Geodesic Congruences and Expansion

A geodesic congruence turns a single free-fall trajectory into a whole flowing family of them — a "beam" of freely falling dust grains or light rays that fills a region of spacetime, one curve through every point. The interesting physics is never the motion of one geodesic (in free fall it feels nothing) but the relative motion of its neighbours: whether the beam swells, contracts, twists, or shears its cross section as it advances. That relative motion is precisely what a falling observer registers as tidal gravity, and it is the raw material from which all of global Lorentzian geometry is built. The bookkeeping is done by the transverse gradient of the tangent field, $B_{\mu\nu}=\nabla_\nu\xi_\mu$ , which measures the failure of nearby geodesics to stay parallel. Split it into its irreducible pieces and three physical quantities fall out: the expansion $\theta$ (fractional rate of change of the bundle's volume), the shear $\sigma_{\mu\nu}$ (volume-preserving distortion), and the rotation $\omega_{\mu\nu}$ (twist). The evolution of $\theta$ along the flow obeys the Raychaudhuri equation — the master focusing law in which $\theta^2$ , $\sigma^2$ , $\omega^2$ , and the Ricci term $R_{\mu\nu}\xi^\mu\xi^\nu$ each appear with a fixed sign. Shear and self-focusing drive $\theta$ down; only rotation resists; and matter enters solely through Ricci, via Einstein's equation. This one equation is the engine of the singularity theorems. Under an energy condition and vanishing rotation it forces an initially converging bundle to a caustic — a conjugate point where neighbouring geodesics cross and $\theta\to-\infty$ — in a bounded parameter interval. Applied to the null generators of horizons it yields the black-hole area theorem, the trapped-surface concept, and the Penrose–Hawking incompleteness proofs: congruences are how the local content of curvature becomes global statements about causal structure.

Key takeaways

A geodesic congruence is a family of geodesics filling a region so that exactly one curve threads each point, defining a smooth unit tangent field $\xi^\mu$ with $\xi^\nu\nabla_\nu\xi^\mu=0$ .
The transverse gradient $B_{\mu\nu}=\nabla_\nu\xi_\mu$ splits uniquely into expansion $\theta$ (trace), shear $\sigma_{\mu\nu}$ (symmetric trace-free), and rotation $\omega_{\mu\nu}$ (antisymmetric).
The Raychaudhuri equation $\dot\theta=-\tfrac13\theta^2-\sigma^2+\omega^2-R_{\mu\nu}\xi^\mu\xi^\nu$ evolves the expansion and ties focusing directly to the Ricci curvature — hence to matter through Einstein's equation.
For irrotational timelike geodesics with $R_{\mu\nu}\xi^\mu\xi^\nu\ge0$ (strong energy condition), a converging bundle ( $\theta_0<0$ ) reaches a caustic within proper time $\tau\le 3/|\theta_0|$ : gravity focuses, never defocuses.
The null version, with coefficient $\tfrac12$ instead of $\tfrac13$ , supplies the focusing at the heart of the Penrose–Hawking singularity theorems and the horizon area theorem.
Key idea: the transverse gradient of the geodesic tangent splits uniquely into expansion, shear, and rotation; Raychaudhuri's equation then turns Ricci curvature — and through Einstein's equation the local matter content — into an unavoidable focusing tendency for the whole bundle.

The core insight

The transverse gradient of the geodesic tangent splits uniquely into expansion (trace), shear (symmetric trace-free), and rotation (antisymmetric). Raychaudhuri's equation then turns the Ricci curvature — and through Einstein's equation the local matter content — into a focusing tendency for the whole bundle.

Where this lives today

The congruence machinery runs inside modern computations and observations. In numerical relativity, every binary black-hole merger locates its newly formed horizon with an apparent-horizon finder that solves for the marginally trapped surface where the outgoing null expansion $\theta=0$ ; the area theorem ( $\theta\ge0$ on the event horizon) then supplies a nondecreasing-area check against the LIGO–Virgo–KAGRA remnant masses for events like GW150914. In cosmology, the shear of a light-ray congruence is the weak-lensing "cosmic shear" that surveys such as Euclid and the Vera Rubin Observatory (LSST) measure across billions of galaxies: the Sachs optical equations — the null Raychaudhuri equation plus a shear-propagation equation — relate observed image distortions to the Ricci focusing of intervening matter and the Weyl tidal focusing of masses off the line of sight, turning distortion statistics into a map of dark matter. And the deepest application, Penrose's 1965 theorem that focusing plus a trapped surface forces geodesic incompleteness, earned the 2020 Nobel Prize — recognition, half a century on, that Raychaudhuri's scalar equation had settled whether real gravitational collapse must end in a singularity.

