13.09.01 · gr-cosmology / microlocal-qft-curved-spacetimes

Globally hyperbolic Lorentzian manifolds

shipped3 tiersLean: none

Anchor (Master): Hawking and Ellis, The Large Scale Structure of Space-Time (Cambridge, 1973), Ch. 6; Bernal and Sánchez, Comm. Math. Phys. 257 (2005) 43; Gérard, Microlocal Analysis of Quantum Fields on Curved Spacetimes (EMS, 2019), Ch. 5

Intuition Beginner

Special relativity gave you the light cone at every point of flat spacetime. The light cone divides the directions you could possibly move into three classes — inside the cone (slower than light, a real worldline), on the cone (a photon's path), and outside the cone (a direction no signal can travel). Causality is a feature of this cone structure: an event can influence another only if a forward-light-cone path connects them.

General relativity keeps the light cones but lets them tilt and warp from point to point. The metric tensor on a curved spacetime still picks out, at each event, a double cone in the tangent space — the set of directions where the metric squared-length is zero. Inside that cone lie the worldlines of matter; outside lie the spacelike directions. The tilt of the cone is the visible face of gravity. Near a black hole, the cones tilt inward so strongly that the future of every point inside the horizon lies at the singularity.

A spacetime is well-behaved if its cones do not let you travel into your own past. Closed timelike curves — paths that come back to where they started in the future direction — are the pathology to avoid. Kurt Gödel found in 1949 that Einstein's equations admit a rotating cosmological solution riddled with them. Such universes break causality so badly that nothing recognisable as physics can happen on them.

The cleanest causality condition is global hyperbolicity. It says two things. First, no closed (or near-closed) causal curves: events are partially ordered by "before" and "after." Second, the spacetime admits a Cauchy hypersurface — a slice on which you can specify initial data for a field, after which the field is determined throughout the whole spacetime. This is the relativistic analogue of saying "the universe has a moment-of-time slice you can start from."

Globally hyperbolic spacetimes are the natural arena for quantum field theory on a curved background, the setting in which the Klein-Gordon equation $(□_{g} + m^{2}) ϕ = 0$ has a unique solution from given initial data, and the setting in which most rigorous theorems about gravitational waves and black-hole evolution are stated. Minkowski spacetime is globally hyperbolic. The exterior of a Schwarzschild black hole is globally hyperbolic. FLRW cosmology is globally hyperbolic. Anti-de Sitter is not — and that failure is one reason why holography in AdS needs boundary conditions imposed by hand.

The deep theorem of the subject is Geroch's: a spacetime is globally hyperbolic if and only if it has a Cauchy hypersurface. The slick refinement, due to Bernal and Sánchez, says further that the spacetime then splits as $M ≅ R \times Σ$ with the time direction extracted cleanly from the geometry. Spacetime becomes " $Σ$ evolving through time," the way Newton always wanted it to look — but extracted from the geometry itself, not imposed by hand.

This unit sets up the geometric apparatus that the rest of the microlocal-QFT chapter depends on. Without global hyperbolicity, the wave equation has no well-defined Cauchy problem, quantisation has no preferred phase space, and the Hadamard-state programme that gives mathematical sense to particle creation, the Unruh effect, and Hawking radiation cannot get started.

Visual Beginner

The cleanest picture is a Cauchy slice with its causal future and past. Imagine spacetime as a thick column. At each point, draw the light cone. A Cauchy hypersurface $Σ$ is a slice across the column with the property that every causal worldline — every timelike or null curve — extended as far as it can go in either direction, crosses $Σ$ exactly once.

The set of events you can reach from $Σ$ by future-directed causal curves is the causal future $J^{+} (Σ)$ . The set you can reach by past-directed causal curves is $J^{-} (Σ)$ . Global hyperbolicity says $M = J^{+} (Σ) \cup J^{-} (Σ)$ — the whole spacetime is the union of $Σ$ 's causal future and past, with $Σ$ as the common boundary.

The right inset of the picture shows the Bernal-Sánchez splitting: once $Σ$ exists, the spacetime decomposes as $R \times Σ$ with a smooth time function $t$ and a lapse $β$ . Each level set $Σ_{t} = {t} \times Σ$ is itself a Cauchy hypersurface; the family ${Σ_{t}}$ is a foliation of $M$ by Cauchy slices. This is the geometric meaning of "spacetime as $Σ$ evolving through time."

Worked example Beginner

Take Minkowski spacetime $R^{1, 3}$ in standard coordinates $(t, x, y, z)$ with metric $d s^{2} = - d t^{2} + d x^{2} + d y^{2} + d z^{2}$ (units $c = 1$ ).

Step 1. The hyperplane $Σ_{0} = {t = 0}$ is a smooth 3-dimensional submanifold of $R^{1, 3}$ , given by $R^{3}$ with coordinates $(x, y, z)$ .

Step 2. The restriction of the metric to $Σ_{0}$ — that is, the metric you see when you only move along $Σ_{0}$ , so $d t = 0$ — is $d s^{2} ∣_{Σ_{0}} = d x^{2} + d y^{2} + d z^{2}$ , the standard Riemannian metric on $R^{3}$ . Positive-definite. So $Σ_{0}$ is a spacelike hypersurface.

Step 3. Check that every inextendible timelike curve crosses $Σ_{0}$ exactly once. A timelike curve in Minkowski is a smooth $γ (τ) = (t (τ), x (τ))$ with $\dot{t} > ∣ \dot{x} ∣$ , so $t$ is strictly increasing along $γ$ . As $τ$ ranges over the maximal interval of definition, $t (τ)$ ranges over an open interval; if the curve is inextendible past Minkowski, that interval is all of $R$ . So $t (τ) = 0$ at exactly one value of $τ$ — the curve crosses $Σ_{0}$ once.

Step 4. The causal future of any point $p = (t_{0}, x_{0})$ is the upper light cone $J^{+} (p) = {(t, x) : t - t_{0} \geq ∣ x - x_{0} ∣}$ . The causal future of $Σ_{0}$ is the union of these cones over all $p \in Σ_{0}$ , which simplifies to ${t \geq 0}$ . Similarly $J^{-} (Σ_{0}) = {t \leq 0}$ . Their union is all of $R^{1, 3}$ , with $Σ_{0}$ the common boundary.

Step 5. The Bernal-Sánchez splitting is direct: $R^{1, 3} = R \times R^{3}$ with $t$ the first factor, $β = 1$ everywhere, and $h_{t} = d x^{2} + d y^{2} + d z^{2}$ independent of $t$ . The lapse is constant; the spatial metric is the same flat metric on every slice. Minkowski is the simplest case of the splitting.

What this tells us: Minkowski spacetime is globally hyperbolic with the simplest possible Cauchy slicing — flat constant-time hyperplanes, all isomorphic, related by Minkowski time translation. Every other globally hyperbolic spacetime is a deformation of this picture. Schwarzschild outside the horizon, FLRW cosmology, gravitational waves on a flat background — all admit foliations by Cauchy hypersurfaces, with $β$ and $h_{t}$ now functions of position and time, but the topological backbone $R \times Σ$ unchanged.

Check your understanding Beginner

Exercise (easy, multiple choice).

A Cauchy hypersurface $Σ$ of a globally hyperbolic spacetime $(M, g)$ is a set on which

A. The metric $g$ vanishes B. Every inextendible timelike curve in $M$ intersects $Σ$ exactly once C. The Einstein equations $G_{μν} = 8 π G T_{μν}$ are imposed D. The Riemann curvature tensor is zero

Hint

Think about Cauchy surfaces as initial-data slices: every possible particle history must register its starting point on the slice.

