Building a hyperbolic 3-manifold the (G,X) way
Guide 2 told you that hyperbolic 3-space H^3 is the generic Thurston geometry — the atoroidal pieces of almost every closed 3-manifold wear it. Now we open the hood on exactly one of the eight geometries and see how H^3-manifolds are actually assembled. Take the model pair (G, X) from Guide 1 with X = H^3, hyperbolic 3-space, and G = its full isometry group. A hyperbolic 3-manifold is then a complete (G,X)-manifold of this kind: a manifold M that wears H^3 locally everywhere, and whose holonomy representation is especially well-behaved.
When such a structure is complete — recall from Guide 1 that this means the developing map is a diffeomorphism onto H^3 — the whole manifold is nothing more than a quotient. Concretely, M = H^3 / Gamma, where Gamma is the image of the holonomy: a subgroup of the isometry group of H^3 that is discrete (its elements do not accumulate) and torsion-free (no nonidentity element has finite order, so it acts freely with no fixed points). The discreteness makes the quotient a manifold rather than a crumpled mess; the torsion-freeness keeps it smooth. So the entire study of hyperbolic 3-manifolds becomes the study of such groups Gamma — a deep instance of the Erlangen-program spirit, where geometry is encoded in a group.
Here is the upper-half-space picture that makes G concrete and worth carrying through the whole guide. Model H^3 as the points (z, t) with z a complex number and t > 0, with the hyperbolic metric ds^2 = (|dz|^2 + dt^2) / t^2. The orientation-preserving isometries are then exactly the action of PSL(2,C) — two-by-two complex matrices of determinant 1, up to sign — extending the Mobius transformations of the boundary plane. So a discrete torsion-free subgroup Gamma of PSL(2,C) is the same data as an oriented hyperbolic 3-manifold. The fundamental group of M is exactly Gamma, and the manifold's geometry is read straight off the matrices.
The Margulis lemma: thick part, thin part
Before rigidity, you need to understand the shape of a hyperbolic 3-manifold, and there is one structural result that organizes everything: the Margulis lemma. Loosely, it says that wherever a hyperbolic manifold is locally thin — wherever there are short loops — the geometry is severely constrained: the short loops must nearly commute, so they generate an almost-abelian group, and the thin region can only look like one of a tiny menu of standard shapes. There is a universal constant, the Margulis constant epsilon(n) depending only on the dimension, that draws the line between 'thin' and 'thick'.
This hands you the thick-thin decomposition, the working anatomy of every finite-volume hyperbolic 3-manifold. Fix epsilon below the Margulis constant. The thin part is the set of points where the shortest noncontractible loop is shorter than epsilon; the thick part is everything else. The thin part is not arbitrary — it is a disjoint union of just two kinds of piece: tubes, solid-torus neighborhoods around short closed geodesics, and cusps, the flaring torus-times-half-line ends we just met. The thick part, meanwhile, is compact and has a diameter and volume bounded purely in terms of the dimension and epsilon.
Mostow-Prasad rigidity: topology decides geometry
Now the headline, and it should genuinely surprise you given Guide 1. There you saw that a single torus carries a whole continuum of distinct flat (G,X)-structures — change the lattice, change the structure, and you get the deformations that become Teichmuller space for surfaces. You might expect the same softness in dimension 3. You would be completely wrong. Mostow rigidity (extended to the finite-volume cusped case by Prasad) says: if M_1 and M_2 are finite-volume hyperbolic manifolds of dimension n >= 3 with isomorphic fundamental groups, then they are isometric — and moreover the isometry is realized by (is homotopic to) the map inducing the isomorphism.
dimension 2 (surfaces): pi_1 isomorphic =/=> isometric
--> a whole moduli of structures (Teichmuller)
dimension n >= 3 (Mostow): pi_1 isomorphic ==> ISOMETRIC
--> the structure is UNIQUE; no deformations
consequence: M hyperbolic, finite volume, n >= 3
==> hyperbolic volume vol(M), shortest geodesic length,
the whole length spectrum ... are TOPOLOGICAL invariantsRead that conclusion slowly, because it is the engine of the whole subject. The homotopy type — even just the fundamental group — of a hyperbolic 3-manifold determines its geometry completely. There is no continuous knob to turn, no moduli, no deformation: a given topological 3-manifold either admits no hyperbolic structure or admits exactly one. The payoff is immediate and astonishing: any geometric quantity you can measure — the hyperbolic volume, the length of the shortest closed geodesic, the full set of geodesic lengths, the symmetry group — is now a topological invariant, computable from and constraining the topology alone. Hyperbolic volume, in particular, becomes one of the most powerful invariants in low-dimensional topology.