The physical picture

Release a small cloud of freely falling dust grains and watch its shape, not its centre. In an infinitesimal time the linear operator $B^\mu{}_\nu=\nabla_\nu\xi^\mu$ maps every separation vector to its neighbour, and that single operator carries exactly the three ways a shape can change: its size, its proportions, and its orientation. The expansion $\theta$ isotropically grows or shrinks the cloud; the shear $\sigma_{\mu\nu}$ stretches it along some axes and squeezes along others at fixed volume (the tidal "spaghettification" pattern); the rotation $\omega_{\mu\nu}$ rigidly twists it. The first figure freezes those three motions side by side.

Figure

The velocity gradient B decomposes into three motions of a test-particle ring, B = (θ/3)h + σ + ω. Expansion θ isotropically changes volume, shear σ distorts at fixed volume, rotation ω twists rigidly. The dashed circle is the initial cross section; only expansion changes the enclosed area

Only the Ricci term carries matter, and Ricci vanishes in vacuum — so a cloud released from rest in the empty field outside a star, with $\theta=\sigma=0$ at that instant, has $\mathrm{d}\theta/\mathrm{d}\tau=0$ there: tidal gravity stretches and squeezes without changing the volume, the exact statement that the Newtonian tidal tensor is trace-free. The volume does not stay fixed for long, though — the shear that promptly develops feeds back through $-\sigma^2$ and drives $\theta$ negative even in vacuum, and a congruence that starts with $\theta\neq0$ (radial infall, for instance) never had zero expansion to begin with. Push the cloud into matter, switch off its rotation, and the equation becomes a one-way street toward a caustic. The second figure makes that inevitability quantitative: the reciprocal expansion falls on a straight line and hits zero (so $\theta\to-\infty$ ) at a time set entirely by the initial convergence.

Figure

Focusing made linear. For an irrotational congruence with R ≥ 0, the reciprocal expansion obeys 1/θ ≥ 1/θ₀ + τ/3, a straight line of slope 1/3. Starting from θ₀ = −0.30 per unit length, 1/θ₀ = −3.33, so the line reaches zero — the caustic where θ → −∞ — no later than τ* = 3/0.30 = 10

A caustic is emphatically not a singularity of spacetime. The metric is perfectly regular there; what breaks down is the congruence, the assumption that one geodesic threads each point. The deep move of the singularity theorems is to show that once a caustic is forced to form, a globally causal spacetime cannot fit the resulting geodesic structure without being geodesically incomplete somewhere — and that incompleteness is the singularity.

The core equations

A timelike geodesic congruence is fixed by two conditions: the curves are geodesics, and the tangent is normalized to proper time. Everything downstream is built on this flow.

\xi^\nu\nabla_\nu\xi^\mu=0,\qquad \xi^\mu\xi_\mu=-1 \qquad (1)

where: $\xi^\mu$ — unit tangent field whose integral curves fill the region $\mathcal{O}$ ; $\nabla_\nu$ — Levi-Civita (metric-compatible, torsion-free) covariant derivative; $\xi^\mu\xi_\mu=-1$ — normalization to proper time for timelike curves (for a null congruence one replaces this by $k^\mu k_\mu=0$ and the affine condition $k^\nu\nabla_\nu k^\mu=0$ ). The deformation of the bundle is carried entirely by the covariant gradient of the tangent. Contracting it against $\xi$ on either index gives zero, so it lives in the 3-space orthogonal to the flow.

B_{\mu\nu}=\nabla_\nu\xi_\mu,\qquad B_{\mu\nu}\xi^\nu=0,\qquad \xi^\mu B_{\mu\nu}=0 \qquad (2)

where: $B_{\mu\nu}$ — velocity gradient, the linear map sending a separation vector to its rate of change; $B_{\mu\nu}\xi^\nu=0$ — follows from the geodesic equation (second index); $\xi^\mu B_{\mu\nu}=0$ — follows from constant norm, $\nabla_\nu(\xi^\mu\xi_\mu)=0$ (first index). So $B$ is purely transverse, acting only on the physical cross section. To decompose $B$ we need the metric of that transverse 3-space — the projector orthogonal to $\xi^\mu$ .

h_{\mu\nu}=g_{\mu\nu}+\xi_\mu\xi_\nu,\qquad h_{\mu\nu}\xi^\nu=0,\qquad h^\mu{}_\mu=3 \qquad (3)

where: $h_{\mu\nu}$ — induced metric on the 3-space orthogonal to $\xi^\mu$ , doubling as the projection operator; $g_{\mu\nu}$ — spacetime metric, signature $(-,+,+,+)$ ; $h^\mu{}_\mu=3$ — the trace equals the transverse dimension (it is $2$ for a null congruence, the origin of the different Raychaudhuri coefficient). Now split $B_{\mu\nu}$ into its trace, symmetric trace-free, and antisymmetric parts. This is the definition of expansion, shear, and rotation.