Answer

B. Every inextendible timelike curve in $M$ intersects $Σ$ exactly once.

A Cauchy hypersurface is defined by the unique-intersection property for inextendible causal worldlines. Specifying initial data on $Σ$ then determines a solution of the wave equation throughout $M$ . Option A is wrong — the metric is non-degenerate everywhere. Option C describes the dynamical content of GR, not the geometric notion of a Cauchy surface. Option D would describe flat (Minkowski) spacetime, not the general definition.

Exercise (medium, multiple choice).

The Bernal-Sánchez 2005 theorem refines Geroch's 1970 theorem by showing that a globally hyperbolic spacetime admits

A. A unique Cauchy hypersurface B. A continuous Cauchy hypersurface, but possibly only one C. A smooth spacelike Cauchy hypersurface and a smooth metric splitting $M ≅ R \times Σ$ with $g = - β^{2} d t^{2} + h_{t}$ D. A flat Cauchy hypersurface

Hint

Geroch's original theorem gave the existence of a (continuous) Cauchy surface. The Bernal-Sánchez improvement supplies smoothness and the metric splitting.

Answer

C. A smooth spacelike Cauchy hypersurface and a smooth metric splitting $M ≅ R \times Σ$ with $g = - β^{2} d t^{2} + h_{t}$ .

Geroch 1970 proved that global hyperbolicity is equivalent to the existence of a continuous Cauchy hypersurface and gave a continuous splitting. Bernal and Sánchez improved this in two papers (2003 Comm. Math. Phys. 243; 2005 Comm. Math. Phys. 257) to a smooth Cauchy hypersurface and a smooth metric splitting. The result removed a long-standing assumption in mathematical GR — until 2003 it was simply asserted that a smooth slicing existed, without proof. Option A is false: every globally hyperbolic spacetime admits infinitely many Cauchy hypersurfaces (any spacelike level set of the time function is one). Option D is false: the Cauchy surface inherits whatever Riemannian metric the spacetime imposes, which need not be flat.

Exercise (medium, numeric).

In Minkowski spacetime $R^{1, 3}$ , the causal future of the origin $p = (0, 0, 0, 0)$ is the upper light cone $J^{+} (p) = {(t, x, y, z) : t \geq x^{2} + y^{2} + z^{2}}$ . Consider the event $q = (5, 3, 4, 0)$ (in units where $c = 1$ ). Is $q$ in the causal future of $p$ ? Give a yes/no answer and a one-line justification with the numerical value of the squared interval $(Δ t)^{2} - (Δ x)^{2}$ .

Hint

A point $q$ is in $J^{+} (p)$ iff $Δ t \geq ∣Δ x ∣$ (timelike or null separation in the forward direction). Compute $∣Δ x ∣$ from the spatial separation.

Answer

Yes; squared interval is $0$ .

Spatial separation: $∣Δ x ∣ = 3^{2} + 4^{2} + 0^{2} = 5$ . Time separation: $Δ t = 5$ . Squared Minkowski interval: $(Δ t)^{2} - ∣Δ x ∣^{2} = 25 - 25 = 0$ . The event $q$ lies on the future null cone of $p$ — it is reachable from $p$ by a light ray. So $q \in J^{+} (p)$ on the boundary (a null worldline, not a timelike one).

Formal definition Intermediate+

A Lorentzian manifold is a pair $(M, g)$ consisting of a smooth $n$ -manifold $M$ ( $n \geq 2$ ) and a smooth symmetric non-degenerate $(0, 2)$ -tensor field $g$ — a metric — whose signature at every point is $(-, +, \dots, +)$ . For relativistic spacetimes $n = 4$ ; the constructions of this unit go through for general $n$ . This unit adopts the mostly-plus signature convention $(-, +, +, +)$ , matching Hawking-Ellis, Wald, Bär-Ginoux-Pfäffle, and Gérard 2019. The opposite mostly-minus convention $(+, -, -, -)$ used in much of the particle-physics literature flips the sign of $g$ and the sign of the d'Alembertian; the geometric content is identical.

Non-zero tangent vectors $v \in T_{p} M$ at a point $p$ are classified by the sign of $g_{p} (v, v)$ :

g_{p} (v, v) < 0 (timelike), g_{p} (v, v) = 0 (null, or lightlike), g_{p} (v, v) > 0 (spacelike) .

The set of null vectors at $p$ forms the null cone $N_{p} \subset T_{p} M$ , a double cone separating timelike vectors (inside) from spacelike vectors (outside). A tangent vector is causal if it is timelike or null and non-zero. A piecewise-smooth curve $γ : I \to M$ is timelike (resp. null, causal, spacelike) if its tangent vector is timelike (resp. null, causal, spacelike) at every regular point.

A time orientation on $(M, g)$ is a continuous choice, at every point, of one of the two connected components of $N_{p} ∖ {0}$ , labelled "future-directed." Equivalently, a smooth nowhere-zero timelike vector field $T$ on $M$ singles out the component containing $T_{p}$ . A Lorentzian manifold equipped with a time orientation is time-orientable; the obstruction to time-orientability is a non-orientable real line bundle obtained by parallel-transporting the null cone around closed loops. All the spacetimes considered here are time-orientable (and the time orientation is fixed once and for all).

Given a time-oriented Lorentzian manifold $(M, g)$ and a point $p \in M$ :

The chronological future $I^{+} (p) := {q \in M : \exists future-directed timelike curve from p to q}$ .
The causal future $J^{+} (p) := {q \in M : \exists future-directed causal curve from p to q} \cup {p}$ .
The chronological past $I^{-} (p)$ and causal past $J^{-} (p)$ are defined symmetrically by reversing time orientation.

For subsets $S \subseteq M$ , $I^{\pm} (S) := ⋃_{p \in S} I^{\pm} (p)$ and similarly for $J^{\pm} (S)$ . By construction $I^{+} (p) \subseteq J^{+} (p)$ and $I^{+} (p)$ is open in $M$ , while $J^{+} (p)$ need not be closed in general — its failure to be closed is a manifestation of the conformal-boundary pathology that distinguishes anti-de Sitter from Minkowski.

A spacetime $(M, g)$ satisfies the strong causality condition at a point $p$ if every neighbourhood $U ∋ p$ contains a smaller neighbourhood $V ∋ p$ such that no causal curve starting in $V$ leaves $U$ and then re-enters it. Equivalently, the topology of $M$ has a basis of "causally convex" open sets at every point. $(M, g)$ is strongly causal if it is strongly causal at every point. Strong causality forbids closed causal curves and the "almost closed" curves that come arbitrarily close to closing.

The central definition of the unit:

Definition (global hyperbolicity). A time-oriented Lorentzian manifold $(M, g)$ is globally hyperbolic if (1) it is strongly causal and (2) for every pair of points $p, q \in M$ , the set $J^{+} (p) \cap J^{-} (q)$ is compact.

A subset $Σ \subset M$ is a Cauchy hypersurface if every inextendible timelike curve in $M$ intersects $Σ$ at exactly one point. A Cauchy hypersurface is automatically a topological codimension-1 submanifold and is achronal (no two points of $Σ$ are connected by a timelike curve).

