Causality Conditions Hierarchy

The ladder is not a museum piece; its top rung is a working hypothesis in nearly every rigorous result about dynamical spacetimes. The Choquet-Bruhat–Geroch theorem says the vacuum Einstein equations, posed as initial data on a slice, have a unique maximal globally hyperbolic development — so the entire mathematical theory of the Cauchy problem in general relativity, including the numerical relativity that produced the gravitational-wave templates matched against LIGO data, is set on globally hyperbolic backgrounds by construction. The strong cosmic censorship conjecture, one of the field's central open problems, is precisely the claim that generic initial data develop into a maximally extended globally hyperbolic spacetime that cannot be smoothly extended past its Cauchy horizon — that nature refuses to let the lower rungs fail. Recent work on the interior of rotating and charged black holes tests censorship head-on, probing whether the Cauchy horizon inside a Kerr black hole is stable or, as censorship demands, destroyed by infalling radiation. In quantum field theory on curved spacetime and its algebraic formulation, global hyperbolicity is the standing assumption that gives the wave operator unique retarded and advanced Green functions and a well-defined vacuum, while the AdS/CFT correspondence lives in the world of the fourth rung — a stably causal but not globally hyperbolic spacetime whose missing Cauchy surface is exactly why the boundary theory must supply the missing data. Even the causal set programme for quantum gravity leans on the ladder, since the reconstruction theorems that let a discrete causal order encode a continuum metric require the approximated spacetime to be at least distinguishing.

Read the causal structure as a relation on events: $p$ precedes $q$ when a future-directed causal curve runs from $p$ to $q$. One would like this to be a partial order — reflexive, antisymmetric, transitive — but a general spacetime refuses even that. At the bottom of the ladder the relation is not antisymmetric: a closed timelike curve makes an event precede itself. Climbing the ladder repairs the order one defect at a time. First no event precedes itself (chronology, then causality). Then distinct events are told apart by the sets of events they can influence (distinguishing). Then no event is menaced by curves that loop arbitrarily close (strong causality). Then the order is robust against jitter in the metric (stable causality). Finally the order is complete enough to slice spacetime into space and run it forward in time (global hyperbolicity). A slogan captures the shift. The lower conditions are about curves that must not exist — no closed timelike curve, no closed causal curve, no almost-closed causal curve. The upper conditions are about objects that do exist — a global time function, a Cauchy surface, a maximal geodesic between causally related points. The dividing line is stable causality, where the whole style of argument flips from forbidding bad curves to constructing good functions. The first figure draws this ladder in full.

The hierarchy is conformally invariant. Whether a vector is timelike, null, or spacelike depends only on its sign under $g$, and a positive rescaling $g\to\Omega^{2}g$ leaves those signs untouched. So every rung is a property of the conformal class $[g]$ alone, which is precisely why a Penrose–Carter diagram — a metric rescaled to drag infinity into finite reach — displays each condition without distortion. Stability, the idea that separates stable causality from ordinary causality, is the subtle one: a spacetime can be perfectly causal yet balance on a knife-edge, so that an arbitrarily faint widening of its light cones produces a closed causal curve. Refusing to sit on that edge is what forces a global clock into existence. The second figure shows the standard witness that stable causality is strictly weaker than global hyperbolicity — anti-de Sitter space — and pins down the number that proves it.

The whole hierarchy is built from two relations on events, so we fix those first. Chronological precedence $p\ll q$ means a future-directed timelike curve runs from $p$ to $q$; causal precedence $p\le q$ means $p=q$ or a future-directed causal (timelike-or-null) curve does. Their reach from a point defines the four fundamental sets.

where: $p\ll q$ — a future-directed timelike curve joins $p$ to $q$ (chronological precedence); $p\le q$ — $p=q$ or a future-directed causal curve joins them; $I^{+}(p)$ — the chronological future, always an open set; $J^{+}(p)$ — the causal future, in general neither open nor closed. The pasts $I^{-},J^{-}$ are defined by reversing the time direction. Every rung of the ladder is a statement about these four sets. The lowest rung simply asks that no event lie in its own chronological future — that time not close up on itself.

where: $p\not\ll p$ — no closed timelike curve passes through $p$, equivalently the relation $\ll$ is irreflexive. Reading it: chronology is the weakest demand one could reasonably make, yet it already fails in exact solutions of Einstein's equations such as the Gödel universe, and it fails automatically on any compact spacetime, which must contain a closed timelike curve. The next rung strengthens "timelike" to "causal," outlawing closed null curves as well — a genuinely stronger demand, since a closed null geodesic can exist where no closed timelike curve does.

where: a causal curve has a tangent that is everywhere timelike or null and future-directed. Reading it: causality $\Rightarrow$ chronology because a timelike curve is causal, but the converse fails, so the two rungs are distinct. Above causality comes the distinguishing condition, which asks that an event be recoverable from the events it can reach.

