The Large Scale Structure of Space-Time — Chapter-by-Chapter Outline

Author: Stephen W. Hawking and George F. R. Ellis First published: 1973 (Cambridge University Press) Edition covered: Original 1973 edition (Cambridge Monographs on Mathematical Physics, ISBN 978-0-521-09906-6). A 50th Anniversary Edition was published in February 2023 (ISBN 978-1-009-25315-4) with a new foreword by Abhay Ashtekar and a new preface by George Ellis; the chapter structure is identical to the original.

Central thesis

General relativity, when treated rigorously as a geometric theory of spacetime, predicts its own breakdown: under physically reasonable conditions — positive energy density, the absence of exotic matter, and the rough agreement of the universe with what astronomers observe — spacetime must contain singularities, points or regions where the known laws of physics cease to apply. These singularities are not artifacts of special symmetry or idealized models; they are generic, topological consequences of how gravity focuses geodesics. The two most important instances are the initial singularity that began the universe (the Big Bang) and the terminal singularities inside gravitationally collapsed stars (black holes).

The book's method is to start from the axiomatic geometry of a smooth four-dimensional Lorentzian manifold, build the physical theory of general relativity on that foundation, develop the theory of causal structure and geodesic focusing, and then deploy those tools in a sequence of singularity theorems of increasing generality. The central intellectual move is the replacement of computational, solution-by-solution reasoning with global, topological reasoning: instead of solving Einstein's equations and examining particular solutions for singularities, Hawking and Ellis show that entire classes of physically reasonable spacetimes are geodesically incomplete — regardless of their detailed geometry.

If the universe satisfies the observed conditions of expansion and positive energy density, then spacetime cannot be extended to the past without encountering a singularity — a place where the present laws of physics break down.

Chapter 1 — The Role of Gravity

Central question

Why should a book about general relativity concern itself with large-scale, global properties of spacetime, rather than local solutions to Einstein's equations?

Main argument

Physics as local laws plus boundary conditions. Hawking and Ellis open by distinguishing two aspects of any physical theory: the local differential equations governing fields (Maxwell's equations, Einstein's equations), and the global or boundary-condition question of what solutions those equations admit across an extended spacetime manifold. Most of physics concentrates on the first; this book is explicitly about the second.

Gravity's unique status. Gravity differs from every other force because it affects the causal structure of spacetime itself. Electromagnetism, strong and weak forces all propagate on a fixed spacetime background; gravity modifies that background. This means the large-scale geometry of the universe is not a passive stage for physics — it is a physical variable with its own dynamics, governed by Einstein's field equations.

Mach's principle and its rejection. The chapter surveys the philosophical tradition — associated with Ernst Mach, amplified by Dirac, Sciama, and Dicke — that holds local inertial frames to be determined by the large-scale distribution of matter. Hawking and Ellis explicitly set this aside. Their project is the opposite: to use locally-determined physical laws (established in the laboratory) to deduce what global spacetime structure must be, without assuming that the universe determines the laws.

The two predictions the book will establish. The authors state the book's two chief results at the outset: (1) massive stars that collapse past a certain point must form black holes containing singularities; (2) the universe, if it satisfies the observed large-scale conditions, must have begun at a singularity. Both conclusions rest on the same theoretical machinery developed in subsequent chapters.

Key ideas

• Physics naturally splits into local law (differential equations) and global boundary condition (what solutions are physically realized).
• Gravity uniquely controls spacetime geometry, making the global structure a physical rather than merely mathematical question.
• Laboratory-established physical laws are assumed to hold throughout spacetime — this is an extrapolation that the book treats as a working hypothesis, not a proven fact.
• The Machian tradition of deriving local laws from cosmic structure is explicitly set aside in favor of the reverse direction.
• The book's programme is stated: prove generic singularity results without assuming special symmetry.

Key takeaway

Gravity's unique role as the geometry of spacetime itself makes the global, large-scale structure of the universe a genuine physical problem, not merely a mathematical one — and the book's goal is to derive unavoidable conclusions about that structure from physically reasonable premises.

Chapter 2 — Differential Geometry

Central question

What mathematical structures on a smooth manifold are required to formulate a covariant theory of spacetime, and how are they systematically constructed?

Main argument

Manifolds (§2.1). A spacetime is modelled as a smooth (C∞) Hausdorff manifold M of dimension four. The chapter defines this rigorously: an atlas of coordinate charts, transition maps, and the notion of differentiability. The authors work at the level of generality needed for general relativity — no background metric is assumed at this stage.

Vectors, covectors, and tensors (§2.2). Tangent vectors are defined as equivalence classes of curves (derivations on smooth functions). The cotangent space, tensor products, and the full tensor algebra T^r_s(p) at each point are constructed. The abstract index notation used throughout the book is introduced here.

Maps of manifolds (§2.3). Diffeomorphisms and their induced pullback and pushforward maps on tensors are defined. This underpins the notion of symmetry (isometries), the Lie derivative, and the equivalence principle.

Exterior differentiation and the Lie derivative (§2.4). The exterior derivative d on differential forms, the interior product, and the Cartan formula for the Lie derivative £_X are developed. These are needed for integration on manifolds and for defining conserved quantities.

Covariant differentiation and the curvature tensor (§2.5). An affine connection ∇ is defined abstractly; covariant derivatives of arbitrary tensors are constructed. The Riemann curvature tensor R^a{bcd} is defined via the commutator of covariant derivatives: [∇c, ∇d]V^a = R^a{bcd} V^b. The Bianchi identity ∇{[e}R{ab]cd} = 0 is proved.

The metric and Levi-Civita connection (§2.6). A Lorentzian metric g{ab} (signature −+++) is introduced. The metric uniquely determines the torsion-free, metric-compatible Levi-Civita connection. The Riemann tensor is decomposed into the Weyl conformal tensor C{abcd} (trace-free part) and the Ricci tensor R{ab} = R^c{acb}, related by the contracted Bianchi identity ∇^a R{ab} = ½ ∇b R.

Hypersurfaces (§2.7). Spacelike, timelike, and null hypersurfaces are classified by the character of their normal. The second fundamental form (extrinsic curvature) K_{ab} of a hypersurface is defined — essential for the Cauchy problem of Chapter 7.

Volume element and Gauss' theorem (§2.8). The canonical volume 4-form ε_{abcd} = √|g| [abcd] is constructed. The covariant divergence theorem is proved; this is needed for conservation laws and for the Penrose–Hawking singularity theorems.