B_{\mu\nu}=\tfrac{1}{3}\theta\,h_{\mu\nu}+\sigma_{\mu\nu}+\omega_{\mu\nu} \qquad (4)

where: $\theta=h^{\mu\nu}B_{\mu\nu}=\nabla_\mu\xi^\mu$ — expansion, the fractional rate of change of transverse volume; $\sigma_{\mu\nu}=B_{(\mu\nu)}-\tfrac13\theta h_{\mu\nu}$ — shear, symmetric trace-free, volume-preserving distortion; $\omega_{\mu\nu}=B_{[\mu\nu]}$ — rotation (twist), antisymmetric. The split is unique and orthogonal, so $\theta$ , $\sigma$ , $\omega$ carry independent information. Differentiating $\theta$ along the flow and commuting derivatives with the Riemann tensor yields the master evolution law. Every term's sign is fixed by geometry.

\frac{\mathrm{d}\theta}{\mathrm{d}\tau}=-\tfrac{1}{3}\theta^2-\sigma_{\mu\nu}\sigma^{\mu\nu}+\omega_{\mu\nu}\omega^{\mu\nu}-R_{\mu\nu}\xi^\mu\xi^\nu \qquad (5)

where: $\mathrm{d}\theta/\mathrm{d}\tau=\xi^\nu\nabla_\nu\theta$ — rate of change of expansion along a geodesic; $\sigma^2\equiv\sigma_{\mu\nu}\sigma^{\mu\nu}\ge0$ — shear scalar, non-negative (purely spatial); $\omega^2\equiv\omega_{\mu\nu}\omega^{\mu\nu}\ge0$ — rotation scalar, the only term opposing focusing; $R_{\mu\nu}\xi^\mu\xi^\nu$ — Ricci focusing term, the sole channel for matter. Self-focusing ( $\theta^2$ ) and shear pull $\theta$ down, rotation props it up, gravity acts through Ricci. Matter enters only through that Ricci term, converted by the trace-reversed Einstein equation. The strong energy condition is exactly the statement that this contribution aids focusing.

R_{\mu\nu}\xi^\mu\xi^\nu=8\pi G\left(T_{\mu\nu}-\tfrac{1}{2}T\,g_{\mu\nu}\right)\xi^\mu\xi^\nu\ge0 \qquad (6)

where: $T_{\mu\nu}$ — stress–energy tensor, $T=g^{\mu\nu}T_{\mu\nu}$ its trace; $G$ — Newton's constant (geometrized units $c=1$ ); the inequality — the strong energy condition, equivalent to $\rho+3p\ge0$ for a perfect fluid. Ordinary matter makes gravity attractive; a cosmological constant or an inflaton violates it and can defocus. Dropping the two helpful terms in the irrotational case and integrating once gives the focusing inequality — the quantitative content of the second figure.

\frac{1}{\theta(\tau)}\ge\frac{1}{\theta_0}+\frac{\tau}{3} \qquad (7)

where: $\theta_0<0$ — initial expansion of a locally converging bundle; $\tau$ — proper time elapsed from the initial slice. The right-hand side climbs at fixed rate $1/3$ and must cross zero by $\tau=3/|\theta_0|$ , where $\theta\to-\infty$ and a caustic forms.

Derivation ADeriving the Raychaudhuri equation

The goal is to obtain the master focusing law from first principles: take the expansion $\theta=B^\mu{}_\mu$ , differentiate it along the flow, and see what geometry forces to appear. The only inputs are the definition $B_{\mu\nu}=\nabla_\nu\xi_\mu$ , the geodesic equation, the Ricci identity for commuting covariant derivatives, and the decomposition (4). What emerges is an equation in which self-focusing, shear, rotation, and Ricci curvature each enter with a sign that is not a convention but a theorem — and that fixed pattern of signs is the whole reason gravity focuses.

Step 1. Propagate the velocity gradient along the flow: $\xi^\rho\nabla_\rho B_{\mu\nu}=\xi^\rho\nabla_\rho\nabla_\nu\xi_\mu$ . Why: the change in the bundle's deformation is what we ultimately want, and $\theta$ is just its trace, so we track the full tensor first.

Step 2. Commute the two covariant derivatives with the Ricci identity $[\nabla_\rho,\nabla_\nu]\xi_\mu=R_{\mu\sigma\rho\nu}\xi^\sigma$ , giving $\xi^\rho\nabla_\rho\nabla_\nu\xi_\mu=\xi^\rho\nabla_\nu\nabla_\rho\xi_\mu+R_{\mu\sigma\rho\nu}\xi^\sigma\xi^\rho$ . Why: this is the single place curvature enters — reordering the derivatives is not free in a curved spacetime, and the price is exactly the Riemann tensor.

