© 2026 NeckPinch · www.neckpinch.com · All rights reserved.
Course 10Book 10A: Three-Manifolds and GeometrizationChapter 1
A Zoo of Three-Manifolds
The 3-sphere, lens spaces, the Poincaré sphere, the Hopf fibration and the shape of space.
Read with Thurston's Three-Dimensional Geometry and Topology, Volume 1, chapter 1 (what is a manifold? gluing polyhedra, the 3-torus and the dodecahedral spaces), and Weeks's The Shape of Space for an accurate popular account. Hatcher's Notes on Basic 3-Manifold Topology is the reference for the rest of this book.
Book 7A stated the Poincaré conjecture: a closed, simply connected 3-manifold is the 3-sphere. Book 10A asks the broader question that Thurston's geometrization conjecture answers: what are all the closed 3-manifolds, and what do they look like? The Ricci flow with surgery is a machine that takes any closed 3-manifold and returns pieces. This book describes the pieces, how they fit together and the geometries they carry, so that when the machine runs in Books 11A–12C you know what it must produce.
This first chapter is a zoo: the 3-sphere and its Hopf fibration, the 3-torus, , lens spaces and the other spherical space forms, the Poincaré homology sphere in three guises, and the Seifert–Weber space. Each will reappear as an example of a piece, a geometry or a fate under the flow.
By the end of this chapter you will be able to:
- describe as the unit quaternions and as two solid tori, and compute the fibres of the Hopf fibration;
- build the 3-torus, and the lens spaces as quotients, and compute their fundamental groups;
- state which groups act freely on , and recognise spherical space forms;
- construct the Poincaré homology sphere and the Seifert–Weber space by gluing the faces of a dodecahedron;
- explain how one could test whether the universe is a nontrivial 3-manifold.
The shape of space
On large scales, space is close to flat (9A.6 The Laplacian and the Bochner Formula), but flatness does not decide topology: a flat universe could be infinite , or a finite 3-torus, or one of the other closed flat 3-manifolds (10A.5 Thurston’s Eight Geometries). If space were finite and smaller than the distance light has travelled since recombination, light from the same region would reach us from different directions. The sphere from which the cosmic microwave background was emitted would intersect its own translated copies, and the sky would contain pairs of matched circles with the same temperature pattern. Neil Cornish, David Spergel and Glenn Starkman proposed this test in 1998. In 2003 Jean-Pierre Luminet and colleagues suggested that a deficit of large-scale power in the WMAP data might be explained if space were the Poincaré dodecahedral space, a positively curved quotient of .
Later searches did not find matched circles. The Planck collaboration ("Planck 2015 results XVIII: Background geometry and topology of the Universe", Astronomy & Astrophysics, 2016) found no evidence of nontrivial topology. For a cubic 3-torus, it showed that the largest ball fitting in the fundamental domain must have radius at least of the distance to the last-scattering surface (99% confidence). If the universe is a nontrivial 3-manifold, it is too large for us to see its topology. Even so, this is a genuine experiment asking which closed 3-manifold we inhabit.
The 3-sphere and the Hopf fibration
The unit sphere has several faces (7A.2 Compactness and Compactification, 8A.5 Lie Groups and Group Actions): the one-point compactification of , via stereographic projection; the group of unit quaternions; and the union of two solid tori and glued along the torus , with the meridian of each glued to the longitude of the other.
The Hopf fibration is the map
Its fibres are the circles : each is the intersection of with a complex line through the origin, hence a great circle (Exercise 1.1). So is a union of disjoint great circles, one for each point of , and any two of them are linked once. Under stereographic projection to the picture is famous (Figure 1.1): one fibre is a vertical line through infinity, one is the unit circle around it, and the rest fill nested tori, each torus made of circles that wind once around each way. The Berger spheres of 9A.5 Computing Curvature and 9B.3 Collapsing and Noncollapsing are with these circles shrunk.
In quantum computing, the state of a qubit is a unit vector , : a point of . Two state vectors that differ by an overall phase describe the same physical state, so the physical states form the quotient of by the circle action, which is , the Bloch sphere. The map from state vectors to points of the Bloch sphere is exactly the Hopf map, and the fibres are the phase circles. The third coordinate is the difference of the two measurement probabilities.
Flat, product and quotient examples
The 3-torus , a cube with opposite faces glued. It has a flat metric and . Someone living inside a small 3-torus would see copies of themselves repeated in a cubic lattice in every direction (Figure 1.2), which is what the matched-circles test looks for on a cosmic scale.
, with . It is the prime manifold that is not irreducible (10A.3 The Prime Decomposition): the sphere does not bound a ball. Its product metric is the neck geometry wrapped up.
Lens spaces. For coprime , the cyclic group generated by , , acts freely on by isometries, and the quotient is the lens space , with (7A.6 Covering Spaces). . Lens spaces are classified by Reidemeister's theorem (1935, extended from piecewise-linear equivalence to homeomorphism by Brody in 1960): and are homeomorphic if and only if . Homotopy equivalence is weaker, for some , so and are homotopy equivalent but not homeomorphic (Exercise 1.3): the fundamental group, even with homotopy type, does not determine a 3-manifold in general. The Poincaré conjecture is the claim that it does when the group is trivial.
Spherical space forms. More generally, a spherical space form is a quotient by a finite group acting freely. The possible groups were classified by Hopf (1926) and later work: cyclic groups, as for lens spaces, and products of cyclic groups with the binary dihedral, binary tetrahedral, binary octahedral and binary icosahedral groups (the preimages in of the rotation groups of regular solids), in suitable combinations. These are exactly the closed 3-manifolds with a metric of constant positive curvature (10A.5 Thurston’s Eight Geometries).