Fibre bundles (§2.9). The tangent bundle TM, cotangent bundle T*M, and frame bundles are introduced as the natural settings for tensor fields. Connections on principal bundles are defined, providing the abstract framework for gauge theories and spinors.

Key ideas

• A spacetime is a four-dimensional smooth Lorentzian manifold; no global coordinate system is assumed.
• The Riemann tensor encodes all intrinsic curvature; it vanishes if and only if the manifold is locally flat.
• The Weyl tensor governs tidal forces and gravitational radiation; the Ricci tensor is directly sourced by matter via Einstein's equations.
• The metric's signature (−+++) is the mathematical encoding of the physical distinction between time and space, and of light cone structure.
• The Levi-Civita connection is the unique torsion-free connection compatible with the metric — this uniqueness is non-trivial and physically important.
• Hypersurfaces and their extrinsic curvature provide the initial-data setting for the evolutionary formulation of general relativity.

Key takeaway

Chapter 2 provides the complete differential-geometric toolkit — manifolds, tensors, connections, curvature, and integration — that all subsequent physics will be expressed in; the care taken here ensures that later results about singularities and causality are genuinely coordinate-independent.

Chapter 3 — General Relativity

Central question

How is Einstein's general theory of relativity formulated as a rigorous geometric theory, and what are the physical content and constraints of its field equations?

Main argument

The spacetime manifold and equivalence principle (§3.1). The physical content of general relativity is encoded in the statement that spacetime is a Lorentzian manifold (M, g_{ab}), and that freely falling observers follow timelike geodesics. The equivalence principle — that in a sufficiently small region, spacetime is locally Minkowskian — is given precise mathematical meaning through the existence of a locally inertial frame at every point.

Matter fields and the stress-energy tensor (§3.2). Matter is described by fields (scalar, electromagnetic, perfect fluid, etc.) on the spacetime manifold. Each matter model contributes a stress-energy tensor T{ab} = T{ba} satisfying ∇^a T{ab} = 0 (local energy-momentum conservation, itself a consequence of the field equations via the Bianchi identity). The dominant and weak energy conditions are discussed here: T{ab} V^a V^b ≥ 0 for all timelike V^a (non-negative energy density for any observer).

Lagrangian formulation (§3.3). Einstein's equations are derived from the Hilbert–Einstein action S = ∫(R + Lm)√|g| d⁴x, where R is the Ricci scalar and Lm is the matter Lagrangian. Varying with respect to g_{ab} yields the field equations directly. This variational approach establishes that general relativity is the unique generally covariant theory of a massless spin-2 field at low energies.

Einstein's field equations (§3.4). The equations G{ab} ≡ R{ab} − ½Rg{ab} = 8πG T{ab} (in units c = 1) relate spacetime curvature (the left-hand side, the Einstein tensor G{ab}) to matter and energy (the right side). The cosmological constant Λ can be added as Λg{ab} on the left. The equations are ten nonlinear second-order partial differential equations for the ten independent components of g_{ab}; four are constraints and six are dynamical. The linearized theory (gravitational waves) is briefly discussed.

Key ideas

• Freely falling bodies follow geodesics — this is the geometrization of gravity.
• The stress-energy tensor T_{ab} must satisfy the dominant energy condition for the singularity theorems to apply: roughly, no observer measures a negative energy density or superluminal energy flux.
• The Einstein tensor G_{ab} is identically divergence-free (contracted Bianchi identity), so the ten equations are not all independent; this is the mathematical reason energy-momentum is automatically conserved.
• The field equations are highly nonlinear: superposition does not hold, and finding exact solutions requires special symmetry assumptions.
• General relativity does not single out a preferred time coordinate; the theory is fully covariant under arbitrary smooth coordinate transformations (diffeomorphism invariance).

Key takeaway

Chapter 3 establishes general relativity as the statement G{ab} = 8πG T{ab} on a Lorentzian manifold, with energy conditions on T_{ab} that will be the key physical input to every singularity theorem in the book.

Chapter 4 — The Physical Significance of Curvature

Central question

How does spacetime curvature physically affect the trajectories of particles and light, and why does positive energy density inevitably cause geodesics to focus and eventually cross?

Main argument

Timelike geodesic congruences — expansion, shear, vorticity (§4.1). Consider a smooth one-parameter family of timelike geodesics (a congruence). The behaviour of neighbouring geodesics is characterized by three kinematic quantities: expansion θ (whether the family is diverging or converging), shear σ{ab} (distortion of cross-sections without change of volume), and vorticity ω{ab} (rotation). The Raychaudhuri equation governing the evolution of θ is:

dθ/dτ = −(θ²/3) − σ{ab}σ^{ab} + ω{ab}ω^{ab} − R_{ab}u^a u^b

where u^a is the unit tangent to the geodesics and R_{ab}u^a u^b is the Ricci curvature contracted with the flow direction. This equation is central to every singularity theorem in the book.

Null geodesic congruences (§4.2). The analogous formalism for null geodesics (light rays) is developed. The expansion θ, shear, and vorticity of a null congruence are defined, and the null Raychaudhuri equation is derived. The optical scalars of a congruence of light rays encode gravitational lensing effects.

Energy conditions and geodesic focusing (§4.3). The strong energy condition (SEC) requires R{ab}V^a V^b ≥ 0 for all timelike V^a, which by Einstein's equations is equivalent to (T{ab} − ½Tg{ab})V^a V^b ≥ 0. The null energy condition (NEC) requires R{ab}k^a k^b ≥ 0 for all null k^a. Under the SEC with ω_{ab} = 0 (irrotational congruence), the Raychaudhuri equation gives:

dθ/dτ ≤ −θ²/3

This implies that if θ = θ₀ < 0 at some point, then θ → −∞ (the congruence collapses to a caustic) within proper time τ ≤ 3/|θ₀|. Positive energy causes geodesics to focus.

Conjugate points (§4.4). A conjugate point to p along a geodesic γ is a point q ≠ p where a nearby geodesic starting at p reconverges to meet γ. Conjugate points arise precisely when the Jacobi fields (solutions to the geodesic deviation equation) vanish. The energy conditions guarantee that conjugate points must exist along any geodesic that starts with a converging congruence.

