (equation 6), so today's past-directed expansion satisfies |θ₀| = 3H₀; take b = 3H₀. b = 3H₀.

Step 2. Apply the focusing bound. The past depth is τ ≤ 3/b = 3/(3H₀) = 1/H₀ — the bound collapses to exactly the Hubble time. τ ≤ 1/H₀.

Step 3. Put in the number. Convert H₀ = 70 km s⁻¹ Mpc⁻¹ = (7.0×10⁴ m s⁻¹)/(3.086×10²² m) = 2.27×10⁻¹⁸ s⁻¹. Then 1/H₀ = 4.41×10¹⁷ s. Dividing by 3.156×10⁷ s per year gives 1/H₀ ≈ 1.40×10¹⁰ yr ≈ 14 Gyr.

Step 4. Compare and interpret. The measured age is about 13.8 Gyr, comfortably below the 14 Gyr ceiling — as it must be, since the theorem gives an upper bound, saturated only for a coasting (empty) universe; adding attractive matter pulls the singularity closer. The Hawking bound τ ≤ 1/H₀ ≈ 14 Gyr is respected by the observed 13.8 Gyr age.

2 · FLRW as an instance, and the de Sitter escape

Show that a spatially flat FLRW universe of dust or radiation satisfies every hypothesis and realizes the incomplete geodesic, then show why de Sitter slips through.

Step 1. Metric and congruence. Take ds² = −dt² + a(t)²δ_{ij}dx^i dx^j with comoving congruence u^μ = (∂_t)^μ. It is geodesic, shear-free and irrotational, so only the expansion survives: θ = 3ȧ/a = 3H > 0 on any constant-t slice, a Cauchy surface — the expansion hypothesis holds.

Step 2. Check the SEC. The relevant curvature is R_{μν}u^μu^ν = R_{tt} = −3ä/a = 4π(ρ + 3p). For dust (p = 0) and radiation (p = ρ/3), ρ + 3p > 0, so R_{μν}u^μu^ν ≥ 0 — the SEC holds and the universe decelerates. SEC satisfied; gravity focuses.

Step 3. Solve Friedmann. The equations give a(t) ∝ t^(2/[3(1+w)]): a ∝ t^(1/2) for radiation (w = 1/3) and a ∝ t^(2/3) for dust (w = 0). Both vanish as t → 0⁺, and the comoving proper time since a = 0 is Δτ = ∫₀^t dt′ = t < ∞. The geodesic reaches a = 0 at finite proper time and cannot be continued (curvature diverges, R ∼ t⁻²): exactly the incomplete past geodesic the theorem guarantees. Finite past proper time to an inextendible boundary.

Step 4. The de Sitter escape. Replace the matter with a positive cosmological constant, p = −ρ, so ρ + 3p = −2ρ < 0 and the SEC fails. The solution a(t) = e^(Ht) never reaches zero: as t → −∞ it decays but stays positive, and every comoving worldline has infinite past proper time. This exponentially expanding chart is only the flat patch — half of de Sitter spacetime, extendible across a past horizon — but the de Sitter spacetime it extends to is geodesically complete. No hypothesis, no conclusion. de Sitter evades Hawking by breaking the one inequality on matter.

Physics of the portrait

The theorem is inseparable from the researcher who, told the Big Bang was probably an artifact of symmetry, set out to prove it was not — with tools Penrose had just built for the opposite end of a star's life.

Origins · 1 minStephen Hawking and how this was discovered

Stephen Hawking

1942–2018 · the cosmological singularity theorem and global methods in general relativity

Stephen Hawking entered gravitational physics at Cambridge just as Roger Penrose (1965) had used trapped surfaces and null geodesics to prove that a collapsing star must form a singularity — breaking the long-held belief that generic collapse would miss itself and bounce. Hawking's decisive move, in his 1966 Adams Prize essay and 1967 Royal Society paper, was to run Penrose's logic in reverse: not the future of a collapsing star but the past of an expanding universe, and not null geodesics but the timelike comoving congruence. The engine was Amal Kumar Raychaudhuri's 1955 focusing identity, and the hypothesis that makes focusing inevitable is R_{μν}u^μu^ν ≥ 0 — through Einstein's equations, the strong energy condition, still called the timelike convergence condition. From these he showed that a broad, symmetry-free class of expanding universes must be past timelike geodesically incomplete: the Big Bang is a theorem, not an accident of the FLRW model. In 1970 he and Penrose fused the timelike and null arguments into the definitive Hawking–Penrose theorem, weakening global hyperbolicity to strong causality plus a generic condition. The irony is that Hawking himself, in 1974, showed quantum fields near a black hole radiate — the same quantum vacuum that violates the very energy conditions the singularity theorems assume. This is the strong energy condition's double life: rigorous enough to prove the Universe had a beginning, fragile enough that quantum matter breaks it.