Mapping tori and knot complements. Gluing the ends of by a diffeomorphism of a surface gives the mapping torus of , a 3-manifold fibring over the circle. Removing an open tubular neighbourhood of a knot in gives a knot complement, a compact 3-manifold with torus boundary; the trefoil and figure-eight complements are the first examples of 10A.4 Seifert Spaces and the JSJ Decomposition and 10A.6 Hyperbolic Three-Manifolds.
Dodecahedral spaces
Take a solid regular dodecahedron and glue each face to the opposite face. A face is a pentagon, and the opposite face is rotated relative to it, so a twist is needed.
- Gluing with a turn () gives the Poincaré homology sphere, also described as , the quotient by the binary icosahedral group of order (7A.6 Covering Spaces), and as the link of the singularity (the Brieskorn sphere ). Its fundamental group is perfect and nontrivial, so it has the homology of but is not : the example that made Poincaré state his conjecture with instead of homology.
- Gluing with a turn () gives the Seifert–Weber space (Weber and Seifert, 1933), which carries a hyperbolic metric.
The edges of the dodecahedron fall into classes under the gluing, and the angles around each glued edge must add up to . In the Poincaré gluing the edges come in classes of , so the dihedral angle must be , slightly more than the Euclidean : the dodecahedron must be spherical, one of the cells of the regular 120-cell in . In the Seifert–Weber gluing they come in classes of , so the angle must be : the dodecahedron must be hyperbolic (Exercise 1.4). The gluing determines the geometry. This is the first glimpse of geometrization.
10A.2 Building Three-Manifolds shows how all closed 3-manifolds can be built (Heegaard splittings, surgery). 10A.3 The Prime Decomposition and 10A.4 Seifert Spaces and the JSJ Decomposition cut them along spheres and tori into pieces. 10A.5 Thurston’s Eight Geometries lists the eight geometries the pieces carry: and its quotients, (the 3-torus), , and five more, including hyperbolic geometry (the Seifert–Weber space) and Nil (circle bundles over tori). 10A.7 Geometrization and Ricci Flow says what the Ricci flow does to each.
History
Heinz Hopf introduced his fibration in 1931. Heinrich Tietze defined lens spaces in 1908; Kurt Reidemeister classified them in 1935. Poincaré constructed his homology sphere in 1904; its dodecahedral description is due to Weber and Seifert (1933), as is the hyperbolic space with the twist. Cornish, Spergel and Starkman proposed the circles-in-the-sky test in 1998; the Planck collaboration's topology analysis appeared in 2014 and 2016.
is the unit quaternions, the compactified , or two solid tori; the Hopf fibration fills it with linked great circles, one for each point of , and it is the map from qubit states to the Bloch sphere. The 3-torus is flat with ; has ; lens spaces are quotients of with , classified by ; spherical space forms are modulo finite groups acting freely. Gluing a dodecahedron's opposite faces with a turn gives the Poincaré homology sphere (spherical), with a turn the Seifert–Weber space (hyperbolic). Matched-circle searches test whether the universe is such a manifold, and have found none at observable scales. 10A.2 Building Three-Manifolds builds every closed 3-manifold from simple pieces.
Exercises
(a) Show that maps to the unit sphere in , and that if and only if . (b) Show that each fibre is a great circle. (c) The fibres over the poles are and ; show that after stereographic projection from they become the unit circle in the horizontal plane and the vertical axis, which are linked.
Solution
(a) . If agrees, then , and , which forces a common phase. (b) The fibre is , the unit circle of a real 2-plane. (c) Writing and projecting from , is the unit circle in , and is the circle , which projects to the -axis together with the point at infinity. The axis passes through the disc bounded by the unit circle once.
Compute of , , and from covering spaces (7A.6 Covering Spaces). Which of these are simply connected after passing to the universal cover, and what is the cover in each case?
Solution
: , cover . : cover , deck group . : , cover . : , cover . All universal covers are simply connected; only is compact among , , .
Using Reidemeister's criterion, show that and are not homeomorphic, and that and are. Using the homotopy criterion , show that and are homotopy equivalent.
Solution
For : , which does not contain ; and , so . For : , and the squares mod are , so ; but , so they are not homeomorphic.
A dodecahedron has vertices, edges and faces. In the Poincaré gluing, the vertices fall into classes of and the edges into classes of ; in the Seifert–Weber gluing, the vertices form class and the edges classes of . (a) Check that both have Euler characteristic , as every closed odd-dimensional manifold must. (b) The regular Euclidean dodecahedron has dihedral angle . Explain why the Poincaré gluing needs a spherical dodecahedron and the Seifert–Weber gluing a hyperbolic one.
Solution
(a) Poincaré: . Seifert–Weber: . (b) Around a glued edge class the dihedral angles must sum to : with edges each angle is , with edges . Spherical polyhedra have larger angles than Euclidean ones of the same shape and hyperbolic ones smaller, and the angle of a regular dodecahedron varies continuously with its size, from the Euclidean value at small size: up to (and beyond) in , down to (and below) in .
At the point , the fibre direction is and the horizontal vectors are , . Compute and show it has length . Deduce that is a Riemannian submersion onto the sphere of radius at this point (and, by symmetry, everywhere), which is why collapsing Berger spheres converge to (9B.3 Collapsing and Noncollapsing).
Solution
to first order, so , of length ; also , since is constant on fibres. So maps to , kills the fibre direction, and is an isometry on horizontal vectors. acts transitively on preserving the fibration and the metrics, so the same holds at every point.
© 2026 NeckPinch (www.neckpinch.com). All content in the guidebook (text, mathematics, figures and exercises) is protected by copyright. All rights reserved. No part may be copied, republished or redistributed without written permission.