Variation of arc-length (§4.5). Beyond the first conjugate point, a geodesic ceases to be a local length-maximizer among timelike curves. This means that if a timelike geodesic γ from p passes through a conjugate point q before reaching r, then there exists a longer timelike curve from p to r — which, in the Lorentzian geometry, means γ is not in the causal boundary between p and r. This variational lemma is the kinematic key used in the singularity theorems.

Key ideas

• The Raychaudhuri equation is the master equation of the book: it shows that positive Ricci curvature (energy) causes geodesic congruences to focus.
• Expansion θ, shear σ, and vorticity ω fully characterize the local behaviour of a geodesic congruence.
• The strong energy condition ensures that gravity is always attractive for geodesic deviation.
• Conjugate points are the geometric signal that a geodesic is becoming extremal: past a conjugate point, there is a shortcut.
• The connection between conjugate points and the failure of geodesics to maximize arc-length is the technical bridge from energy conditions to causal pathology.

Key takeaway

Chapter 4 proves the focusing lemma — that physically reasonable energy conditions force geodesic families to develop caustics — which is the indispensable engine of every singularity theorem in the book.

Chapter 5 — Exact Solutions

Central question

What are the global causal and geometric structures of the physically most important exact solutions to Einstein's equations, and what singularities and horizons do they contain?

Main argument

Minkowski space (§5.1). The flat solution T_{ab} = 0, Λ = 0. Its global structure is that of ℝ⁴ with signature (−,+,+,+). The conformal compactification (Penrose diagram) is introduced: by a conformal rescaling, the entire Minkowski spacetime is mapped to a finite region, exposing the structure of null, timelike, and spacelike infinities (ℐ⁺, ℐ⁻, i⁰, i⁺, i⁻). This diagrammatic technique is generalized to all subsequent solutions.

De Sitter and anti-de Sitter spaces (§5.2). The maximally symmetric solutions with Λ > 0 (de Sitter) and Λ < 0 (anti-de Sitter). De Sitter space has topology ℝ × S³ and contains a cosmological event horizon: observers are causally cut off from large portions of the spacetime. Its Penrose diagram reveals spacelike future and past infinities, unlike Minkowski's null infinity.

Robertson-Walker spaces (§5.3). The cosmologically relevant solutions describing a spatially homogeneous, isotropic universe filled with a perfect fluid. The metric is ds² = −dt² + a(t)²[dr²/(1−kr²) + r²dΩ²] where k ∈ {−1,0,+1} gives the spatial curvature and a(t) is the scale factor. The Friedmann equations for ȧ and ä are derived. When the strong energy condition holds and ä ≤ 0, a(t) → 0 in finite past time, giving the initial Big Bang singularity — a point of divergent density and curvature.

Spatially homogeneous cosmological models (§5.4). More general cosmologies that are spatially homogeneous but not necessarily isotropic (Bianchi models). These are classified by the Lie algebra of their three-dimensional symmetry group. Some Bianchi models (e.g. Bianchi IX, the Mixmaster universe) exhibit chaotic oscillatory approach to the singularity — the Mixmaster behaviour in which the three scale factors undergo BKL (Belinsky-Khalatnikov-Lifshitz) oscillations.

Schwarzschild and Reissner-Nordström solutions (§5.5). The unique spherically symmetric vacuum solution is the Schwarzschild metric ds² = −(1−2M/r)dt² + (1−2M/r)⁻¹dr² + r²dΩ². The coordinate singularity at r = 2M (the Schwarzschild radius) is distinguished from the curvature singularity at r = 0. The Kruskal–Szekeres extension reveals the complete global structure: two exterior regions, a future singularity, and a past singularity, connected by an Einstein-Rosen bridge. The Reissner-Nordström solution for a charged mass is also analyzed; it has an inner Cauchy horizon as well as an outer event horizon.

The Kerr solution (§5.6). The unique stationary, axisymmetric vacuum solution for a rotating mass. The metric has an ergosphere (outside the event horizon) and inner and outer event horizons. Penrose's process of energy extraction from the ergosphere is described. The global structure is significantly more complex than Schwarzschild's.

Gödel's universe (§5.7). An exact solution with rotating dust (Λ ≠ 0) that admits closed timelike curves (CTCs) — paths in spacetime that loop back to their own past. This solution is examined as an illustration that Einstein's equations alone do not forbid causality violation; additional conditions on causal structure are needed.

Taub-NUT space (§5.8). A vacuum solution with unusual topology: the spatial sections are three-spheres, and the solution has a parameter N (the NUT charge) that gives it a gravitomagnetic character. Taub-NUT space contains a Misner string (a coordinate singularity analogous to a Dirac string in electromagnetism) and illustrates how extensions of spacetime can be non-unique, motivating the global techniques of Chapter 6.

Further exact solutions (§5.9). A survey of other important solutions: plane gravitational waves, the Kasner anisotropic cosmology, and product solutions, each illustrating particular features of spacetime geometry.

Key ideas

• The Penrose (conformal) diagram maps the entire global causal structure of a spacetime to a finite picture; its construction depends only on the conformal class of the metric.
• The Schwarzschild solution's apparent singularity at r = 2M is a coordinate artifact; the Kruskal extension shows it is a regular event horizon.
• The Friedmann Big Bang singularity is a generic feature of Robertson-Walker models satisfying the strong energy condition — not a consequence of special symmetry.
• Gödel's universe demonstrates that causality violation is not automatically excluded by Einstein's equations; it must be imposed as a separate physical requirement.
• Taub-NUT space shows that maximal analytic extension of a spacetime can be non-unique, making the global choice of topology a physical question.

Key takeaway

Chapter 5 surveys the zoo of exact solutions to illustrate the range of possible global structures — singularities, horizons, closed timelike curves — and to establish the concrete geometries that the abstract machinery of Chapters 6–10 will analyze.

Chapter 6 — Causal Structure

Central question

What is the complete mathematical theory of causality in a curved spacetime, and what are the conditions under which a spacetime admits a well-posed global physics?

Main argument

Orientability (§6.1). The spacetime manifold must be time-orientable: there must be a globally consistent choice of future and past for every light cone. This is imposed as a physical axiom — without it, the notions of cause and effect are undefined globally.

Causal curves (§6.2). A curve is causal (or non-spacelike) if its tangent vector is everywhere timelike or null. The chronological future I⁺(p) of a point p is the set of all points reachable from p by a future-directed timelike curve; the causal future J⁺(p) includes the null boundary. The past analogues I⁻(p) and J⁻(p) are defined symmetrically. Their topological and set-theoretic properties are systematically developed.