where: the first clause is future-distinguishing, the second past-distinguishing; a spacetime is distinguishing when both hold. Reading it: no two distinct events may share a chronological future or a chronological past. Distinguishing is strictly stronger than causality and strictly weaker than strong causality, the rung that forbids curves returning arbitrarily near their start.

where: $U,V$ — open neighborhoods of $p$; "meets $V$ once" — the curve's intersection with $V$ is connected, so once a curve leaves $V$ it never returns. Reading it: strong causality forbids the almost-closed causal curves that distinguishing spacetimes may still harbor, and it has an exact topological meaning — it holds precisely when the Alexandrov topology built from the diamonds $I^{+}(p)\cap I^{-}(q)$ reproduces the manifold topology. The watershed rung above it demands robustness under a widening of the cones.

where: $\tilde g\succ g$ — the cones of $\tilde g$ strictly contain those of $g$ at every point; a global time function is a continuous $t:M\to\mathbb{R}$ with $p\le q,\ p\ne q\Rightarrow t(p)<t(q)$. Reading it: this is Hawking's theorem — the robustness of causality against cone-opening is the same statement as owning a clock that ticks forward along every causal curve. The summit of the ladder replaces "no bad curves" with a compactness demand that turns out to be everything.

where: $J^{+}(p)\cap J^{-}(q)$ — the causal diamond between $p$ and $q$; a Cauchy surface is an achronal set met exactly once by every inextendible causal curve. Reading it: this is the modern Bernal–Sánchez form of global hyperbolicity, and it is equivalent to a Cauchy surface and to a smooth splitting $M\cong\mathbb{R}\times\Sigma$. The older Leray–Hawking–Ellis definition also assumed strong causality; that assumption is now known to be implied, not needed.

Concrete spacetimes locate each rung and, more importantly, the gap between neighboring rungs. Each row below is a witness that one condition can hold while the next one up fails.

1 · The Gödel critical radius (fully worked)

In Gödel's rotating universe, written in cylindrical coordinates about an axis, the circles of constant $t$, $r$, $z$ with $\phi$ running over $[0,2\pi)$ have squared length proportional to $\sinh^{2}r-\sinh^{4}r$. Find the radius beyond which these circles become closed timelike curves, and state what rung of the ladder Gödel therefore violates.

Step 1. Read off the sign that decides causal character. A circle of constant $t,r,z$ is a closed curve; its tangent is timelike exactly when its squared norm is negative, i.e. when the coefficient $g_{\phi\phi}\propto\sinh^{2}r-\sinh^{4}r=\sinh^{2}r\,(1-\sinh^{2}r)$ turns negative. Since $\sinh^{2}r\ge 0$ always, the sign is controlled by the factor $(1-\sinh^{2}r)$.

Step 2. Locate the sign change. The coefficient is positive (spacelike circle) for $\sinh^{2}r<1$ and negative (timelike circle) for $\sinh^{2}r>1$, so the critical radius solves $\sinh^{2}r_{c}=1$, that is $\sinh r_{c}=1$.

Step 3. Evaluate. Inverting, $r_{c}=\operatorname{arcsinh}(1)=\ln\!\big(1+\sqrt2\big)$. Numerically $1+\sqrt2=2.41421$, so $r_{c}=\ln 2.41421=\mathbf{0.8814}$ in units of the Gödel length scale. For $r>r_{c}$ the $\phi$-circles are genuine closed timelike curves, and because the spacetime is homogeneous one passes through every event.

Step 4. Place Gödel on the ladder. A closed timelike curve is the direct negation of chronology, so Gödel fails the lowest rung and therefore every rung above it. $r_c = \operatorname{arcsinh}(1)=\ln(1+\sqrt2)\approx 0.8814$, and beyond it Gödel violates the chronology condition. This is the exact solution that first proved general relativity permits time travel, and it is why chronology must be imposed as an extra assumption, not derived from the field equations.

2 · Why anti-de Sitter has no Cauchy surface (fully worked)

Global anti-de Sitter space in static conformal coordinates has line element $\mathrm{d}s^{2}=\sec^{2}\!\rho\,(-\mathrm{d}t^{2}+\mathrm{d}\rho^{2}+\sin^{2}\!\rho\,\mathrm{d}\Omega^{2})$ with $\rho\in[0,\pi/2)$. Compute the coordinate time a radial light ray takes to travel from the center to the boundary, and use it to explain why no causal diamond is compact.