Counterexamples to common slips

A spacetime can be strongly causal without being globally hyperbolic: anti-de Sitter is strongly causal, but $J^{+} (p) \cap J^{-} (q)$ need not be compact because causal geodesics reach the timelike conformal boundary at infinity in finite affine parameter.
The causal future $J^{+} (p)$ is not in general closed. In Minkowski it is closed, but in spacetimes with timelike boundary or where causal geodesics escape to infinity, $J^{+} (p)$ may fail to be closed; one corrects this by intersecting with $J^{-} (q)$ for some $q$ in the future of $p$ , which under global hyperbolicity restores compactness.
A Cauchy hypersurface need not be smooth, even in a smooth spacetime — Geroch 1970 only secured the topological / continuous version. The Bernal-Sánchez 2003 / 2005 refinement made smoothness a theorem rather than an assumption.
"Achronal" is weaker than "spacelike." A null hypersurface (e.g. a light cone) is achronal but not spacelike. A Cauchy hypersurface is required to be achronal; the Bernal-Sánchez theorem says one can choose it spacelike, but the definition admits null pieces.
A globally hyperbolic spacetime need not be geodesically complete. Schwarzschild's exterior is globally hyperbolic but radial timelike geodesics fall through the horizon in finite proper time and continue into the interior, which lies outside the exterior region.

Key theorem with proof Intermediate+

Theorem (Geroch 1970; Bernal-Sánchez 2003, 2005). Let $(M, g)$ be a connected time-oriented Lorentzian manifold. The following are equivalent.

$(M, g)$ is globally hyperbolic.
$(M, g)$ admits a Cauchy hypersurface.
$(M, g)$ admits a smooth spacelike Cauchy hypersurface $Σ$ , and there exists a smooth diffeomorphism $Ψ : R \times Σ \sim M$ such that each slice $Ψ ({t} \times Σ)$ is a smooth spacelike Cauchy hypersurface, and in the coordinates pulled back from $R \times Σ$ the metric takes the form $g = - β^{2} d t^{2} + h_{t}$ , where $β : M \to R_{> 0}$ is a smooth positive function (the lapse) and $h_{t}$ is a smoothly $t$ -varying Riemannian metric on $Σ$ .

The equivalence $(1) \Leftrightarrow (2)$ is Geroch's theorem ^{[Geroch 1970]}. The strengthening to the smooth statement $(3)$ is the Bernal-Sánchez theorem ^{[Bernal-Sánchez 2003 + 2005]}.

Proof of $(2) \Rightarrow (1)$ . Suppose $(M, g)$ admits a Cauchy hypersurface $Σ$ . Strong causality follows: if there were a closed or almost-closed causal curve $γ$ , the unique-intersection property of $Σ$ would be violated either by $γ$ meeting $Σ$ twice or by a perturbation of $γ$ doing so. For compactness of $J^{+} (p) \cap J^{-} (q)$ , one shows that any sequence in $J^{+} (p) \cap J^{-} (q)$ has a limit point, using the fact that the Cauchy hypersurface partitions $M$ into past and future and that the intersection $J^{+} (p) \cap J^{-} (q)$ lies in a sandwich between two compact subsets of Cauchy slices.

Proof of $(1) \Rightarrow (2)$ (Geroch's construction). The hard direction. Assume $(M, g)$ is globally hyperbolic. The strategy is to construct a continuous time function $τ : M \to R$ — a function strictly increasing along every future-directed causal curve — whose level sets $τ^{- 1} (c)$ are Cauchy hypersurfaces.

Pick a probability measure $μ$ on $M$ in the same measure class as a fixed volume form. For each point $p \in M$ define $$ \tau(p) := \log \frac{\mu(I^-(p))}{\mu(I^+(p))}. $$ The function $τ$ is well-defined because $μ (I^{\pm} (p))$ are positive and finite (the latter because global hyperbolicity implies $I^{\pm} (p)$ has compact closure intersection with any $I^{-} (q) \cap I^{+} (r)$ , and one chooses $μ$ with appropriate decay). Strong causality and the compactness condition together force $τ$ to be continuous and strictly increasing along future-directed causal curves. The level set $Σ_{0} := τ^{- 1} (0)$ is then shown to be a Cauchy hypersurface: every inextendible timelike curve $γ$ has $τ \circ γ$ strictly increasing and ranging over all of $R$ , so it crosses $τ = 0$ exactly once.

Sketch of $(2) \Rightarrow (3)$ (Bernal-Sánchez). Geroch's time function $τ$ is only continuous. Bernal-Sánchez showed in 2003 ^{[Bernal-Sánchez 2003]} that it can be smoothed to a smooth time function whose level sets are smooth spacelike hypersurfaces, by a Morse-theoretic / convolution argument: convolve $τ$ against a smooth bump in a Cauchy-development neighbourhood, perturb to remove critical points, and verify that the resulting smooth function still has every level set a Cauchy hypersurface. In 2005 ^{[Bernal-Sánchez 2005]} they extended this to produce the metric splitting: choose orthogonal coordinates adapted to the foliation ${Σ_{t}}_{t}$ , identify $β$ as the reciprocal magnitude of the gradient $d τ$ (so that $τ$ is proper time along the integral curves of the future-directed unit normal), and pull back the spatial metric to obtain $h_{t}$ .

Proof of $(3) \Rightarrow (2)$ . Immediate: the slice $Σ_{0} = Ψ ({0} \times Σ)$ is a Cauchy hypersurface by construction. $■$

Bridge. The Geroch-Bernal-Sánchez theorem builds toward 13.09.02, where the well-posedness of the Klein-Gordon Cauchy problem $(□_{g} + m^{2}) ϕ = 0$ with data on a Cauchy hypersurface $Σ$ becomes the basic existence theorem of curved-spacetime field theory, and appears again in 13.09.03, where the wave-front-set characterisation of Hadamard states (Radzikowski 1996) is stated on a globally hyperbolic background. The foundational reason this geometric condition is the right input to QFT is that compactness of $J^{+} (p) \cap J^{-} (q)$ together with the smooth metric splitting $g = - β^{2} d t^{2} + h_{t}$ is exactly what makes the principal symbol of $□_{g} + m^{2}$ a strictly hyperbolic differential operator with well-posed Cauchy problem and finite-speed-of-propagation Green's functions $E^{\pm}$ . This is exactly the bridge between Lorentzian geometry and microlocal analysis: a global condition on the manifold (existence of a Cauchy hypersurface) is equivalent to a local-PDE condition on the wave operator (well-posed initial-value problem). Putting these together with the time-orientation choice, the FNW deformation argument ^{[Hawking-Ellis 1973]} then produces Hadamard states on every globally hyperbolic spacetime. The pattern generalises through the Cauchy-horizon-instability literature (Christodoulou 1994; Dafermos 2003) and connects to the cosmic-censorship conjecture 13.06.01, where the question of whether generic gravitational collapse stays inside the globally hyperbolic regime is the central open problem of mathematical relativity.

Exercises Intermediate+

Exercise 2 (easy, multiple choice).

A spacetime $(M, g)$ that is strongly causal but for which $J^{+} (p) \cap J^{-} (q)$ fails to be compact for some $p, q \in M$ is

A. Globally hyperbolic B. Not globally hyperbolic C. Closed-timelike-curve-permitting D. Necessarily anti-de Sitter

Hint

Read the definition of global hyperbolicity carefully: both conditions are required.

Answer

B. Not globally hyperbolic.

Both conditions in the definition (strong causality and compactness of $J^{+} (p) \cap J^{-} (q)$ for all $p, q$ ) are required. Strong causality alone does not suffice. Anti-de Sitter is an example: strongly causal but with non-compact $J^{+} \cap J^{-}$ because causal geodesics escape to the timelike conformal boundary. Option C is wrong because strong causality already forbids closed timelike curves. Option D is wrong because many spacetimes share the failure mode, not only anti-de Sitter.

Exercise 3 (medium, symbolic).

Show that the chronological future $I^{+} (p)$ of any point $p$ in a Lorentzian manifold is an open subset of $M$ .

Hint

Use the fact that smooth timelike curves have an open neighbourhood of perturbed timelike curves with the same endpoints up to small variation. Equivalently, use that timelike-ness is an open condition on tangent vectors.