Achronal boundaries (§6.3). The achronal boundary ∂I⁺(S) of a set S is the boundary of the chronological future of S; it is generated by null geodesic segments that can have future endpoints but no past endpoints. This structure is crucial for defining event horizons precisely.

Causality conditions (§6.4). A hierarchy of progressively stronger causality conditions is introduced:

• Chronology condition: No closed timelike curves (CTCs). Violated in Gödel's universe.
• Causality condition: No closed causal (non-spacelike) curves.
• Strong causality: No almost-closed causal curves (no causal curves that loop nearly back to their starting point).
• Stable causality: The spacetime admits a global time function t such that dt is everywhere past-directed timelike (equivalently, the causal hierarchy is preserved under small metric perturbations). These conditions form a strict hierarchy; stable causality is the physically reasonable assumption that rules out causal pathologies.

Cauchy developments and Cauchy horizons (§6.5). Given a partial Cauchy surface S (a spacelike hypersurface that no causal curve intersects more than once), its future Cauchy development D⁺(S) is the region of spacetime fully determined by initial data on S — every past-directed inextendible causal curve through a point of D⁺(S) crosses S. The boundary ∂D⁺(S) is the future Cauchy horizon H⁺(S), generated by null geodesic segments that have no past endpoints on S. Beyond the Cauchy horizon, physics on S cannot predict what happens.

Global hyperbolicity (§6.6). A spacetime is globally hyperbolic if it satisfies the strong causality condition and the sets J⁺(p) ∩ J⁻(q) are compact for all p, q. Geroch's theorem is proved: a globally hyperbolic spacetime admits a Cauchy surface S such that the spacetime is diffeomorphic to S × ℝ. This is the gold-standard causality condition — it implies the Cauchy problem is well-posed and the spacetime has no causal pathologies.

Existence of geodesics (§6.7). In a globally hyperbolic spacetime, the causal diamond J⁺(p) ∩ J⁻(q) is compact, which implies there always exists a causal geodesic of maximal length between causally related points p and q. This geodesic maximization result is the causal analogue of the Riemannian shortest-path theorem, and it plays a central role in the proofs of the singularity theorems.

The causal boundary of spacetime (§6.8). The construction of ideal boundary points (b-boundary) that represent the "edge" of spacetime — singularities and points at infinity. Schmidt's b-boundary construction attaches an ideal boundary ∂M to M so that incomplete geodesics have endpoints there.

Asymptotically simple spaces (§6.9). Spacetimes that are asymptotically flat (vanishing curvature at infinity in an appropriate sense) can be conformally compactified to attach null infinity ℐ⁺ and ℐ⁻ (Penrose's scri). This provides the rigorous definition of isolated gravitating systems and of gravitational radiation.

Key ideas

• The chronological/causal futures I⁺, J⁺ are the fundamental objects of causal analysis; their topological properties are non-trivially constrained by causality conditions.
• Global hyperbolicity is the correct causality condition for well-posed dynamics: it ensures both that Cauchy surfaces exist and that geodesics can be maximized.
• The Cauchy horizon H⁺(S) is the boundary of predictability from initial data on S; its existence signals a breakdown in the determinism of general relativity.
• Stable causality excludes not only actual CTCs but also spacetimes arbitrarily close (in the C⁰ topology on metrics) to having them.
• The compactness of J⁺(p) ∩ J⁻(q) is the key property linking global hyperbolicity to the existence of length-maximizing geodesics.

Key takeaway

Chapter 6 provides the complete causal-structure toolkit — the I⁺/J⁺ formalism, Cauchy developments, global hyperbolicity, and the existence of maximizing geodesics — which the singularity theorems of Chapter 8 assemble into their proofs.

Chapter 7 — The Cauchy Problem in General Relativity

Central question

Does general relativity, as a system of partial differential equations, have a well-posed initial-value formulation — given data on a spatial slice, is the future evolution uniquely determined?

Main argument

The nature of the problem (§7.1). The ten Einstein equations G{ab} = 8πG T{ab} are not all dynamical: four are constraint equations (analogous to ∇·B = 0 in electromagnetism) that the initial data must satisfy, and six are genuinely evolutionary. The gauge freedom (diffeomorphism invariance) means that the solution is determined only up to a diffeomorphism — one must choose a gauge to get a unique metric.

The reduced Einstein equations (§7.2). By introducing a background metric g̃{ab} and imposing four harmonic gauge (de Donder gauge) conditions — ∇^b(g{ab} − ½g g̃{ab}) = 0 (in coordinates, □x^μ = 0) — the ten Einstein equations reduce to a system of ten wave equations of the form □g{ab} = F{ab}(g, ∂g), where F{ab} is a non-linear function of the metric and its first derivatives. This is the reduced system.

The initial data (§7.3). On a spacelike 3-surface Σ, the initial data consist of: the induced Riemannian metric h{ij} and the extrinsic curvature K{ij}. These must satisfy the Gauss-Codazzi constraint equations — four equations that are necessary conditions for the data to be embeddable in a Lorentzian spacetime satisfying Einstein's equations. The constraints form an underdetermined elliptic system (two scalar constraints and one vector constraint per point).

Second-order hyperbolic equations (§7.4). The mathematical theory of quasi-linear symmetric hyperbolic systems is developed. The key result (Leray's theorem) is that such systems have unique local solutions given smooth initial data; the domain of dependence of a region is bounded by the characteristics (light cones).

Existence and uniqueness for vacuum equations (§7.5). Combining the harmonic reduction with Leray's theory, Choquet-Bruhat's theorem is proved: given smooth constraint-satisfying initial data (h{ij}, K{ij}) on Σ, there exists a unique (up to diffeomorphism) local Cauchy development — a neighbourhood of Σ in which the vacuum Einstein equations hold. The solution is indeed hyperbolic and signals propagate within light cones.

Maximal development and stability (§7.6). Choquet-Bruhat and Geroch's theorem on the maximal development is stated: among all Cauchy developments of given initial data, there is a unique maximal one that contains every other development as an open subset. The stability of solutions (small perturbations of initial data give small changes in the spacetime, within the domain of dependence) is also discussed.

Einstein equations with matter (§7.7). The Cauchy problem for Einstein's equations coupled to specific matter models — perfect fluids, Maxwell fields, scalar fields — is analyzed. For each well-behaved matter model, the coupled system remains hyperbolic and the Cauchy problem retains its well-posedness.

