What it is
Every closed orientable 3-manifold splits uniquely as a connected sum of prime pieces (Kneser–Milnor). Each prime piece splits further along incompressible tori (the Jaco–Shalen–Johannson decomposition). Thurston conjectured in 1982 that each resulting piece admits a complete, finite-volume, locally homogeneous metric modelled on one of eight geometries: the three constant-curvature geometries, two product geometries, and three twisted ones. Scott's survey is the standard description of them.
Thurston proved the conjecture for Haken manifolds. The general case — including every closed manifold with finite fundamental group, and hence the Poincaré conjecture — remained open until Ricci flow.
How Ricci flow proves it
Run Ricci flow with surgery for all time and rescale by t.
- Thick part. Where the manifold is locally non-collapsed, the rescaled metric converges to hyperbolic metrics of finite volume. Hamilton had analysed this in his work on non-singular solutions; Perelman made it work with surgery.
- Thin part. Where it collapses with curvature bounded below, collapsing theory shows the topology is a graph manifold, made of Seifert-fibred pieces glued along tori.
- The boundary between them. The tori separating the two parts are incompressible.
Together these give the geometric decomposition. The collapsing step was worked out independently by Shioya–Yamaguchi, Morgan–Tian, Kleiner–Lott, Bessières–Besson–Boileau–Maillot–Porti, and others.
Where it stands
Geometrization is a theorem, and it underpins modern 3-manifold topology. Agol and Wise's work on virtually Haken and virtually fibred hyperbolic 3-manifolds, and the classification of 3-manifold groups, all build on it. On the Ricci flow side, Bamler proved that in the long-time picture only finitely many surgeries occur and curvature decays like t−1. For the topology, see the 3-manifolds topic.