Answer

Let $q \in I^{+} (p)$ , so there is a smooth future-directed timelike curve $γ : [0, 1] \to M$ with $γ (0) = p$ and $γ (1) = q$ . Pick a coordinate chart $(U, ϕ)$ around $q$ . Within $U$ the metric $g$ varies continuously, and the set of future-directed timelike vectors at each point is an open cone in the tangent space. The curve $γ$ enters $U$ at some time $t_{0} < 1$ ; for any $q^{'} \in U$ sufficiently close to $q$ , the straight-line segment in coordinates from $γ (t_{0})$ to $q^{'}$ is a smooth perturbation of the segment from $γ (t_{0})$ to $q$ , and remains timelike for $q^{'}$ in some open neighbourhood of $q$ . Concatenating $γ ∣_{[0, t_{0}]}$ with this perturbed segment gives a future-directed timelike curve from $p$ to $q^{'}$ . So a neighbourhood of $q$ lies in $I^{+} (p)$ . Since $q \in I^{+} (p)$ was arbitrary, $I^{+} (p)$ is open.

Exercise 5 (medium, symbolic).

Show that the Schwarzschild exterior region $M_{ext} = R \times (r_{s}, \infty) \times S^{2}$ with the Schwarzschild metric is globally hyperbolic. Identify a Cauchy hypersurface.

Hint

Consider the level sets of the Schwarzschild coordinate $t$ . Verify (a) $Σ_{0} = {t = 0}$ is spacelike for $r > r_{s}$ ; (b) every inextendible timelike curve in $M_{ext}$ crosses $Σ_{0}$ exactly once. The second part uses the staticity of the Schwarzschild metric and the fact that the timelike Killing vector $\partial_{t}$ generates time translations.

Answer

The Schwarzschild metric for $r > r_{s}$ is $$ ds^2 = -\Big(1 - \tfrac{r_s}{r}\Big) dt^2 + \Big(1 - \tfrac{r_s}{r}\Big)^{-1} dr^2 + r^2 d\Omega^2. $$ On $Σ_{0} = {t = 0}$ the induced metric is $d s^{2} ∣_{Σ_{0}} = (1 - r_{s} / r)^{- 1} d r^{2} + r^{2} d Ω^{2}$ , positive-definite for $r > r_{s}$ . So $Σ_{0}$ is spacelike.

For the unique-crossing property: a future-directed timelike curve $γ (λ) = (t (λ), r (λ), θ (λ), φ (λ))$ in $M_{ext}$ has $\dot{t} > 0$ everywhere (from the timelike condition and time-orientability with $\partial_{t}$ future-directed for $r > r_{s}$ ). So $t (λ)$ is strictly increasing. An inextendible timelike curve in $M_{ext}$ has either (i) $t (λ) \to \pm \infty$ at both ends — in which case it crosses every $t = c$ exactly once; or (ii) approaches the horizon $r = r_{s}$ in finite parameter, in which case the curve leaves $M_{ext}$ and is not inextendible within $M_{ext}$ . In either case the inextendible-curve unique-crossing requirement of the Cauchy-surface definition is met within $M_{ext}$ .

Compactness of $J^{+} (p) \cap J^{-} (q)$ for $p, q \in M_{ext}$ follows from the static structure: any causal curve from $p$ to $q$ stays in a bounded $t$ -range $[t (p), t (q)]$ , and a Schwarzschild-radial bound on causal travel time confines the curve to a compact set bounded above and below in $r$ .

Exercise 6 (medium, symbolic).

Anti-de Sitter spacetime $AdS_{2}$ can be described as the universal cover of the hyperboloid $- T^{2} - V^{2} + X^{2} = - 1$ in $R^{2, 1}$ , with induced metric. In suitable coordinates $(t, x)$ it becomes $d s^{2} = sec^{2} x (- d t^{2} + d x^{2})$ for $x \in (- π /2, π /2)$ , $t \in R$ . Show that a null geodesic starting at the origin $(t, x) = (0, 0)$ reaches the boundary $x = π /2$ in finite affine parameter. Conclude that $AdS_{2}$ is not globally hyperbolic.

Hint

A null geodesic has $0 = d s^{2} = sec^{2} x (- d t^{2} + d x^{2})$ , so $d t = \pm d x$ . Parametrise by affine parameter and use a conformal-rescaling argument.

Answer

The null condition $d t = d x$ along an outgoing right-moving null geodesic gives the path $t = x$ in coordinates. Reparametrising by an affine parameter $λ$ — for a null geodesic in a conformally-rescaled metric $\tilde{g} = Ω^{2} g_{flat}$ with $g_{flat} = - d t^{2} + d x^{2}$ , the affine parameter scales as $d λ \propto Ω^{- 2} d s_{flat}$ — gives the affine parameter $λ$ as $\int_{0}^{x_{0}} sec^{- 2} (x) d x$ along the null geodesic from $x = 0$ to $x = x_{0}$ . This integral is finite as $x_{0} \to π / 2^{-}$ : $\int_{0}^{π /2} cos^{2} x d x = π /4 < \infty$ .

So an outgoing null geodesic reaches the conformal boundary $x = π /2$ in finite affine parameter, and similarly an incoming geodesic comes from $x = - π /2$ in finite affine parameter. Information can flow in from the boundary in finite proper time. Pick a point $p$ in the interior; $J^{-} (p)$ is unbounded toward the past boundary, and for $q$ sufficiently in the future, $J^{+} (p) \cap J^{-} (q)$ extends out toward the boundary and fails to be compact. $AdS_{2}$ is therefore not globally hyperbolic. Field theory on $AdS$ requires boundary conditions imposed by hand — the central technical point of the AdS/CFT dictionary.

Exercise 7 (hard, symbolic).

Let $(M, g)$ be a globally hyperbolic spacetime and let $Σ_{1}, Σ_{2}$ be two smooth spacelike Cauchy hypersurfaces with $Σ_{2} \subset J^{+} (Σ_{1})$ . Prove that $Σ_{1}$ and $Σ_{2}$ are diffeomorphic.

Hint

Use the Bernal-Sánchez splitting $M ≅ R \times Σ$ to flow $Σ_{1}$ onto $Σ_{2}$ along integral curves of the future-directed unit normal to the foliation. Show this flow is a diffeomorphism.

Answer

By Bernal-Sánchez 2005 there is a smooth Cauchy time function $t : M \to R$ and a smooth diffeomorphism $Ψ : R \times Σ \sim M$ with $t \circ Ψ = π_{1}$ (the projection to the first factor) and $Ψ ({c} \times Σ)$ a Cauchy hypersurface for each $c \in R$ .

For our given Cauchy surfaces $Σ_{1}, Σ_{2}$ , the time function $t$ need not be constant on either, but a refinement exists. Choose $t_{1}$ constant on $Σ_{1}$ and let $T$ be the gradient flow of $t$ (a smooth complete vector field by the metric splitting form). The flow $Φ_{s}^{T} : M \to M$ moves Cauchy slices to Cauchy slices: $Φ_{s}^{T} (Σ_{1})$ is a Cauchy slice for each $s$ .

For any $p \in Σ_{2}$ , the integral curve of $T$ through $p$ extends both directions and crosses $Σ_{1}$ at exactly one point $π (p) \in Σ_{1}$ (by the Cauchy-surface property of $Σ_{1}$ applied to the timelike integral curve). The map $π : Σ_{2} \to Σ_{1}$ is smooth and bijective. Smoothness of the inverse comes from the smooth foliation: for $q \in Σ_{1}$ , the integral curve of $T$ from $q$ runs into $Σ_{2}$ at a unique point, by the Cauchy-surface property of $Σ_{2}$ . The composition is the identity on each side, so $π$ is a diffeomorphism.