Step 1. Write the condition for a radial null ray. Set $\mathrm{d}s^{2}=0$ and $\mathrm{d}\Omega=0$. The common conformal factor $\sec^{2}\!\rho$ divides out, leaving $-\mathrm{d}t^{2}+\mathrm{d}\rho^{2}=0$, so $\mathrm{d}t=\pm\,\mathrm{d}\rho$ — radial light travels at $45^{\circ}$ in the $(\rho,t)$ plane, exactly as the figure draws it.

Step 2. Integrate to the boundary. A future-directed ray from the center $\rho=0$ at $t=0$ obeys $t=\rho$, so it reaches the conformal boundary $\rho=\pi/2$ at coordinate time $\Delta t=\pi/2=\mathbf{1.5708}$. The boundary is a finite coordinate-time away even though it is infinitely far in proper distance, because $\sec^{2}\!\rho$ blows up there.

Step 3. Draw the causal consequence. Pick any interior event $p$; its causal future $J^{+}(p)$ runs off to the boundary in finite time and the causal diamond $J^{+}(p)\cap J^{-}(q)$ therefore includes limit points that lie on the boundary, which is not part of $M$. The diamond is thus not compact — it is missing its boundary limit points.

Step 4. Read the verdict off the ladder. Anti-de Sitter is causal and even stably causal (the coordinate $t$ is a global time function), but compactness of diamonds fails, so by the modern criterion it is not globally hyperbolic and possesses no Cauchy surface. A signal reaches infinity in $\Delta t=\pi/2\approx1.571$, so evolution needs boundary data at $\rho=\pi/2$ and is not fixed by an initial slice alone — the physical reason AdS/CFT must specify what happens at the boundary.

3 · Reading the strict inclusions off a tilted cylinder

Take two-dimensional Minkowski space and roll the spatial axis into a circle, giving a cylinder $\mathbb{R}_t\times S^{1}_x$; then progressively tilt the light cones toward the compact $x$-direction. Track which rung fails as the tilt increases.

Step 1. At zero tilt the cylinder is globally hyperbolic: each circle $t=\text{const}$ is a Cauchy surface, met once by every inextendible causal curve, so all conditions hold at once.

Step 2. As the cones tilt, the future null direction acquires a component around the circle. Before either null direction becomes tangent to $S^{1}$, no causal curve can close, but causal curves can be made to spiral arbitrarily close to their starting circle — strong causality is the first rung to fall while causality and distinction survive.

Step 3. At the critical tilt one null direction becomes exactly tangent to $S^{1}$: that null generator closes into a closed null curve, breaking causality while chronology (no closed timelike curve yet) still holds — the concrete witness that causality is strictly stronger than chronology.

Step 4. Past the critical tilt the cones open enough that closed timelike curves appear and even chronology fails. The single family of cylinders walks down the ladder one rung at a time, exhibiting each strict inclusion as a distinct tilt threshold — the cleanest laboratory for the whole hierarchy.

The causality conditions are a monument built by a small group of relativists in the 1960s and 70s, but if one name belongs on the watershed rung it is Stephen Hawking, who proved that stable causality is the same thing as owning a clock and, with George Ellis, wrote the book that fixed the ladder in its modern form.

Texts. Hawking & Ellis, The Large Scale Structure of Space-Time (1973), ch. 6 — the founding treatment of the ladder · Wald, General Relativity (1984), ch. 8 — global hyperbolicity and the Cauchy problem · Penrose, Techniques of Differential Topology in Relativity (1972) — the causal-structure toolkit · Beem, Ehrlich & Easley, Global Lorentzian Geometry, 2nd ed. (1996), ch. 3 · Minguzzi & Sánchez, "The causal hierarchy of spacetimes" (2008) — the definitive modern survey with every counterexample · Misner, Thorne & Wheeler, Gravitation (1973), ch. 34 for light cones and causal order. The key anchor results are: the Gödel critical radius $r_c=\operatorname{arcsinh}(1)=\ln(1+\sqrt2)\approx0.8814$, beyond which $\phi$-circles become closed timelike curves; the anti-de Sitter boundary transit time $\Delta t=\pi/2\approx1.571$, derived from $\mathrm{d}t=\mathrm{d}\rho$, showing causal diamonds are non-compact; Hawking's equivalence between stable causality and the existence of a global time function; and the Bernal–Sánchez form of global hyperbolicity. The Gödel and AdS line elements used throughout are the standard textbook forms.

Builds on: Causal Structure of Spacetime · Chronological and Causal Futures and Pasts · Achronal Sets and Edges · Conformal Transformations and Causal Invariance Leads to: Global Hyperbolicity · Domain of Dependence and Cauchy Horizons · Cauchy Surfaces · Penrose–Carter Conformal Diagrams

Ask the tutor: Ask me to walk a specific spacetime — Gödel, Taub–NUT, anti-de Sitter, or Kerr's interior — up or down the ladder and pinpoint the exact rung where its causal structure breaks.