Key ideas

• The four constraint equations restrict which initial data sets are physically admissible; they cannot be chosen freely.
• Harmonic gauge is a technical device for revealing the hyperbolic character of the equations; the physical content is gauge-invariant.
• Leray's theory of hyperbolic systems is the mathematical bedrock for well-posedness; it guarantees that information cannot propagate faster than light.
• The maximal development is the largest spacetime consistent with given initial data; it may be geodesically incomplete, in which case the spacetime contains a singularity.
• The authors note that Chapter 7 is not logically needed for Chapters 8–10, but its content establishes that the spacetimes in those chapters are genuine solutions, not artifacts.

Key takeaway

Chapter 7 establishes that general relativity is a well-posed initial-value theory: smooth initial data on a spacelike surface uniquely determine a maximal spacetime, which may nevertheless be geodesically incomplete — and that incompleteness is what the singularity theorems characterize.

Chapter 8 — Space-Time Singularities

Central question

What is the correct definition of a spacetime singularity, and under what physically reasonable conditions can singularities be proven to exist?

Main argument

The definition of singularities (§8.1). A naive definition of singularity as "a point where curvature diverges" fails because such points are not part of the manifold M — they have been "cut out." The correct definition, following Schmidt, is b-incompleteness: a spacetime is singular if it contains an inextendible causal geodesic of finite generalized affine parameter (b-length). The b-boundary ∂M consists of the equivalence classes of such incomplete geodesic endpoints — the singular points. A b-incomplete geodesic represents either a particle that ceases to exist after finite proper time (timelike incompleteness) or a light ray that terminates (null incompleteness).

The singularity theorems (§8.2). Four theorems of increasing generality are proved. All share the same logical structure: a set of physically reasonable conditions + a "trapped" initial condition → geodesic incompleteness.

Theorem 1 (Penrose, 1965): If (a) the NEC holds, (b) the spacetime is globally hyperbolic with a non-compact Cauchy surface, and (c) there exists a closed trapped surface S (a compact spacelike 2-surface on which both ingoing and outgoing null geodesics have negative expansion θ < 0), then the spacetime contains a future null-incomplete geodesic.

Theorem 2 (Hawking, 1967): If (a) the SEC holds, (b) there are no closed timelike curves, (c) every timelike or null geodesic contains a point where the curvature is non-zero (generic condition), and (d) there exists a compact spacelike hypersurface (a closed universe), then the spacetime contains a past-incomplete timelike geodesic — an initial singularity.

Theorem 3 (Hawking–Penrose, 1970): The most general theorem: if (a) the SEC holds for timelike geodesics, (b) every timelike and null geodesic meets a point where k{[a}R{b]cd[e}k_{f]}k^c k^d ≠ 0, (c) no closed timelike curves exist, and (d) there exists either a closed trapped surface, or a trapped point (where all future-directed null geodesics are converging), or a compact achronal set, then the spacetime is geodesically incomplete. This unifies collapse and cosmological singularities in a single theorem.

Theorem 4: A variant using the non-spacelike generic condition and a compact Cauchy surface, giving incompleteness without assuming expansion of the universe.

The proof strategy. Each theorem follows the same route: (1) the focusing lemma (Chapter 4) shows that any geodesic satisfying the trapped condition will develop a conjugate point; (2) beyond the conjugate point, the geodesic is no longer a local length maximizer; (3) global hyperbolicity (Chapter 6) provides a maximizing geodesic from any point to the Cauchy surface; (4) the existence of a maximizing geodesic past a conjugate point is a contradiction; (5) therefore no such complete geodesic can exist.

The description of singularities (§8.3). Schmidt's b-boundary construction places the singularities as boundary points of M. The topology of ∂M is analyzed; it may be a single point (all incomplete geodesics converge to the same boundary point) or a more complicated set.

The character of singularities (§8.4). The theorems establish incompleteness but say nothing about curvature divergence. The authors show, however, that at least some singularities guaranteed by the theorems must involve curvature blow-up — specifically, the singularities predicted by Theorems 1 and 2 cannot be merely Cseries-discontinuities in curvature.

Imprisoned incompleteness (§8.5). A pathological case where incomplete geodesics remain within a compact region — they oscillate forever without leaving, never reaching a "boundary." The Taub-NUT solution contains an example. Such singularities arise from causal pathologies rather than curvature blow-up.

Key ideas

• Geodesic incompleteness (b-incompleteness) is the correct definition of a spacetime singularity; curvature divergence is a consequence in many but not all cases.
• The trapped surface condition is the crucial physical trigger for the singularity theorems: once ingoing and outgoing null geodesics both converge on a compact surface, collapse is inevitable.
• The generic condition (every geodesic experiences some focusing at some point) is a very weak assumption satisfied by any spacetime with local gravitational fields.
• The Hawking–Penrose theorem (1970) is the most general: it applies to both cosmological and black-hole singularities simultaneously.
• The singularity theorems are existence results: they guarantee incompleteness without specifying the detailed structure of the singularity.

Key takeaway

Chapter 8 proves that singularities — geodesically incomplete curves — are generic in general relativity: any spacetime satisfying physically reasonable energy conditions and containing a trapped region or a closed universe cannot be complete, establishing that singularities are not mathematical pathologies but physical predictions.

Chapter 9 — Gravitational Collapse and Black Holes

Central question

What is the global causal structure of spacetimes containing gravitationally collapsed objects, how are black holes rigorously defined, and what are the fundamental properties of their event horizons?

Main argument

Stellar collapse (§9.1). The chapter opens with the physics of gravitational collapse: a star exhausting its nuclear fuel loses the pressure support that balances gravity. For sufficiently massive stars (above the Oppenheimer-Volkoff limit), no equilibrium configuration exists; the star undergoes complete gravitational collapse. The singularity theorems guarantee incompleteness; the question is whether the resulting singularity is hidden from external observers (cosmic censorship) or naked.

The collapse of a spherically symmetric star is modelled using the interior and exterior Schwarzschild solutions matched at the star's surface. As the star contracts through its Schwarzschild radius r = 2M, the surface passes into a region from which light cannot escape. External observers see the surface asymptotically approaching r = 2M as t → ∞ (due to gravitational redshift), but the star physically crosses r = 2M in finite proper time.