Step 3. On the reordered term use the geodesic equation: $\xi^\rho\nabla_\nu\nabla_\rho\xi_\mu=\nabla_\nu(\xi^\rho\nabla_\rho\xi_\mu)-(\nabla_\nu\xi^\rho)(\nabla_\rho\xi_\mu)=-B^\rho{}_\nu B_{\mu\rho}$ , since $\xi^\rho\nabla_\rho\xi_\mu=0$ kills the first piece. Why: geodesic motion collapses the awkward third-derivative term into a clean quadratic in $B$ , the source of the $\theta^2$ , $\sigma^2$ , $\omega^2$ terms.

Step 4. Assemble $\xi^\rho\nabla_\rho B_{\mu\nu}=-B_{\mu\rho}B^\rho{}_\nu+R_{\mu\sigma\rho\nu}\xi^\sigma\xi^\rho$ , a matrix Riccati equation for the velocity gradient. Why: this is the full tensor evolution; the expansion equation is its trace and the shear and rotation propagation equations are its trace-free and antisymmetric parts.

Step 5. Take the trace over $\mu,\nu$ . The Riemann term contracts to Ricci, $R^\mu{}_{\sigma\mu\rho}=R_{\sigma\rho}$ , leaving $\dfrac{\mathrm{d}\theta}{\mathrm{d}\tau}=-B_{\mu\nu}B^{\nu\mu}-R_{\sigma\rho}\xi^\sigma\xi^\rho$ . Why: tracing is exactly what turns the full curvature tensor into the Ricci curvature — which is why only Ricci, and never the tidal Weyl part, appears in the expansion law.

Step 6. Insert the decomposition (4). Cross terms vanish by orthogonality of the irreducible parts, and antisymmetry flips one sign: $B_{\mu\nu}B^{\nu\mu}=\tfrac13\theta^2+\sigma_{\mu\nu}\sigma^{\mu\nu}-\omega_{\mu\nu}\omega^{\mu\nu}$ . Why: the crucial minus sign on $\omega^2$ comes from $\omega^{\nu\mu}=-\omega^{\mu\nu}$ , and it is why rotation opposes focusing while shear reinforces it.

Step 7. Substitute to reach $\dfrac{\mathrm{d}\theta}{\mathrm{d}\tau}=-\tfrac13\theta^2-\sigma^2+\omega^2-R_{\mu\nu}\xi^\mu\xi^\nu$ , the Raychaudhuri equation, with $\tfrac13$ traced to the 3-dimensional transverse space. Why: every sign is now manifest and physical — this is the finished law that the singularity theorems will exploit. *Why the coefficients: the $\tfrac13$ coefficient equals $1/(D-1)$ with transverse dimension $D-1=3$ , and re-tracing with $h^\mu{}_\mu=2$ reproduces the null coefficient $\tfrac12$ .*

Go deeper · Derivation B · 1 minThe focusing theorem and the caustic bound

Derivation BThe focusing theorem and the caustic bound

The goal is to show that once a timelike geodesic bundle is converging anywhere, gravity plus geometry guarantee it collapses to a caustic within a bounded proper time — the result that makes the singularity theorems possible. Assume the congruence is irrotational ( $\omega=0$ , justified by the next derivation) and that the strong energy condition holds, so $R_{\mu\nu}\xi^\mu\xi^\nu\ge0$ . The argument is a one-line differential inequality followed by an integration, and it delivers not just "a caustic forms" but the explicit deadline $\tau\le3/|\theta_0|$ .

Step 1. With $\omega=0$ , both $\sigma^2\ge0$ and $R_{\mu\nu}\xi^\mu\xi^\nu\ge0$ appear with minus signs, so $\dfrac{\mathrm{d}\theta}{\mathrm{d}\tau}=-\tfrac13\theta^2-\sigma^2-R_{\mu\nu}\xi^\mu\xi^\nu\le-\tfrac13\theta^2$ . Why: we discard exactly the terms that can only accelerate collapse, turning an equation into a clean inequality that still bounds the worst case.

Step 2. Change variables to $u=1/\theta$ . Then $\dot u=-\theta^{-2}\dot\theta\ge\tfrac13$ , a constant lower bound on the slope. Why: the nonlinear $\theta^2$ term becomes linear under the reciprocal, converting a hard ODE into a trivial one that we can integrate by inspection.