Topologically: all Cauchy slices of a globally hyperbolic spacetime are diffeomorphic to each other. The topology of the slice is a global invariant of the spacetime.

Exercise 8 (hard, symbolic).

Let $(M, g)$ be a globally hyperbolic spacetime and $Σ \subset M$ a Cauchy hypersurface. Show that $M = J^{+} (Σ) \cup J^{-} (Σ)$ with $J^{+} (Σ) \cap J^{-} (Σ) = Σ$ .

Hint

For any $p \in M$ , consider an inextendible timelike curve through $p$ . It must cross $Σ$ at exactly one point. Use the direction of the crossing to assign $p$ to $J^{+} (Σ)$ or $J^{-} (Σ)$ .

Answer

Cover. For $p \in M$ , pick any future-directed timelike curve through $p$ . Extend it past $p$ both ways to an inextendible timelike curve $γ$ . By the Cauchy-surface property of $Σ$ , $γ$ crosses $Σ$ at exactly one point $q$ . If $q$ is at or before $p$ along $γ$ , then there is a future-directed timelike curve from $q$ to $p$ , so $p \in I^{+} (Σ) \subseteq J^{+} (Σ)$ . Symmetrically, if $q$ is at or after $p$ , then $p \in J^{-} (Σ)$ . Either way $p \in J^{+} (Σ) \cup J^{-} (Σ)$ .

Boundary. If $p \in J^{+} (Σ) \cap J^{-} (Σ)$ , there are future-directed and past-directed causal curves $γ_{1}, γ_{2}$ from points of $Σ$ to $p$ and from $p$ to points of $Σ$ respectively. Concatenating, the curve $γ_{2} \cdot γ_{1}^{- 1}$ is a future-directed causal curve from $Σ$ to $Σ$ passing through $p$ . If $p \in / Σ$ , the concatenation traverses $Σ$ at least twice (going up through $Σ$ at the start and again at $p$ or after). The unique-intersection property of $Σ$ forces $p \in Σ$ — so $J^{+} (Σ) \cap J^{-} (Σ) \subseteq Σ$ . The reverse inclusion is immediate.

Exercise 9 (hard, symbolic).

Gödel's 1949 stationary rotating cosmological solution has metric (in suitable coordinates with $a > 0$ ) $$ ds^2 = -dt^2 + dx^2 - \tfrac{1}{2} e^{2ax} dy^2 + dz^2 + 2 e^{ax} dt, dy. $$ Show that the closed curves $γ_{r} : [0, 2 π] \to M$ , $γ_{r} (σ) = (0, x_{0}, r cos σ, r sin σ, 0)$ for appropriate radius $r$ at a fixed $x_{0}$ , become timelike for $r$ sufficiently large. Conclude that Gödel spacetime is not strongly causal and therefore not globally hyperbolic.

Hint

Compute $g (\overset{γ}{˙}_{r}, \overset{γ}{˙}_{r})$ where $\overset{γ}{˙}_{r} = (- r sin σ, 0, 0, 0, r cos σ) \cdot \partial / \dots$ — wait, parametrise more carefully. Consider closed circles in the $(y, z)$ plane and examine when the $- \frac{1}{2} e^{2 a x} d y^{2}$ contribution dominates.

Answer

Parametrise the closed curve as $γ : [0, 2 π] \to M$ by $γ (σ) = (0, x_{0}, r cos σ, r sin σ, 0)$ in coordinates $(t, x, y, z)$ . Wait — $z$ should be the fifth coordinate; here is the correct Gödel curve. Take the closed circle $σ \mapsto (0, x_{0}, r cos σ, 0, r sin σ)$ (revising indices: $t = 0, x = x_{0}, y = r cos σ, t^{'} = 0, z = r sin σ$ , so the curve lies in the $(y, z)$ plane and is closed).

Tangent vector: $\overset{γ}{˙} = (- r sin σ) \partial_{y} + (r cos σ) \partial_{z}$ . Computing the squared norm: $$ g(\dot\gamma, \dot\gamma) = g_{yy} (r\sin\sigma)^2 + g_{zz} (r\cos\sigma)^2 = -\tfrac{1}{2} e^{2ax_0}(r\sin\sigma)^2 + (r\cos\sigma)^2. $$ Averaged over $σ \in [0, 2 π]$ : $⟨ g (\overset{γ}{˙}, \overset{γ}{˙})⟩ = r^{2} [1 - \frac{1}{2} e^{2 a x_{0}}] /2$ . For $e^{2 a x_{0}} > 2$ (i.e. $x_{0} > \frac{l n 2}{2 a}$ ), the average is negative; for sufficiently large $r$ , the curve is timelike on average and can be deformed to a globally timelike closed curve by a small perturbation.

So Gödel spacetime contains closed timelike curves — its causality is the most pathological grade (chronology fails). It is not strongly causal and therefore not globally hyperbolic. Gödel's original 1949 Rev. Mod. Phys. 21 447 paper presented exactly this as a sharp consequence of admitting solutions of Einstein's equations with rotating-dust causal pathology.

Lean formalization Intermediate+

Mathlib has no Lorentzian-manifold infrastructure as of 2026-05. The closest layers are Geometry.Manifold.SmoothManifoldWithCorners and Geometry.Manifold.MetricSpace (smooth manifolds and positive-definite Riemannian metrics) plus LinearAlgebra.QuadraticForm and LinearAlgebra.BilinearForm.Nondegenerate (signature-aware bilinear forms at the algebraic level). The pseudo-Riemannian-metric-on-a-smooth-manifold structure, the time-orientation predicate, and the causality conditions of this unit do not yet exist as Mathlib objects.

The Lean 4 SpaceTime project (M. Larson and collaborators, separate from Mathlib) has a partial Lorentzian-metric layer and some causal-structure apparatus; the Bernal-Sánchez metric splitting and Geroch's time-function construction are not formalised in any current Lean library.

lean_status: none reflects this gap; no Lean module ships with this unit. the Mathlib gap analysis names the specific missing infrastructure. Tyler's review attests intermediate-tier correctness; the master-tier Bernal-Sánchez proof statement and the Geroch time-function construction are flagged for external Lorentzian-geometry review.

Advanced results Master

Three structural developments take the basic global-hyperbolicity framework to the depth Gérard 2019 Ch. 5 and the curved-spacetime QFT programme require.

The causal hierarchy. Causality conditions on a Lorentzian manifold form a chain of strictly nested classes ^{[Hawking-Ellis 1973 Ch. 6.4]}. From weakest to strongest: chronology (no closed timelike curves) $\supset$ causality (no closed causal curves) $\supset$ distinguishing (each point has a neighbourhood basis that distinguishes points by $I^{\pm}$ ) $\supset$ strong causality (no almost-closed causal curves) $\supset$ stable causality (existence of a global time function) $\supset$ global hyperbolicity. Each strict inclusion is exhibited by an explicit counterexample — chronology fails in Gödel; strong causality fails in Reissner-Nordström extensions past the Cauchy horizon; stable causality holds without global hyperbolicity for anti-de Sitter and Minkowski with a point removed. The Penrose 1972 lecture notes ^{[Penrose 1972]} and the Bernal-Sánchez 2007 Class. Quantum Grav. 24 745 ("Globally hyperbolic spacetimes can be defined as 'causal' instead of 'strongly causal'") show that one may weaken "strongly causal" to "causal" in the definition of global hyperbolicity without changing the class — the compactness of $J^{+} (p) \cap J^{-} (q)$ promotes causality to strong causality automatically.