Black holes — the event horizon (§9.2). A black hole is defined precisely as a region of spacetime from which signals cannot escape to future null infinity ℐ⁺. The event horizon B = M \ J⁻(ℐ⁺) is the boundary of the causal past of ℐ⁺ — the past set of points that can send signals to distant observers. Key properties established:

• The event horizon is a null hypersurface generated by null geodesics that have no future endpoints and may have past endpoints.
• The event horizon is an achronal set (no two of its points are timelike-related).
• Hawking's area theorem: if the NEC holds and the spacetime is globally hyperbolic, the area of cross-sections of the event horizon is a non-decreasing function of time. This is the black hole analogue of the second law of thermodynamics (and motivated Bekenstein's later identification of entropy with area).
• The topology theorem: cross-sections of the event horizon of a stationary black hole in four dimensions are topologically 2-spheres (S²).

The authors also prove that the event horizon must be a trapped surface: the outgoing null geodesics from the horizon have zero expansion (θ = 0), meaning the horizon is in a critical state between focusing and defocusing.

The final state of black holes (§9.3). Hawking's rigidity theorem (stationary implies static or axisymmetric) and the uniqueness theorems (no-hair theorems) are discussed:

• A stationary vacuum black hole must be the Schwarzschild solution (characterized by mass M alone) or the Kerr solution (characterized by mass M and angular momentum J).
• A stationary charged black hole is the Reissner-Nordström (M, Q) or Kerr-Newman (M, J, Q) solution.
• Thus, the final state of any gravitational collapse must be one of these family members — all information about the collapsing matter (baryon number, shape, etc.) is lost except M, J, Q.

The Penrose process and the Christodoulou irreducible mass are described: energy can be extracted from a rotating black hole's ergosphere, but this decreases its angular momentum (and hence its horizon area), saturating at the Schwarzschild solution.

Key ideas

• A black hole is a causally defined object: the region from which escape to ℐ⁺ is impossible. This definition requires knowing the entire future of the spacetime.
• The event horizon is teleological: its location today depends on what falls into the black hole in the future.
• The area theorem (Hawking 1971) implies black holes obey a second-law-like constraint; this would later be identified as the second law of black hole thermodynamics.
• The no-hair theorems assert that the endpoint of collapse is fully specified by (M, J, Q) — a radical information loss about the collapsing matter.
• Cosmic censorship — that singularities formed in gravitational collapse are always hidden behind event horizons — is assumed but not proved; the authors note it as a major open problem.

Key takeaway

Chapter 9 defines black holes rigorously through causal structure, proves the area theorem and topology theorem for event horizons, and establishes the no-hair result that the final state of collapse is one of a small family of stationary solutions characterized entirely by mass, charge, and angular momentum.

Chapter 10 — The Initial Singularity in the Universe

Central question

Does the observed universe — expanding, filled with matter and radiation detected in the cosmic microwave background — necessarily have had an initial singularity in its past?

Main argument

The expansion of the universe (§10.1). The observational evidence for a large-scale expanding universe is reviewed: Hubble's law (recession velocity v = H₀d), the homogeneity and isotropy implied by the cosmological principle, and the discovery of the cosmic microwave background radiation (CMB) by Penzias and Wilson (1965). The CMB is the decisive evidence: it shows the universe was once in a state of thermal equilibrium at high temperature and density. The isotropy of the CMB is used directly in the singularity argument.

Applying the singularity theorems to the universe (§10.2). Hawking and Ellis show that the CMB provides the trapped condition needed for the singularity theorem. The argument runs as follows:

1. The CMB photons we observe today were emitted from a last-scattering surface at redshift z ≈ 1100. Their observed near-perfect isotropy implies that the past light cone of the current observer converges to a very small region at the epoch of last scattering.

2. More precisely: the focusing of null geodesics (incoming CMB photons) along the past light cone of any observer, combined with the fact that the universe has positive Ricci curvature (satisfied if the energy conditions hold and the matter density is positive), implies that the past light cone develops a conjugate point within finite affine parameter — i.e., the past light cone reconverges.

3. A reconverging past light cone is analogous to a trapped surface for the cosmological case. By Theorem 2 (Hawking 1967), this implies past timelike geodesic incompleteness: the universe's past cannot be extended indefinitely, and the cosmological model must contain an initial singularity.

This is a non-perturbative result: it does not assume exact Robertson-Walker symmetry. It applies to any spacetime whose past light cones show the observed degree of isotropy, regardless of local inhomogeneities.

The nature and implications of the singularities (§10.2 continued). The authors discuss what the initial singularity means physically: it is a boundary of spacetime where the density, temperature, and curvature diverge, and where the known laws of physics break down. They note that the singularity theorems cannot characterize the structure of the singularity in detail — whether it is isotropic (as in FLRW) or anisotropic (as in BKL oscillations). The Weyl curvature hypothesis (that the initial singularity has near-zero Weyl tensor, corresponding to low gravitational entropy) is discussed as an additional physical condition, not derivable from the singularity theorems themselves.

The chapter ends by noting that the singularity theorems establish the incompleteness of spacetime under general relativity, but that this incompleteness signals the breakdown of the theory itself — quantum gravity will be needed to understand what happens at or before the initial singularity.

Key ideas

• The cosmic microwave background radiation provides the observational input (past-directed focusing of null geodesics) that activates the singularity theorem.
• The argument does not require exact isotropy or homogeneity: the observed approximate isotropy is sufficient.
• The initial cosmological singularity is a theorem, not a model assumption: it follows from the observed state of the universe plus general relativity plus the energy conditions.
• The singularity theorems cannot say what happened "before" the singularity, because the singularity is a boundary of spacetime, not a point within it.
• The regime near the singularity — extreme density and curvature — is precisely where quantum effects are expected to be important, suggesting that a quantum theory of gravity is needed to resolve the singularity.

Key takeaway

Chapter 10 shows that the observed expanding universe, combined with the positive energy conditions and the focusing effect shown by the CMB, satisfies the conditions of the Hawking singularity theorem — the Big Bang singularity is not an assumption of the standard cosmological model but a theorem derivable from it.

The book's overall argument

1. Chapter 1 (The Role of Gravity) — Establishes the program: gravity uniquely controls spacetime geometry, making the global structure of the universe a physical problem; the book will use local physics to derive global conclusions about singularities and black holes.

2. Chapter 2 (Differential Geometry) — Builds the mathematical foundation: a Lorentzian manifold with a Levi-Civita connection, Riemann and Weyl curvature tensors, and the integration machinery needed for a covariant physical theory.