History

1955

— Raychaudhuri derives his focusing equation for geodesic congruences

1965

— Penrose proves the first singularity theorem from trapped surfaces

1967

— Hawking states the cosmological singularity theorem in Proc. R. Soc. A

1970

— Hawking and Penrose fuse the collapse and cosmological arguments

2003

— Borde, Guth and Vilenkin remove the energy condition entirely

Going further

Texts. Hawking & Ellis, The Large Scale Structure of Space-Time (1973), §8.2 (Theorem 4) and §4.4 (Raychaudhuri focusing) · Wald, General Relativity (1984), §9.5 (Theorem 9.5.1) with the focal-point lemmas of §9.3 · Misner, Thorne & Wheeler, Gravitation (1973), §34.6 (singularity theorems overview) · Carroll, Spacetime and Geometry (2004), §9.5 (energy conditions) and the singularity-theorem appendix · Hartle, Gravity (2003), Ch. 22 · Schutz, A First Course in General Relativity, 2nd ed., Ch. 10. Key relations used above: the focusing bound $3/|\theta_0|$ follows from integrating $\mathrm{d}(\theta^{-1})/\mathrm{d}\tau \ge \tfrac{1}{3}$ ; in FLRW, $\theta = 3H$ and $R_{tt} = 4\pi(\rho+3p) = -3\ddot{a}/a$ ; the SEC reads $\rho+3p\ge 0$ ; and $1/H_0 \approx 14\,\text{Gyr}$ for $H_0 = 70\,\text{km}\,\text{s}^{-1}\,\text{Mpc}^{-1}$ . Primary literature: Penrose, Phys. Rev. Lett. 14, 57 (1965); Hawking, Proc. R. Soc. A 300, 187 (1967); Hawking & Penrose, Proc. R. Soc. A 314, 529 (1970); Borde, Guth & Vilenkin, Phys. Rev. Lett. 90, 151301 (2003).

Prerequisites. Global Hyperbolicity · Cauchy Surfaces · Causal Structure of Spacetime · Domain of Dependence and Cauchy Horizons. Leads to. Penrose–Carter Conformal Diagrams · Causality Conditions Hierarchy · Asymptotic Flatness.

Ask the tutor: Ask me to walk through why the reciprocal 1/θ linearizes the Raychaudhuri inequality, or how inflation's negative pressure flips the sign of R_{μν}u^μu^ν and lets the past run forever.

What is the difference between the Hawking and Penrose singularity theorems?

They share one skeleton — focusing plus a causality condition — in complementary settings. Penrose (1965) treats gravitational collapse: he focuses null geodesics using a trapped surface and the null energy condition, concluding null incompleteness inside a forming black hole. Hawking (1966–67) treats cosmology: he focuses timelike geodesics using an expanding Cauchy surface and the strong energy condition, concluding past timelike incompleteness — the Big Bang. The 1970 Hawking–Penrose theorem merges them, replacing global hyperbolicity with strong causality plus a generic condition, so a trapped surface, a reconverging point, or a compact spacelike slice each triggers incompleteness — collapse and cosmology in one theorem.

Does the theorem prove the curvature was infinite at the Big Bang?

No, and this is the single most common overstatement. The theorem proves timelike geodesic incompleteness: some comoving observer's worldline has finite proper length into the past and cannot be extended. Whether the obstruction is a genuine curvature (scalar) singularity — where an invariant like R_{μνρσ}R^{μνρσ} diverges — is a separate question the theorem does not touch. In FLRW it does happen to be a curvature singularity (R ∼ t⁻²), but one can build incomplete spacetimes whose incompleteness is milder. The theorem's power is that it says nothing about the nature of the boundary to prove the boundary exists.

Why the strong energy condition specifically, rather than the weak or null one?

Because Hawking focuses timelike geodesics, and the Raychaudhuri term that must be sign-definite for a timelike congruence is R_{μν}u^μu^ν with u timelike. Requiring R_{μν}u^μu^ν ≥ 0 for every timelike u is, through the trace-reversed Einstein equations, exactly the strong energy condition (ρ + 3p ≥ 0 and ρ + p ≥ 0). Penrose, focusing null geodesics, needs only R_{μν}k^μk^ν ≥ 0 for null k — the weaker null energy condition (ρ + p ≥ 0). The energy condition is dictated entirely by which geodesics you focus, and that by whether your initial data is an expanding surface (timelike) or a trapped surface (null).

If ordinary matter obeys the SEC, how does inflation escape?

Inflation is not ordinary matter. It is driven by a scalar field held high on its potential, whose effective equation of state is w = p/ρ ≈ −1, so ρ + 3p ≈ −2ρ < 0 — the strong energy condition fails. Physically, the large negative pressure produces repulsive gravity, which accelerates the expansion (ä > 0) and lets the congruence avoid focusing. So during inflation Hawking's hypothesis is not met and the theorem makes no claim — not a flaw, but the theorem correctly declining to constrain a regime it was never about. Closing that gap took the energy-condition-free Borde–Guth–Vilenkin argument.

What does "global hyperbolicity" actually buy the proof?

It guarantees the existence of a longest curve. The focusing lemma is a local fact: geodesics that get too long must contain a focal point. But to reach a contradiction you must confront an actual maximal geodesic with that fact, and the mere existence of long curves is not enough — you need a genuine maximizer. Global hyperbolicity (a Cauchy surface, plus the compactness of causal-curve spaces it implies) is precisely the condition under which the length functional attains its supremum between a surface and a point (Avez–Seifert). Drop it and maximizers can fail to exist, which is why the 1970 Hawking–Penrose theorem replaced it with a subtler strong-causality-plus-generic package.

Is the bound τ ≤ 3/b saying the universe is at most 3/b old?

Almost — with care about what "old" means. The bound says every comoving observer has at most proper time 3/b of past history before running into an incomplete geodesic, where b is the lower bound on θ on the chosen slice. For the comoving congruence θ = 3H, so 3/b = 1/H: the past proper time is capped by the Hubble time. It is an upper bound, saturated only by an empty (coasting) universe; add attractive matter and the actual look-back time is shorter, which is why the real 13.8 Gyr sits under the 14 Gyr Hubble-time ceiling. The theorem does not compute the age; it caps it.