Domain of dependence. For a subset $S \subset M$ in a time-oriented Lorentzian manifold, the future domain of dependence $D^{+} (S)$ is the set of points $p$ such that every past-inextendible causal curve through $p$ intersects $S$ . The past domain of dependence $D^{-} (S)$ is defined symmetrically; the full domain of dependence $D (S) := D^{+} (S) \cup D^{-} (S)$ is the set of all $p$ such that data on $S$ determines a unique solution at $p$ for any reasonable hyperbolic equation. The Cauchy-surface characterisation now reads: $Σ$ is a Cauchy hypersurface iff $D (Σ) = M$ . The Cauchy horizon $H^{+} (S) := \overline{D^{+} (S)} ∖ D^{+} (S)$ is the boundary of predictability — the surface beyond which initial data on $S$ no longer determines the solution. Reissner-Nordström and Kerr black-hole interiors contain Cauchy horizons; whether these horizons are unstable under generic perturbation is the content of the strong cosmic censorship conjecture (Penrose 1979; Christodoulou 1994 Comm. Math. Phys. 161 195 for the Maxwell-charged scalar; Dafermos-Luk 2017 arXiv:1710.01722 for Kerr).

The Bernal-Sánchez splitting in detail. The 2005 metric splitting $g = - β^{2} d t^{2} + h_{t}$ admits a sharp form. Bernal and Sánchez 2005 ^{[Bernal-Sánchez 2005]} proved that one can take $β \equiv 1$ in this splitting by reparametrising the time function: if $τ$ is any smooth Cauchy time function with gradient $d τ$ nowhere zero and $∣ d τ ∣_{g}$ smooth and bounded above and below on compact subsets of each Cauchy slice, then the rescaled function $t := \int_{0}^{τ} β (s, \cdot) d s$ has unit-lapse splitting $g = - d t^{2} + h_{t}$ . The shift from variable lapse to unit lapse is geometric — a reparametrisation of the time function — and physically natural: $t$ is proper time along the integral curves of the future-directed unit normal to the Cauchy foliation. Sánchez 2010 ("The geometric structure of conformally globally hyperbolic spacetimes") extended the splitting to the conformally globally hyperbolic case relevant to Penrose's conformal compactification programme.

The unit-lapse form $g = - d t^{2} + h_{t}$ is the cleanest setting for the Klein-Gordon Cauchy problem: in this gauge $(□_{g} + m^{2}) ϕ = 0$ reads $- \partial_{t}^{2} ϕ + \frac{1}{d e t h _{t}} \partial_{i} (det h_{t} h_{t}^{ij} \partial_{j} ϕ) + m^{2} ϕ = 0$ , which is a second-order quasilinear hyperbolic PDE in time, manifestly amenable to energy-estimate methods. The Bernal-Sánchez splitting is therefore not just a topological-product statement: it is the geometric setup in which curved-spacetime QFT inherits the well-posedness machinery of flat-space Klein-Gordon theory.

Synthesis. Global hyperbolicity is the foundational reason that classical and quantum field theory on a curved Lorentzian background acquire well-defined dynamics. The Geroch theorem identifies the geometric condition — compactness of causal diamonds plus strong causality — with the analytic condition — well-posed Cauchy problem for hyperbolic operators. The central insight of the Bernal-Sánchez refinement is that the smooth metric splitting $g = - β^{2} d t^{2} + h_{t}$ is automatic, not assumed, once a continuous Cauchy time function exists; this builds toward 13.09.02, where the Klein-Gordon Cauchy problem is solved on the splitting and the causal-propagator $E = E^{+} - E^{-}$ is constructed. The pattern recurs throughout mathematical relativity and appears again in 13.09.03, where the wave-front-set definition of Hadamard states (Radzikowski 1996) is stated on globally hyperbolic manifolds, and in 13.06.01, where the cosmic-censorship conjecture asks whether generic gravitational collapse preserves global hyperbolicity or produces Cauchy horizons whose interior breaks predictability. The bridge between Lorentzian geometry and quantum field theory passes through the existence of Cauchy slices: without them, no rigorous QFT in a curved background; with them, the entire microlocal / Hadamard programme of Gérard 2019 unfolds.

Full proof set Master

Proposition (Achronality of Cauchy hypersurfaces). Let $(M, g)$ be a time-oriented Lorentzian manifold and let $Σ \subset M$ be a Cauchy hypersurface. Then $Σ$ is achronal — no two points of $Σ$ are connected by a timelike curve.

Proof. Suppose toward contradiction that $p, q \in Σ$ with $p \neq = q$ and $q \in I^{+} (p)$ , so there is a future-directed timelike curve $γ$ from $p$ to $q$ . Extend $γ$ in both directions to an inextendible timelike curve $\tilde{γ}$ . By the Cauchy-hypersurface property, $\tilde{γ}$ intersects $Σ$ exactly once. But $\tilde{γ}$ already passes through $p \in Σ$ and through $q \in Σ$ with $p \neq = q$ , two distinct intersection points — contradiction. So no such $p, q$ exist, and $Σ$ is achronal. $□$

Proposition (Continuity of the future-causal-future map). Let $(M, g)$ be globally hyperbolic. The map $J^{+} (\cdot, \cdot) : M \times M \to Cl (M)$ sending $(p, q) \mapsto J^{+} (p) \cap J^{-} (q)$ takes values in compact subsets of $M$ , and the assignment is upper semi-continuous in the Hausdorff topology on compact subsets.

Proof (sketch). Compactness of $J^{+} (p) \cap J^{-} (q)$ is the second clause of the definition of global hyperbolicity. Upper semi-continuity in the Hausdorff metric: suppose $p_{n} \to p, q_{n} \to q$ and $x_{n} \in J^{+} (p_{n}) \cap J^{-} (q_{n})$ with $x_{n} \to x$ . Choose causal curves $γ_{n}$ from $p_{n}$ to $x_{n}$ to $q_{n}$ . The sequence ${γ_{n}}$ lies in a fixed compact set (eventually inside $J^{+} (p_{0}) \cap J^{-} (q_{0})$ for any open neighbourhood pair $(p_{0}, q_{0})$ of $(p, q)$ ). A compactness argument on causal curves (using the existence of a smooth time function $t : M \to R$ , on which $t \circ γ_{n}$ is monotone with bounded image, so by Arzelà-Ascoli a subsequence converges uniformly to a causal curve $γ$ from $p$ through $x$ to $q$ ) gives $x \in J^{+} (p) \cap J^{-} (q)$ . Hence the limit set is contained in the limit of the sets — upper semi-continuity. $□$

Proposition (Cauchy hypersurfaces are diffeomorphic). Let $(M, g)$ be globally hyperbolic with Bernal-Sánchez splitting $Ψ : R \times Σ \sim M$ . Then every smooth spacelike Cauchy hypersurface of $M$ is diffeomorphic to $Σ$ .