3. Chapter 3 (General Relativity) — Erects the physical theory on that foundation: Einstein's field equations G{ab} = 8πGT{ab}, derived from the Hilbert action, with energy conditions on T_{ab} that will be the key physical input to every singularity theorem.

4. Chapter 4 (The Physical Significance of Curvature) — Proves the focusing lemma: the Raychaudhuri equation shows that the strong/null energy conditions force geodesic families to develop conjugate points, meaning complete geodesics past those points would violate length-maximization — the central contradiction used in every singularity proof.

5. Chapter 5 (Exact Solutions) — Examines the global structure of the most important solutions (Minkowski, Schwarzschild, Kerr, Friedmann) to build intuition and establish that singularities and horizons occur in specific known cases — motivating the search for general, non-symmetric proofs.

6. Chapter 6 (Causal Structure) — Develops the full theory of causality: I⁺/J⁺ sets, Cauchy developments, global hyperbolicity, and the existence of maximizing geodesics in globally hyperbolic spacetimes — providing the causal infrastructure that the singularity theorems require.

7. Chapter 7 (The Cauchy Problem) — Establishes that general relativity is a well-posed initial-value theory: given consistent initial data, there is a unique maximal development; that development may be incomplete — the singularity theorems characterize precisely when it must be.

8. Chapter 8 (Space-Time Singularities) — Delivers the book's central theorems: using the focusing lemma (Chapter 4) and the causal machinery (Chapter 6), the Penrose, Hawking, and Hawking-Penrose theorems prove that physically reasonable spacetimes with trapped regions or closed spatial sections are necessarily geodesically incomplete — they contain singularities.

9. Chapter 9 (Gravitational Collapse and Black Holes) — Applies the general theory to astrophysical collapse: defines black holes via causal structure, proves the area theorem and topology theorem for event horizons, and establishes the no-hair uniqueness theorems showing that collapsed stars inevitably produce one of a small family of stationary black holes.

10. Chapter 10 (The Initial Singularity in the Universe) — Closes the cosmological argument: the observed cosmic microwave background provides the trapped condition required by the Hawking singularity theorem, proving that the Big Bang is not a model assumption but a mathematical consequence of observed cosmology plus general relativity.

Common misunderstandings

Misunderstanding: The singularity theorems prove that curvature diverges to infinity.

The theorems prove only that certain geodesics are incomplete in finite affine parameter (b-incompleteness). Curvature divergence is a common but not universal consequence; the theorems themselves are silent about the detailed physical character of the singularity. Some b-incomplete geodesics (imprisoned incompleteness, Section 8.5) do not involve curvature blow-up at all.

Misunderstanding: The Big Bang singularity follows from the assumption that the universe is perfectly homogeneous and isotropic (FLRW symmetry).

The singularity theorems are explicitly non-symmetric. The cosmological argument in Chapter 10 requires only the observed approximate isotropy of the CMB and the energy conditions — not exact FLRW symmetry. The result holds for a wide class of inhomogeneous cosmologies satisfying those conditions.

Misunderstanding: Hawking and Ellis prove that singularities are physically real, not just mathematical artifacts.

The theorems establish geodesic incompleteness within the mathematical framework of classical general relativity. The authors are explicit that near the singularity, where densities and curvatures reach Planck-scale values, general relativity itself breaks down and a quantum theory of gravity is required. The singularities are predictions of a theory known to be incomplete in that regime.

Misunderstanding: The event horizon of a black hole is where time "stops" or where extreme physical effects are experienced.

The event horizon is a globally defined causal boundary: an observer crossing it experiences nothing locally special (for a large enough black hole, tidal forces are negligible at the horizon). The dramatic physics occurs at the singularity inside the horizon, not at the horizon itself. The horizon's significance is teleological — it is defined by the entire future causal structure.

Misunderstanding: The book requires that trapped surfaces always correspond to black holes.

Trapped surfaces trigger the singularity theorems but do not by themselves imply black holes. A trapped surface implies a future singularity (in the presence of energy conditions), but whether that singularity is hidden by an event horizon (Penrose's cosmic censorship hypothesis) is a separate conjecture that Hawking and Ellis state as an open problem, not a theorem.

Misunderstanding: The Cauchy problem chapter (Chapter 7) is central to the singularity theorems.

Hawking and Ellis explicitly note that Chapter 7 is included for its intrinsic interest and completeness, but is "not really needed for the remaining three chapters." The singularity theorems in Chapter 8 use the causal structure of Chapter 6 but do not rely on the well-posedness results of Chapter 7.

Central paradox / key insight

The book's central paradox is this: general relativity, which is formulated as a smooth geometric theory on a differentiable manifold, predicts the destruction of that very smoothness — the existence of spacetime points where the manifold cannot be extended.

In any other physical theory, a prediction of mathematical breakdown signals the failure of the model. In general relativity, Hawking and Ellis show that the breakdown is not a flaw in the theory but a robust theorem: under the same energy conditions that make gravity a classical theory (positive energy density, no exotic matter), spacetime must develop incomplete geodesics. The theory is self-undermining in an unavoidable way.

The resolution of the paradox lies outside classical general relativity. The singularity theorems are universally interpreted as pointing to the necessity of a quantum theory of gravity — one that supersedes general relativity at Planck-scale energies and resolves the singularities into something finite. In this sense, the most important result of the book is a proof that general relativity has a built-in expiration date: it cannot be the final theory.

The singularities predicted by the classical theory represent not a failure of mathematics but a genuine physical prediction that the theory describing them is incomplete — a self-announcement of the need for quantum gravity.

Important concepts

Spacetime manifold

A smooth, four-dimensional, Hausdorff, paracompact manifold M equipped with a Lorentzian metric g_{ab} of signature (−,+,+,+). The metric encodes the causal structure (light cones) and the geometry (curvature) simultaneously.

Lorentzian metric

A non-degenerate symmetric (0,2) tensor field g{ab} on M with signature (−,+,+,+). Vectors V^a are timelike if g{ab}V^a V^b < 0, null if = 0, spacelike if > 0. The metric determines which curves are causal (non-spacelike).

Geodesic

A curve whose tangent vector is parallel-transported along itself: ∇_{u}u^a = 0 (affinely parameterized). Timelike geodesics are the paths of freely falling massive particles; null geodesics are the paths of light rays.