Step 3. Integrate from the initial slice where $\theta=\theta_0<0$ : $\dfrac{1}{\theta(\tau)}\ge\dfrac{1}{\theta_0}+\dfrac{\tau}{3}$ . Since $\theta_0<0$ , the starting value $1/\theta_0$ is negative and the right side rises steadily toward zero. Why: this is the straight line of the second figure — the whole focusing story is a line of slope $1/3$ climbing to the axis.

Step 4. The right side reaches zero at $\tau=3/|\theta_0|$ . Because $1/\theta\to0^-$ there, $\theta\to-\infty$ : a conjugate point (caustic) has formed no later than $\tau_*=3/|\theta_0|$ . Why: a diverging expansion at finite proper time is precisely a point where neighbouring geodesics cross, the seed the global theorems need. *Numerically, $\theta_0=-0.30$ per unit length gives $\tau_*=10$ , and shear or matter only shortens it.*

Go deeper · Derivation C · 1 minHypersurface orthogonality kills the rotation

Derivation CHypersurface orthogonality kills the rotation

The goal is to justify the irrotational assumption the focusing theorem leans on. We show that a geodesic congruence orthogonal to a family of hypersurfaces automatically has $\omega_{\mu\nu}=0$ , so whenever the flow is a gradient — as for the fundamental observers in cosmology or the generators of a horizon defined by a level function — the one term that could resist collapse is simply absent. The engine is Frobenius' integrability theorem, which converts a statement about surfaces into an algebraic condition on the twist of $\xi$ .

Step 1. Suppose the congruence is orthogonal to the level sets of a scalar $\Phi$ , so $\xi_\mu\propto\nabla_\mu\Phi$ . Frobenius' theorem states that such a hypersurface-orthogonal 1-form obeys $\xi_{[\mu}\nabla_\nu\xi_{\rho]}=0$ . Why: this integrability condition is exactly the geometric meaning of "the flow is everywhere perpendicular to a stack of surfaces".

Step 2. Write the rotation as the projected antisymmetrized gradient, $\omega_{\mu\nu}=h^\rho{}_\mu h^\sigma{}_\nu\nabla_{[\sigma}\xi_{\rho]}$ . Why: $\omega$ is by definition the transverse, antisymmetric part of $\nabla\xi$ , so any condition on $\nabla_{[\nu}\xi_{\mu]}$ controls it directly.

Step 3. Expand the Frobenius condition and contract with $\xi^\rho$ . Using the geodesic condition $a_\mu=\xi^\nu\nabla_\nu\xi_\mu=0$ , the acceleration term drops and the projected antisymmetric part of $\nabla\xi$ is forced to vanish. Why: the acceleration is the only obstruction to the identification, and geodesics have none, so the implication is clean.

Step 4. Conclude $\omega_{\mu\nu}=0$ : a geodesic, hypersurface-orthogonal congruence has no twist. Why: this removes the sole $+\omega^2$ term from Raychaudhuri's equation, leaving focusing unopposed — the precise hypothesis the singularity theorems require.

Limiting cases

The decomposition and the Raychaudhuri equation specialize sharply with the causal character of the congruence and its symmetry:

Case	Result	Meaning
Timelike congruence	$\dfrac{\mathrm{d}\theta}{\mathrm{d}\tau}=-\tfrac{1}{3}\theta^2-\sigma^2+\omega^2-R_{\mu\nu}\xi^\mu\xi^\nu$	Transverse space is 3-dimensional, hence the $\tfrac13\theta^2$ coefficient; $\theta$ is the fractional rate of change of a 3-volume per unit proper time $\tau$ .
Null congruence	$\dfrac{\mathrm{d}\theta}{\mathrm{d}\lambda}=-\tfrac{1}{2}\theta^2-\sigma^2+\omega^2-R_{\mu\nu}k^\mu k^\nu$	Transverse space is 2-dimensional, giving $\tfrac12\theta^2$ ; $\theta$ is the fractional rate of change of cross-sectional area per affine parameter $\lambda$ , with $k^\mu k_\mu=0$ .
Hypersurface-orthogonal	$\omega_{\mu\nu}=0\iff\xi_{[\mu}\nabla_\nu\xi_{\rho]}=0$	By Frobenius' theorem the flow is a gradient; rotation vanishes and the focusing term loses its only opponent.
Irrotational, energy condition	$R_{\mu\nu}\xi^\mu\xi^\nu\ge0,\ \omega=0\Rightarrow\theta^{-1}\ge\theta_0^{-1}+\tfrac{1}{3}\tau$	If $\theta_0<0$ anywhere, $\theta\to-\infty$ within $\tau\le3/\|\theta_0\|$ : the focusing theorem.
FLRW cosmology	$\theta=3H,\ \sigma=\omega=0\Rightarrow\dot H+H^2=-\tfrac{4\pi G}{3}(\rho+3p)$	The fundamental observers form a shear-free, irrotational geodesic congruence; Raychaudhuri becomes the acceleration equation.