Proof. Exercise 7 of this unit, expanded: given two smooth spacelike Cauchy hypersurfaces $Σ_{1}, Σ_{2}$ , use the flow of the future-directed unit-normal vector field $n = β^{- 1} \partial_{t}$ (or any timelike vector field transverse to both surfaces) to define a smooth bijection $Σ_{1} \to Σ_{2}$ . Cauchy-surface uniqueness of crossing point ensures the map is well-defined and bijective; smooth dependence on the flow ensures it is a diffeomorphism. The topology of the Cauchy slice is therefore a global topological invariant of $(M, g)$ . $□$

Connections Master

Tensors on smooth manifolds 13.02.01 supplies the underlying tensorial language: a Lorentzian metric is a $(0, 2)$ -tensor field of indefinite signature, the null cone is the zero locus of $g (v, v)$ , and the time-orientation is a section of a real line bundle constructed from the metric. The metric splitting $g = - β^{2} d t^{2} + h_{t}$ is a tensor-field decomposition adapted to the Cauchy foliation.
Geodesics and parallel transport 13.02.02 provides the integration of the causal structure: causal geodesics are the curves whose tangent vectors lie inside or on the null cone at every point, and the existence of causal geodesics between causally-related points is a basic question for which global hyperbolicity supplies an existence-and-uniqueness apparatus (Avez-Seifert theorem: every globally hyperbolic spacetime contains a maximising causal geodesic between any pair of causally related points).
Schwarzschild solution 13.05.01 is the canonical non-flat globally hyperbolic spacetime: the exterior region is globally hyperbolic with $t = const$ Cauchy hypersurfaces, and the maximal Kruskal extension is globally hyperbolic on the union of regions I and II. The cosmic-censorship conjecture, which asks whether realistic collapse always produces black holes whose exterior remains globally hyperbolic, is the central open question of mathematical relativity.
Einstein field equations 13.04.01 govern the time evolution of the metric on globally hyperbolic spacetimes. Choquet-Bruhat 1952 ^{[Choquet-Bruhat 1952]} proved local well-posedness of the Cauchy problem for the vacuum Einstein equations starting from initial data on a Cauchy hypersurface; the global structure of the resulting maximal Cauchy development is governed by global hyperbolicity of the resulting spacetime. The Bernal-Sánchez splitting underlies the ADM formulation of GR as a Hamiltonian initial-value problem.
Klein-Gordon equation on a globally hyperbolic spacetime [13.09.02, pending] is the immediate downstream unit. Well-posedness of the Cauchy problem $(□_{g} + m^{2}) ϕ = 0$ with data $(ϕ_{0}, ϕ_{1}) \in C_{c}^{\infty} (Σ) \oplus C_{c}^{\infty} (Σ)$ rests on the Bernal-Sánchez splitting and the strict-hyperbolicity of $□_{g} + m^{2}$ in the resulting gauge; the advanced and retarded fundamental solutions $E^{\pm}$ have supports contained in $J^{\pm} (supp)$ thanks to finite-speed-of-propagation, the analytic shadow of compactness of $J^{+} (p) \cap J^{-} (q)$ .
Hadamard states via wave-front-set criterion [13.09.03, pending] uses the globally hyperbolic background to formulate Radzikowski's wave-front-set definition of Hadamard quasi-free states. The Cauchy surface gives a canonical symplectic phase space $(S, σ) = (C_{c}^{\infty} (Σ) \oplus C_{c}^{\infty} (Σ), σ)$ for the CCR algebra; the choice of Hadamard state is then determined by the microlocal singularity structure of the two-point function on $M \times M$ .
Cosmology — FLRW models [13.08.01, pending] are spatially homogeneous and isotropic globally hyperbolic spacetimes. The Cauchy hypersurfaces are constant-cosmic-time slices $Σ_{t}$ , each a Riemannian 3-manifold of constant curvature (flat, spherical, or hyperbolic). The Bernal-Sánchez splitting is realised explicitly by the FLRW metric $g = - d t^{2} + a (t)^{2} h$ .
Black holes [13.06.01, pending] raise the cosmic-censorship questions: the Reissner-Nordström and Kerr interiors contain Cauchy horizons past which strong causality and global hyperbolicity fail. Whether generic gravitational collapse produces stably globally hyperbolic spacetimes — the strong cosmic-censorship conjecture — is undecided.

Historical & philosophical context Master

The notion that causality on a Lorentzian manifold imposes a global topological constraint emerged in stages. Jean Leray in 1953 ^{[Leray 1953]}, in mimeographed lecture notes from the Institute for Advanced Study in Princeton, introduced the notion of a causal domain for hyperbolic systems on a Lorentzian manifold — the local form of what later became global hyperbolicity. Leray's notes were never formally published but circulated widely; Choquet-Bruhat's 1952 Acta Math. paper ^{[Choquet-Bruhat 1952]} proved the local existence and uniqueness of solutions to the vacuum Einstein equations on small Lorentzian neighbourhoods, building on the same hyperbolic-PDE framework.

Robert Geroch's 1970 J. Math. Phys. paper ^{[Geroch 1970]}, "Domain of dependence," made global hyperbolicity the central organising concept of causal structure. Geroch defined a Cauchy hypersurface, proved its existence in any globally hyperbolic spacetime via the measure-theoretic time-function construction outlined in the proof above, and showed that the maximal Cauchy development of given initial data is globally hyperbolic — establishing the central existence-uniqueness theorem for the Einstein vacuum equations as a problem in Lorentzian geometry. Stephen Hawking and George Ellis's 1973 monograph The Large Scale Structure of Space-Time ^{[Hawking-Ellis 1973]} consolidated the causal-structure theory and applied it to the singularity theorems (Penrose 1965; Hawking 1966; Hawking-Penrose 1970) that established geodesic incompleteness as a generic feature of gravitational collapse and cosmological initial singularities. Roger Penrose's 1972 CBMS lecture notes Techniques of Differential Topology in Relativity ^{[Penrose 1972]} gave the modern presentation of the causality hierarchy and the conformal-compactification programme.

The smoothness of the Cauchy hypersurface and the metric splitting were, until 2003, simply asserted on physical grounds. Hawking-Ellis and Wald both proved their global-hyperbolicity theorems at the continuous level. Antonio Bernal and Miguel Sánchez in two papers, Comm. Math. Phys. 243 (2003) 461 ^{[Bernal-Sánchez 2003]} and Comm. Math. Phys. 257 (2005) 43 ^{[Bernal-Sánchez 2005]}, closed this gap: they constructed a smooth Cauchy time function whose level sets are smooth spacelike Cauchy hypersurfaces, and they extracted the smooth metric splitting $g = - β^{2} d t^{2} + h_{t}$ as a consequence. The improvement matters technically — the well-posedness theory for the Klein-Gordon and Maxwell equations on a curved background, the BFK 1996 / Hollands-Wald 2001 renormalisation programme, and the microlocal-analytic programme of Gérard 2019 all assume smooth slicings without circular reference. Bernal-Sánchez 2007 Class. Quantum Grav. 24 745 further showed that "strongly causal" in the definition of global hyperbolicity can be relaxed to "causal" — the compactness condition automatically promotes causality to strong causality.

The Geroch-Bernal-Sánchez framework supplies the geometric arena for algebraic quantum field theory on curved spacetimes (Brunetti-Fredenhagen-Verch 2003 Comm. Math. Phys. 237 31), where each globally hyperbolic spacetime corresponds to a $C^{*}$ -algebra of observables and the assignment is functorial under isometric embeddings of globally hyperbolic regions. Gérard 2019 ^{[Gérard 2019 Ch. 5]} is the canonical modern textbook entry to this material. The setting is the indispensable geometric prerequisite to the entire microlocal-analytic study of quantum fields on curved backgrounds initiated by Radzikowski 1996 (the wave-front-set definition of Hadamard states) and extended through the modern Hollands-Wald renormalisation programme.