Raychaudhuri equation

The evolution equation for the expansion θ of a congruence of timelike geodesics: dθ/dτ = −(θ²/3) − σ{ab}σ^{ab} + ω{ab}ω^{ab} − R_{ab}u^a u^b. Under the strong energy condition with ω = 0, the right side is ≤ −θ²/3, proving focusing.

Energy conditions

Inequalities on the stress-energy tensor T_{ab}:

• Weak energy condition (WEC): T_{ab}V^a V^b ≥ 0 for all timelike V^a (non-negative energy density).
• Strong energy condition (SEC): (T{ab} − ½Tg{ab})V^a V^b ≥ 0 for timelike V^a; equivalent to R_{ab}V^a V^b ≥ 0. Used in cosmological singularity theorems.
• Null energy condition (NEC): T_{ab}k^a k^b ≥ 0 for all null k^a; weakest condition, used in the Penrose black-hole theorem and the area theorem.
• Dominant energy condition (DEC): WEC plus: T^a_{\ b}V^b is causal for all causal V^a (energy flux is never superluminal).

Conjugate points

Two points p and q on a geodesic γ are conjugate if there is a non-trivial Jacobi field (solution to the geodesic deviation equation D²J^a/dτ² + R^a_{bcd}u^b J^c u^d = 0) that vanishes at both p and q. Their existence signals that the geodesic is not a local length maximizer past q.

Trapped surface

A compact spacelike 2-surface S such that both families of null geodesics orthogonal to S have negative expansion (θ < 0). Physically: even outgoing light rays are converging. Occurs inside the Schwarzschild horizon at r < 2M.

Closed trapped surface

A compact (without boundary) trapped surface. Its existence is the key hypothesis in the Penrose singularity theorem for black holes.

B-incompleteness (b-incompleteness)

A geodesic γ : [0, a) → M is b-incomplete if it cannot be extended beyond the endpoint a (it "ends" in finite affine parameter) while a and M remain smooth. A spacetime is singular if it contains a b-incomplete causal geodesic.

Causal future J⁺(p)

The set of all points that can be reached from p by future-directed causal (non-spacelike) curves. The boundary ∂J⁺(p) is generated by future-directed null geodesics from p with no focal points.

Chronological future I⁺(p)

The set of all points reachable from p by future-directed timelike curves. I⁺(p) is an open set; J⁺(p) is its closure in causally well-behaved spacetimes.

Global hyperbolicity

A spacetime is globally hyperbolic if (a) it satisfies strong causality (no almost-closed causal curves) and (b) J⁺(p) ∩ J⁻(q) is compact for all p, q. Equivalent by Geroch's theorem to the existence of a Cauchy surface — a spacelike hypersurface met exactly once by every inextendible causal curve.

Cauchy surface

A closed achronal hypersurface Σ such that D⁺(Σ) ∪ D⁻(Σ) = M: every point in the spacetime lies in the domain of dependence of Σ. The existence of a Cauchy surface is equivalent to global hyperbolicity and guarantees well-posedness of the initial value problem.

Cauchy horizon H⁺(S)

The future boundary of the future Cauchy development D⁺(S) of a partial Cauchy surface S: H⁺(S) = D̄⁺(S) \ I⁻(D⁺(S)). Points on the Cauchy horizon are not determined by initial data on S — they represent the failure of global predictability.

Event horizon

The causal boundary B = ∂J⁻(ℐ⁺) — the boundary of the region that can communicate with future null infinity ℐ⁺. The event horizon is a null hypersurface whose area (by Hawking's area theorem) cannot decrease.

Penrose (conformal) diagram

A two-dimensional representation of the causal structure of a spacetime obtained by conformally rescaling the metric to bring infinity to a finite location. Null geodesics appear at 45° angles; the entire global causal structure is visible in a bounded diagram.

Weyl tensor

The trace-free part of the Riemann curvature tensor: C{abcd} = R{abcd} − (g{a[c}R{d]b} − g{b[c}R{d]a}) + (1/3)Rg{a[c}g{d]b}. The Weyl tensor vanishes in conformally flat spacetimes; it governs tidal forces and free gravitational radiation, and is zero in FLRW cosmologies (the Weyl curvature hypothesis).

Singularity theorems

A family of theorems (Penrose 1965, Hawking 1966–1967, Hawking-Penrose 1970) proving that physically reasonable spacetimes satisfying energy conditions and containing trapped regions or compact spatial sections are necessarily geodesically incomplete — they contain singularities.

References and Web Links

Primary book and edition information

• Hawking, S. W. and Ellis, G. F. R. The Large Scale Structure of Space-Time. Cambridge University Press, 1973. ISBN 978-0-521-09906-6.

  • Cambridge University Press book page
  • Cambridge table of contents page

• Hawking, S. W. and Ellis, G. F. R. The Large Scale Structure of Space-Time: 50th Anniversary Edition. Cambridge University Press, 2023. ISBN 978-1-009-25315-4. (Includes new foreword by Abhay Ashtekar and new preface by George Ellis; chapter structure is identical to the 1973 original.)

  • 50th Anniversary Edition on Cambridge

Background and overview

• Wikipedia article on The Large Scale Structure of Space-Time
• Semantic Scholar entry with citation data

Key foundational papers the book builds on

• Penrose, R. "Gravitational Collapse and Space-Time Singularities." Physical Review Letters 14 (1965): 57–59. (First singularity theorem for black holes.)
• Hawking, S. W. "The Occurrence of Singularities in Cosmology." Proceedings of the Royal Society A 294 (1966): 511–521.
• Hawking, S. W. and Penrose, R. "The Singularities of Gravitational Collapse and Cosmology." Proceedings of the Royal Society A 314 (1970): 529–548. (The general Hawking-Penrose theorem.)
• Raychaudhuri, A. K. "Relativistic Cosmology I." Physical Review 98 (1955): 1123–1126. (The Raychaudhuri equation.)

Differential geometry class notes based on Hawking-Ellis Chapter 2

• ETSU Differential Geometry Notes from Hawking and Ellis

Singularities and black holes — further reading

• Stanford Encyclopedia of Philosophy: Singularities and Black Holes
• Royal Society: A critical appraisal of the singularity theorems

Additional chapter summaries and study resources

These are secondary summaries and should be used alongside, rather than instead of, the original book.

• Stephen Hawking official site — book description
• AbeBooks — 50th Anniversary Edition listing with description