Why it is true: rigidity comes from the boundary
You should not take a theorem this strong on faith, so here is the honest architecture of the proof — a survey, not the full argument, which is a serious course. The whole strategy is to push the isomorphism out to infinity. An isomorphism pi_1(M_1) -> pi_1(M_2) lifts to a map between the universal covers, both of which are H^3. That lift is generally a crude, distorted map — but it is a quasi-isometry: it preserves distances up to a bounded multiplicative-plus-additive error. The key fact about negatively curved spaces is that a quasi-isometry of H^3, however ugly in the interior, extends to a genuine, well-defined map on the sphere at infinity, the boundary 2-sphere of H^3.
On that boundary 2-sphere the map turns out to be a quasiconformal map — it distorts angles by only a bounded amount. (This is your first taste of the quasiconformal machinery that Guide 5 builds in full; here we just need its punchline.) The deep analytic step, due to Mostow, is an ergodicity argument: the boundary action of Gamma is so chaotically mixing that a quasiconformal map commuting with it cannot afford even infinitesimal distortion almost anywhere. The distortion is forced to vanish, so the boundary map is exactly conformal — a genuine Mobius transformation. A conformal map of the sphere at infinity extends to an isometry of H^3, and that isometry descends to the isometry M_1 -> M_2 we wanted. Distortion squeezed to zero on the boundary; rigidity in the interior.
- Start from an isomorphism pi_1(M_1) -> pi_1(M_2) of fundamental groups (the only input — purely topological).
- Lift it to a quasi-isometry of the universal covers H^3 -> H^3; in negative curvature this is automatic from the group isomorphism.
- Extend the quasi-isometry to the boundary 2-sphere; on the boundary it is a quasiconformal map distorting angles by only a bounded amount.
- Run Mostow's ergodicity argument: the chaotic boundary action forces the quasiconformal distortion to vanish almost everywhere, so the boundary map is conformal.
- A conformal boundary map extends to an isometry of H^3 that descends to the sought isometry M_1 -> M_2 — the structures coincide.
Two honest caveats about the hypotheses, since the slogan 'topology determines geometry' over-promises if you drop a condition. First, the dimension must be at least 3: the argument fails completely for surfaces, where H^2 has a circle at infinity and quasiconformal maps of a circle are far too flexible — that flexibility is precisely Teichmuller space, and it is no accident that Guide 4 is about surfaces. Second, finite volume is essential: infinite-volume hyperbolic 3-manifolds (think of an infinite handlebody) deform freely and are not rigid; their deformation theory is a rich subject of its own. State the hypotheses, never the slogan alone.
What rigidity buys, and where the rung turns
The practical upshot is a dictionary between topology and geometry that simply has no analogue in lower dimensions. Because hyperbolic volume is a topological invariant, the set of all hyperbolic volumes of 3-manifolds becomes a meaningful object — and Thurston and Jorgensen showed it is a well-ordered subset of the reals, with the figure-eight knot complement realizing one of the smallest volumes. Mostow rigidity is also what licenses computer programs (the famous SnapPea / SnapPy) to recognize a hyperbolic 3-manifold from its triangulation: since the geometry is unique, computing it once and comparing invariants is a legitimate classification strategy — exactly the kind of structural recognition Guide 2 warned that geometrization alone does not hand you.
Resist one tempting over-reading, though. Rigidity does not say hyperbolic 3-manifolds are boring or all alike — quite the opposite. It says each one is a sharp, individual crystal: no two distinct ones can be deformed into each other, so the landscape is a vast discrete museum of utterly rigid specimens, not a continuous family. And it does not say recognizing them is easy in practice; deciding whether two triangulations give the same manifold is computationally hard even granted that the geometry is unique. Rigidity gives you a well-defined fingerprint, not a fast one.
So we end on the great hinge of this rung. Rigidity is a dimension-3-and-up phenomenon; it leans entirely on H^2 having a circle at infinity that is too floppy, while H^3 has a sphere that is just rigid enough. Flip that observation around and you get the next two guides: because surfaces are NOT rigid, their hyperbolic structures form a rich continuous space. Guide 4 builds that space — Teichmuller space — together with the mapping class group acting on it, and Guide 5 develops the quasiconformal maps you just glimpsed into the full analytic engine and the Weil-Petersson geometry of moduli. Rigidity in dimension 3 and flexibility in dimension 2 are two faces of one coin; you have now seen the rigid face up close.