Bibliography Master

Primary literature:

Geroch, R., "Domain of dependence", J. Math. Phys. 11 (1970), 437-449. [Originating paper for the global-hyperbolicity equivalence with existence of a Cauchy surface.]
Bernal, A. N. & Sánchez, M., "On smooth Cauchy hypersurfaces and Geroch's splitting theorem", Comm. Math. Phys. 243 (2003), 461-470. [Smooth Cauchy hypersurface theorem.]
Bernal, A. N. & Sánchez, M., "Smoothness of time functions and the metric splitting of globally hyperbolic spacetimes", Comm. Math. Phys. 257 (2005), 43-50. [Smooth metric-splitting theorem $g = - β^{2} d t^{2} + h_{t}$ .]
Bernal, A. N. & Sánchez, M., "Globally hyperbolic spacetimes can be defined as 'causal' instead of 'strongly causal'", Class. Quantum Grav. 24 (2007), 745-749.
Leray, J., Hyperbolic differential equations, Institute for Advanced Study, Princeton (1953, mimeographed lecture notes). [Originating notion of a causal domain.]
Choquet-Bruhat, Y., "Théorème d'existence pour certains systèmes d'équations aux dérivées partielles non linéaires", Acta Math. 88 (1952), 141-225. [Local well-posedness of the vacuum Einstein equations on Lorentzian neighbourhoods.]
Gödel, K., "An example of a new type of cosmological solutions of Einstein's field equations of gravitation", Rev. Mod. Phys. 21 (1949), 447-450. [Closed-timelike-curve cosmological solution; canonical chronology-violating counterexample.]

Singularity-theorem literature:

Penrose, R., "Gravitational collapse and space-time singularities", Phys. Rev. Lett. 14 (1965), 57-59.
Hawking, S. W., "The occurrence of singularities in cosmology. III. Causality and singularities", Proc. R. Soc. A 300 (1967), 187-201.
Hawking, S. W. & Penrose, R., "The singularities of gravitational collapse and cosmology", Proc. R. Soc. A 314 (1970), 529-548.

Cosmic-censorship literature:

Penrose, R., "Singularities and time-asymmetry", in General Relativity: An Einstein Centenary Survey (eds. S. W. Hawking and W. Israel, Cambridge, 1979), 581-638.
Christodoulou, D., "Examples of naked singularity formation in the gravitational collapse of a scalar field", Ann. of Math. 140 (1994), 607-653.
Dafermos, M. & Luk, J., "The interior of dynamical vacuum black holes I: The $C^{0}$ -stability of the Kerr Cauchy horizon", arXiv:1710.01722 (2017).

Modern monographs and textbooks:

Hawking, S. W. & Ellis, G. F. R., The Large Scale Structure of Space-Time (Cambridge University Press, 1973). [Canonical textbook on causal structure of spacetime; Ch. 6 is the standard reference for the causality hierarchy and the singularity theorems.]
Wald, R. M., General Relativity (University of Chicago Press, 1984). [Ch. 8 covers the causality conditions and global hyperbolicity at the modern level.]
Penrose, R., Techniques of Differential Topology in Relativity, CBMS-NSF Regional Conf. Ser. 7 (SIAM, 1972). [Causal hierarchy and conformal compactification.]
O'Neill, B., Semi-Riemannian Geometry with Applications to Relativity (Academic Press, 1983). [Standard mathematical-style textbook on pseudo-Riemannian and Lorentzian geometry; Ch. 14 covers causality.]
Beem, J. K., Ehrlich, P. E. & Easley, K. L., Global Lorentzian Geometry, 2nd ed. (CRC Press, 1996). [Comprehensive monograph on Lorentzian causal structure.]
Bär, C., Ginoux, N. & Pfäffle, F., Wave Equations on Lorentzian Manifolds and Quantization, ESI Lectures in Mathematics and Physics (EMS, 2007). [Free PDF at arXiv:0806.1036; companion classical-PDE volume for the Gérard programme.]
Gérard, C., Microlocal Analysis of Quantum Fields on Curved Spacetimes, ESI Lectures in Mathematics and Physics (EMS, 2019). [Modern textbook entry to microlocal-analytic curved-spacetime QFT; Ch. 5 covers global hyperbolicity as a geometric prerequisite to the Klein-Gordon Cauchy problem and the Hadamard-state programme.]

Sánchez expository surveys:

Sánchez, M., "Causal hierarchy of spacetimes, temporal functions and smoothness of Geroch's splitting", Mat. Contemp. 29 (2005), 127-155. [Author's review of the 2003 + 2005 results.]
Minguzzi, E. & Sánchez, M., "The causal hierarchy of spacetimes", in Recent Developments in Pseudo-Riemannian Geometry (eds. D. V. Alekseevsky and H. Baum, EMS, 2008), 299-358. [Comprehensive modern review of causality conditions.]

Prerequisites

13.02.01
13.02.02

Tier anchors

beginner: Penrose, The Road to Reality (Cape, 2004), Ch. 17 (selected sections on causal structure and Cauchy surfaces); Hartle, Gravity (2003), Ch. 4 (light cones and causal future)
intermediate: O'Neill, Semi-Riemannian Geometry (Academic, 1983), Ch. 14; Bär-Ginoux-Pfäffle, Wave Equations on Lorentzian Manifolds and Quantization (EMS, 2007), Ch. 1; Wald, General Relativity (1984), Ch. 8
master: Hawking and Ellis, The Large Scale Structure of Space-Time (Cambridge, 1973), Ch. 6; Bernal and Sánchez, Comm. Math. Phys. 257 (2005) 43; Gérard, Microlocal Analysis of Quantum Fields on Curved Spacetimes (EMS, 2019), Ch. 5

References

Hawking, S. W. & Ellis, G. F. R. — The Large Scale Structure of Space-Time (Cambridge, 1973) · Ch. 6 (Causal structure); §6.6 (Cauchy surfaces and global hyperbolicity)
Wald, R. M. — General Relativity (University of Chicago Press, 1984) · Ch. 8 (Causal structure); §8.3 (Global hyperbolicity and Cauchy surfaces)
Geroch, R. — 'Domain of dependence', J. Math. Phys. 11 (1970) 437 · Theorem 11 (global hyperbolicity ⇔ existence of a Cauchy surface)
Bernal, A. N. & Sánchez, M. — 'On smooth Cauchy hypersurfaces and Geroch's splitting theorem', Comm. Math. Phys. 243 (2003) 461 · Theorem 1 (smooth spacelike Cauchy hypersurface in any globally hyperbolic spacetime)
Bernal, A. N. & Sánchez, M. — 'Smoothness of time functions and the metric splitting of globally hyperbolic spacetimes', Comm. Math. Phys. 257 (2005) 43 · Theorem 1.1 (smooth metric splitting M ≅ ℝ × Σ with g = -β² dt² + h_t)
Gérard, C. — Microlocal Analysis of Quantum Fields on Curved Spacetimes (EMS Press, 2019) · Ch. 5 (Lorentzian manifolds and the wave equation); §5.1 (causal structure); §5.2 (globally hyperbolic spacetimes); §5.3 (Cauchy problem for the Klein-Gordon equation)
Bär, C., Ginoux, N. & Pfäffle, F. — Wave Equations on Lorentzian Manifolds and Quantization (EMS, 2007) · Ch. 1 (Preliminaries on Lorentzian manifolds); §1.3 (Cauchy hypersurfaces); free PDF at arXiv:0806.1036
O'Neill, B. — Semi-Riemannian Geometry with Applications to Relativity (Academic Press, 1983) · Ch. 14 (Causality in Lorentz manifolds)
Leray, J. — 'Hyperbolic differential equations', Institute for Advanced Study, Princeton (1953, mimeographed lecture notes) · Originating notion of a causal domain for hyperbolic systems on a Lorentzian manifold
Penrose, R. — Techniques of Differential Topology in Relativity, CBMS-NSF Regional Conf. Series 7 (SIAM, 1972) · §5 (Causality conditions); §6 (Cauchy surfaces and stable causality)
Choquet-Bruhat, Y. — 'Théorème d'existence pour certains systèmes d'équations aux dérivées partielles non linéaires', Acta Math. 88 (1952) 141 · Local well-posedness for hyperbolic systems on a Lorentzian manifold; complement to Geroch's global hyperbolicity

Estimated time

beginner: 18m
intermediate: 40m
master: 60m