Common trapA caustic is not a spacetime singularity

When $\theta\to-\infty$ at a conjugate point it is tempting to declare spacetime singular there. It is not. A caustic is a failure of the congruence, not of the manifold: neighbouring geodesics of that family have crossed, so the "one curve per point" description breaks down — exactly as light rays focused by a lens cross at a focal point in perfectly ordinary flat space. The metric, curvature, and geodesics themselves remain smooth through a caustic. The singularity theorems never claim the caustic is the singularity; they use its forced existence, combined with global causal hypotheses (global hyperbolicity, a trapped surface or a Cauchy slice), to prove that some geodesic is inextendible — geodesic incompleteness, the technical definition of a singularity, generically located elsewhere. Confusing the local caustic with the global incompleteness is the commonest conceptual error in reading the theorems.

Worked examples

1 · The Big Bang as a caustic of the cosmic congruence (fully worked)

The fundamental (comoving) observers in a Friedmann–Lemaître–Robertson–Walker universe form an irrotational, shear-free timelike geodesic congruence. Show that the Raychaudhuri equation reduces to the cosmological acceleration equation, and use it to place the Big Bang as a caustic of this congruence in a matter-dominated model. Take $H_0=70\ \mathrm{km\,s^{-1}\,Mpc^{-1}}$ .

Step 1. For FLRW the expansion is isotropic, so $\theta=\nabla_\mu\xi^\mu=3\dot a/a=3H$ , with $\sigma=\omega=0$ . Then $\mathrm{d}\theta/\mathrm{d}\tau=3\dot H$ . Raychaudhuri (5) becomes $3\dot H=-\tfrac13(3H)^2-R_{\mu\nu}\xi^\mu\xi^\nu$ , i.e. $\dot H+H^2=-\tfrac13R_{\mu\nu}\xi^\mu\xi^\nu$ . This is $\ddot a/a=-\tfrac13R_{\mu\nu}\xi^\mu\xi^\nu$ .

Step 2. Insert matter through (6). For a perfect fluid $R_{\mu\nu}\xi^\mu\xi^\nu=4\pi G(\rho+3p)$ , so $\dfrac{\ddot a}{a}=-\dfrac{4\pi G}{3}(\rho+3p)$ — the acceleration equation. Ordinary matter ( $\rho+3p>0$ ) decelerates the expansion; the strong energy condition is precisely $\rho+3p\ge0$ .

Step 3. Take pressureless dust ( $p=0$ ), the Einstein–de Sitter model, where $a\propto\tau^{2/3}$ . Then $H=\dfrac{2}{3\tau}$ and the congruence expansion is $\theta=3H=\dfrac{2}{\tau}$ . Check consistency: $\mathrm{d}\theta/\mathrm{d}\tau=-2/\tau^2$ , while $-\tfrac13\theta^2-4\pi G\rho=-\tfrac{4}{3\tau^2}-\tfrac{2}{3\tau^2}=-\dfrac{2}{\tau^2}$ using $8\pi G\rho=3H^2=4/(3\tau^2)$ . The two sides agree exactly — Raychaudhuri is internally consistent for FLRW.

Step 4. Run the flow backward. Since $\theta=2/\tau\to+\infty$ as $\tau\to0^+$ , the entire comoving congruence emanates from a single caustic at $\tau=0$ — the Big Bang, where every fundamental geodesic converges as the expansion diverges. Its distance in the past is the age $\tau_0=2/(3H_0)$ . With $H_0=70\ \mathrm{km\,s^{-1}\,Mpc^{-1}}=2.27\times10^{-18}\ \mathrm{s^{-1}}$ , the present expansion is $\theta_0=3H_0=6.8\times10^{-18}\ \mathrm{s^{-1}}$ and $\tau_0=2.9\times10^{17}\ \mathrm{s}\approx9.3\ \mathrm{Gyr}$ . The Big Bang is a congruence caustic $\approx9.3$ Gyr in the past of an Einstein–de Sitter universe.

2 · The caustic deadline for a converging bundle (fully worked)

An irrotational timelike geodesic congruence in a region satisfying the strong energy condition is measured to have expansion $\theta_0=-0.30$ per unit length (a locally converging bundle) on an initial slice. By what proper time must a caustic form, and how do shear and matter change the answer?

Step 1. Use the focusing inequality (7): $1/\theta(\tau)\ge1/\theta_0+\tau/3$ . Here $1/\theta_0=1/(-0.30)=-3.33$ units of length. The reciprocal expansion starts at $-3.33$ and rises at the fixed rate $1/3$ .

