Skip to content

Long-time Behavior and Thin–Thick Decomposition

"The thick part has bounded geometry and consists of hyperbolic pieces, while the thin part collapses with bounded curvature." — Perelman, Ricci flow with surgery on three-manifolds, arXiv:math/0303109, §§6–7

In Article 03 we showed that the Ricci flow with surgery exists on \([0, \infty)\) for every closed, oriented 3-manifold. With this the flow is rescued as an analytic object – but we still do not know what the manifold becomes as \(t \to \infty\). This article follows Perelman 0303109 §§6–7 and shows that the flow asymptotically falls into identifiable pieces: hyperbolic components and so-called graph manifolds – the last piece of the geometrization puzzle.

1. Rescaling and the right scale

Looking directly at \(g(t)\) as \(t \to \infty\) is meaningless: on a hyperbolic piece the diameter grows linearly and the curvature decays like \(1/t\). The correct quantity is the rescaled metric

\[ \hat g(t) := \frac{1}{4t}\, g(t). \]

The factor \(1/(4t)\) is chosen so that a constant hyperbolic metric of sectional curvature \(-1/(4t)\) becomes stationary under \(\hat g(t)\). The Ricci-flow system is thereby turned into an asymptotic soliton flow whose fixed points are exactly hyperbolic metrics of sectional curvature \(-1/4\).

2. The thin–thick decomposition

Let \(w > 0\) be a small parameter. For each time \(t > 0\) define

\[ M_{\text{thick}}(w, t) = \{\, x \in M_t \ \big| \ \mathrm{vol}(B_{\hat g(t)}(x, 1)) \ge w \,\}, \qquad M_{\text{thin}}(w, t) = M_t \setminus M_{\text{thick}}(w, t). \]

On the thick part the volume of a unit ball is bounded below – by the \(\kappa\)-non-collapse theorem the curvature is then \(C^k\)-controlled and Cheeger–Gromov convergence arguments apply. On the thin part the volume collapses without the curvature blowing up: precisely the kind of collapse forbidden locally by \(\kappa\)-non-collapse is allowed here globally.

3. Convergence of the thick pieces to hyperbolic geometry

Theorem (Perelman 0303109 §7.3, hyperbolization of the thick part). For every sequence \(t_i \to \infty\) there is a subsequence, a finite collection of complete hyperbolic 3-manifolds of finite volume \((H_1, h_1), \dots, (H_k, h_k)\), and a sequence of smooth maps $$ \varphi_i : H_1 \sqcup \dots \sqcup H_k \to M_{t_i} $$ such that \(\varphi_i^* \hat g(t_i) \to h_1 \sqcup \dots \sqcup h_k\) in \(C^\infty_{\text{loc}}\).

The images \(\varphi_i(H_j)\) are called persistent hyperbolic pieces. They are bounded by embedded incompressible 2-tori in \(M_t\) – tori whose fundamental group injects into \(\pi_1(M)\). These tori become the cutting surfaces of the JSJ decomposition in Act 3 (cf. geometrization conjecture).

4. The thin part: collapse with bounded curvature

The thin part is analytically more dramatic. Here Perelman invokes a theorem from the Cheeger–Fukaya–Gromov theory of collapse:

Theorem (Perelman 0303109 §7.4, collapsing theorem). There is \(w_0 > 0\) such that for \(w < w_0\) and all sufficiently large \(t\) the thin part \(M_{\text{thin}}(w, t)\) is a graph manifold, i.e. a 3-manifold that decomposes along embedded incompressible tori into Seifert-fibered pieces.

Perelman only sketched the full proof; it was completed in two independent works:

  • Shioya & Yamaguchi (2005), Volume collapsed three-manifolds with a lower curvature bound. Math. Ann. 333.
  • Kleiner & Lott (2014), Locally collapsed 3-manifolds. Astérisque 365 – today's canonical proof, with no lower curvature bound assumed.

5. Putting it together: geometrization

Combining the thick (hyperbolic) and thin (Seifert-fibered) pieces with the components removed by surgery in Article 03 as spherical space forms yields:

Decomposition layer Originates from
prime decomposition neck cuts of surgery
spherical space forms \(S^3/\Gamma\) discarded components
hyperbolic pieces persistent thick part
Seifert-fibered pieces thin part with bounded curvature
JSJ tori boundary tori between thick and thin

This is – piece by piece – exactly Thurston's geometrization conjecture (cf. Act 1, Article 05). The main theorem is therefore proved for every closed, oriented 3-manifold.

6. What does not follow immediately

The proof sketched here gives geometrization with possibly infinitely many surgeries on \([0, \infty)\). For the Poincaré conjecture a much shorter argument suffices: if \(M\) is simply connected, the flow becomes extinct in finite time. This extinction theorem is Perelman's third preprint 0307245; we treat it in Article 05: finite extinction. Only with that is the Poincaré conjecture provable without the full thin–thick machinery (although geometrization will still need it).

7. Which obstacles fall now

Obstacle (cf. Article 01) Resolution in this article
O4: long-time existence is not enough rescaled metric \(\hat g = g/(4t)\)
O4': convergence on hyperbolic pieces Cheeger–Gromov on \(M_{\text{thick}}\)
O4'': what happens on \(M_{\text{thin}}\)? collapsing theorem → Seifert-fibered graph manifolds
O5 (partial): read topology off the limit thick = hyperbolic, thin = Seifert, boundaries = JSJ tori

Cross-references

Sources

  • Perelman, G. (2003). Ricci flow with surgery on three-manifolds. arXiv:math/0303109, §§6–7.
  • Morgan, J. & Tian, G. (2014). The Geometrization Conjecture. CMI/AMS – worked-out long-time argument.
  • Kleiner, B. & Lott, J. (2008). Notes on Perelman's papers. Geom. Topol. 12, §§90–93.
  • Kleiner, B. & Lott, J. (2014). Locally collapsed 3-manifolds. Astérisque 365 – collapsing theorem without lower curvature bound.
  • Shioya, T. & Yamaguchi, T. (2005). Volume collapsed three-manifolds with a lower curvature bound. Math. Ann. 333.
  • Cao, H.-D. & Zhu, X.-P. (2006). A complete proof of the Poincaré and geometrization conjectures. Asian J. Math. 10, §§7.5–7.7.