Step 2. A caustic occurs when $\theta\to-\infty$ , i.e. when the right-hand side reaches zero: $-3.33+\tau_*/3=0\Rightarrow\tau_*=3/|\theta_0|=3/0.30=\mathbf{10}$ units of length. The bundle must reach a conjugate point by $\tau_*=10$ — this is the caustic marked in the second figure.

Step 3. The inequality is a worst-case bound. Any shear ( $\sigma^2>0$ ) or focusing matter ( $R_{\mu\nu}\xi^\mu\xi^\nu>0$ ) adds a further negative term to $\mathrm{d}\theta/\mathrm{d}\tau$ , so the true caustic arrives no later than $\tau_*=10$ and generically sooner. Only rotation ( $\omega^2>0$ ) could push it past the deadline — and hypersurface orthogonality forbids that.

Physics of the portrait

The equation at the centre of this article carries one name because one person, working largely in isolation in Calcutta, saw that the collapse of a matter bundle could be read off from a single scalar without ever choosing coordinates.

Origins · 1 minAmal Kumar Raychaudhuri and how this was discovered

Amal Kumar Raychaudhuri

1923–2005 · the covariant focusing equation and the road to the singularity theorems

Amal Kumar Raychaudhuri derived his equation in 1955, in a short Physical Review paper titled "Relativistic Cosmology I," while at the Indian Association for the Cultivation of Science in Calcutta. His aim was concrete and unfashionable: to ask, without assuming the high symmetry that made every known cosmological model artificially smooth, whether a cloud of dust in general relativity must collapse to a singularity. Following the expansion of a congruence covariantly — no special coordinates, no imposed symmetry — he showed that ordinary attractive matter drives $\theta$ inexorably downward, so that an initially collapsing, non-rotating cloud reaches a singularity of the flow in finite proper time. The result cut against the widespread hope, encouraged by the idealized Friedmann models, that a generic universe might "bounce" and avoid a beginning. For a decade the equation sat mostly unremarked. Then, in the mid-1960s, Roger Penrose and Stephen Hawking recognized that Raychaudhuri's focusing, promoted to null geodesics and combined with global causal reasoning and the notion of a trapped surface, was exactly the local ingredient needed to prove that singularities are generic — not artifacts of symmetry. The Penrose–Hawking theorems, and with them the modern understanding that black holes and the Big Bang harbour genuine singularities, rest on the equation Raychaudhuri wrote down. He remained a devoted teacher in Calcutta for the rest of his career; the covariant decomposition into expansion, shear, and rotation that frames the equation was cast in its enduring form by Jordan, Ehlers, and Sachs a few years later.

History

1955

— Raychaudhuri publishes the focusing equation for the expansion of a geodesic congruence (Phys. Rev. 98, 1123)

1961

— Jordan, Ehlers, and Sachs give the covariant kinematic decomposition into expansion, shear, and rotation

1965

— Penrose proves the first singularity theorem using null geodesic focusing and trapped surfaces

1970

— Hawking and Penrose prove the general singularity theorem for cosmology and gravitational collapse

2020

— Penrose shares the Nobel Prize in Physics for showing that black-hole formation is a robust prediction of general relativity

Going further

Texts. Wald, General Relativity (1984), §9.2 (congruences) and §9.3–9.5 (singularity theorems) · Misner, Thorne, and Wheeler, Gravitation (1973), §22.5 and Box 22.1 · Carroll, Spacetime and Geometry (2004), §3.10 · Hartle, Gravity (2003), Ch. 22 (for the FLRW reduction) · Hawking and Ellis, The Large Scale Structure of Space-Time (1973), §4.1–4.4 · Poisson, A Relativist's Toolkit (2004), Ch. 2 · Raychaudhuri, Phys. Rev. 98, 1123 (1955). The $-\tfrac13\theta^2$ (timelike) and $-\tfrac12\theta^2$ (null) coefficients follow from the transverse-space dimension $1/(D-1)$ . The $+\omega^2$ sign follows from antisymmetry. The focusing bound $\tau_*=3/|\theta_0|$ is obtained by integrating $\mathrm{d}(1/\theta)/\mathrm{d}\tau\ge\tfrac13$ , giving $\tau_*=10$ for $\theta_0=-0.30$ . The FLRW reduction gives $\theta=3H$ and $\ddot a/a=-\tfrac{4\pi G}{3}(\rho+3p)$ , with the Einstein–de Sitter identity $\theta=2/\tau$ and age $\tau_0=2/(3H_0)\approx9.3$ Gyr for $H_0=70\ \mathrm{km\,s^{-1}\,Mpc^{-1}}$ .

Builds on: Geodesics and the geodesic equation · Riemann and Ricci curvature · Null Geodesics and Light Cones · Causal Structure of Spacetime Leads to: Global Hyperbolicity · Domain of Dependence and Cauchy Horizons · Penrose–Carter Conformal Diagrams · Conformal Infinity and Asymptotic Structure

Ask the tutor: Ask me to walk through how Penrose promoted this focusing argument to null geodesics and turned a trapped surface into a proof of geodesic incompleteness, or to derive the shear propagation equation as the trace-free partner of Raychaudhuri.

What is the difference between expansion, shear, and rotation, physically?

They are the three independent ways an infinitesimal ball of test particles can deform as the congruence flows. Expansion $\theta$ is the isotropic part: it grows or shrinks the ball's volume without changing its shape. Shear $\sigma_{\mu\nu}$ is the symmetric trace-free part: it stretches the ball along some axes and squeezes along others, turning a sphere into an ellipsoid at fixed volume — the fingerprint of tidal gravity. Rotation $\omega_{\mu\nu}$ is the antisymmetric part: it rigidly twists the ball with no change of shape or volume. Being orthogonal pieces of $B_{\mu\nu}$ , the three carry independent information and evolve by separate propagation equations.

Why must the congruence be irrotational for the singularity theorems?

Rotation is the only Raychaudhuri term that opposes focusing — it enters with a plus sign because $\omega_{\mu\nu}$ is antisymmetric. Physically a swirling bundle has centrifugal support: the rotation holds neighbouring geodesics apart, just as angular momentum stops a spinning gas cloud from collapsing to a point. If $\omega^2$ is large enough it overwhelms the focusing terms and no caustic forms. The theorems therefore restrict to irrotational congruences, which by Frobenius' theorem are exactly the hypersurface-orthogonal ones — the generators of a Cauchy slice or of a horizon. Rotating collapse is genuinely subtler, one reason the interior of a Kerr black hole is so much richer than the non-rotating case.

Why is the coefficient 1/3 for timelike but 1/2 for null congruences?

The coefficient of $\theta^2$ is $1/(D-1)$ , where $D-1$ is the dimension of the genuinely transverse space — the cross section whose volume $\theta$ measures. For a timelike congruence the tangent removes one timelike direction, leaving a 3-dimensional spatial cross section, so the coefficient is $1/3$ and $\theta$ tracks a volume. For a null congruence the tangent $k^\mu$ is itself null and lies within the wavefront it is orthogonal to; modding out that degenerate direction leaves only a 2-dimensional transverse space, so the coefficient is $1/2$ and $\theta$ measures the rate of change of cross-sectional area. Getting this dimension wrong is the most common source of an incorrect Raychaudhuri coefficient.

How is the Raychaudhuri equation related to the geodesic deviation equation?

They are the same physics at different levels. Geodesic deviation governs a single separation vector $\eta^\mu$ : it is Lie-dragged by the flow, so $\xi^\nu\nabla_\nu\eta^\mu=B^\mu{}_\nu\eta^\nu$ , and differentiating again brings in the Riemann tensor as a tidal force. Raychaudhuri is what you get by tracking the whole cross section at once — it is the trace of the propagation equation for $B_{\mu\nu}$ , which governs every deviation vector simultaneously. Geodesic deviation is the equation for one thread of the bundle; Raychaudhuri is the equation for the collective volume of all of them.

Does the equation apply to accelerated, non-geodesic congruences?

Yes, with an extra term. If the curves are not geodesics — say, observers held static in a gravitational field — then $a^\mu=\xi^\nu\nabla_\nu\xi^\mu\ne0$ , and the generalized equation acquires a $+\nabla_\mu a^\mu$ contribution alongside the $\theta^2$ , $\sigma^2$ , $\omega^2$ , and Ricci terms. That acceleration divergence can locally oppose focusing, which is why a rocket firing its engines, or a star supported by pressure gradients, need not collapse. The geodesic form used in the singularity theorems is the special case $a^\mu=0$ : genuine free fall, where nothing but gravity and the congruence's own kinematics act.

What does the black-hole area theorem have to do with all this?

The event horizon is generated by null geodesics that never leave it, so it is a null congruence obeying the null Raychaudhuri equation. Suppose the generators' expansion ever went negative. With the null energy condition ensuring $R_{\mu\nu}k^\mu k^\nu\ge0$ , the equation would force $\theta\to-\infty$ — a caustic — in finite affine parameter, where a generator would cross another and could leave the horizon. But horizon generators, by definition, have no future endpoints. The contradiction forces $\theta\ge0$ everywhere on the horizon, so its area can never decrease. That is Hawking's area theorem, the classical seed of black-hole thermodynamics and the analogue of the second law.