Skip to content

Ricci Flow with Surgery

"We construct a solution which is a smooth Ricci flow on its time-of-definition, punctuated by a discrete set of surgery times at which the metric is modified." — Perelman, Ricci flow with surgery on three-manifolds, arXiv:math/0303109 (2003)

In Article 02 we understood the singularities of the Ricci flow in dimension 3 locally: every high-curvature region looks like a neck, a cap, or a spherical space form. In this article we follow Perelman's second preprint 0303109 and assemble these local pieces into a global flow with surgery that continues across each singularity – the technical heart of the proof of the geometrization conjecture.

1. The idea: cut just before the singularity

Hamilton's original program failed at obstacle O3 from Article 01: understanding singularities locally is not enough; one must operatively remove them, without the proof disintegrating into endless special cases.

Perelman's solution is strikingly simple in principle:

  1. Watch the flow until the curvature first exceeds a fixed threshold \(\Omega\) at some point (just before the singularity).
  2. Localize the high-curvature region using the canonical neighborhood theorem – it splits into necks, caps, and compact components that are spherical space forms.
  3. Cut along the central 2-sphere of every neck and glue in a standard solution (a fixed, explicit metric on \(D^3\)).
  4. Discard compact components already identified as spherical space forms – they are topologically known and contribute nothing further to the flow.
  5. Resume the Ricci flow on the modified manifold smoothly until the next singularity occurs.

The result is a sequence \((M_t, g(t))\) of smooth Ricci flows on time intervals \([t_{i-1}, t_i]\) separated by a discrete set of surgery times \(0 < t_1 < t_2 < \dots\).

2. δ-necks and the cut locus

A precise cut requires a neck that is "thin enough." Let \(\delta > 0\) be a small parameter.

Definition (δ-neck). A point \((x, t)\) lies in a \(\delta\)-neck of scale \(r\) if the parabolic rescaling of \(g(t)\) around \(x\) is \(\delta\)-close in \(C^{\lfloor 1/\delta \rfloor}\) to the round cylindrical model \(S^2_1 \times (-1/\delta, 1/\delta)\).

Lemma (existence of a \(\delta\)-neck, 0303109 §4). For all sufficiently small \(\delta\), there is a curvature bound \(h(\delta)\) such that every point with \(|\mathrm{Rm}|(x,t) \ge h(\delta)\) that lies in an \(\varepsilon\)-neck (in the sense of Act 2, Article 06) contains a central \(\delta\)-neck.

The cut locus is chosen on the central 2-sphere \(\Sigma\) of such a neck. The finer condition \(\delta \ll \varepsilon\) ensures that the neck is long and thin enough to splice in a standard solution without disturbing the future curvature evolution.

3. The standard solution

The standard solution is a fixed, complete, asymptotically cylindrical metric \(\bar g\) on \(\mathbb{R}^3\) with the following properties (0303109 §2, worked out in Cao–Zhu §7.3):

  • \(\bar g\) is rotationally symmetric and has positive sectional curvature.
  • Outside a large ball, \((\mathbb{R}^3, \bar g)\) is isometric to the round half-cylinder \(S^2_1 \times [0, \infty)\).
  • The Ricci flow with initial datum \(\bar g\) exists smoothly on \([0, 1)\), shrinks to a point at \(t = 1\), and is \(\kappa\)-non-collapsed.

The standard solution serves as a model cap: after the cut, the unwanted neck piece is replaced by a chunk of \(\bar g\), smoothly glued through a cut-off function. Existence and uniqueness of its flow is itself a technical theorem (Cao–Zhu, Kleiner–Lott Ch. 12).

4. The surgery algorithm

The construction depends on three parameter sequences chosen simultaneously:

Parameter Role
\(\varepsilon_i \to 0\) accuracy of the canonical neighborhood
\(\delta_i \to 0\) quality of the neck at the cut, \(\delta_i \ll \varepsilon_i\)
\(r_i \to 0\) scale at which canonical neighborhoods kick in
\(h_i \to 0\) surgery threshold, \(h_i \ll \delta_i r_i\)

Definition (Ricci flow with surgery). A family \(\{(M_t, g(t))\}_{t \ge 0}\) is a Ricci flow with \((r, \delta)\)-surgery if on each interval \([t_{i-1}, t_i)\) it satisfies \(\partial_t g = -2\,\mathrm{Ric}\) smoothly, surgery is performed at all \(\delta_i\)-necks at times \(t_i\), and the canonical neighborhood property with parameters \((\varepsilon_i, r_i)\) is preserved up to \(t_i\).

5. The surgery theorem (0303109 §5)

Theorem (Perelman, surgery exists globally). For every closed, oriented 3-manifold \((M, g_0)\) there exist sequences \(\varepsilon_i, \delta_i, r_i, h_i \to 0\) such that a Ricci flow with \((r, \delta)\)-surgery on \([0, \infty)\) exists.

The central difficulty is the inductive loop: to keep canonical neighborhoods alive at the step \(i \to i+1\) one needs \(\kappa\)-non-collapse despite the previous surgeries. Perelman shows:

  • Local \(\kappa\)-non-collapse (Article 06 in Act 2) persists under surgery, provided \(\delta_i\) is chosen small enough.
  • The number of surgeries on every finite time interval is finite, since each one removes a fixed amount of volume (\(\ge c\, h_i^3\)).

6. What surgery does not destroy

Surgery is designed to preserve every structure essential to the proof:

  • Hamilton–Ivey pinching (cf. Article 02) survives every surgery step because the standard solution itself has positive curvature.
  • κ-non-collapse survives with a smaller but positive constant \(\kappa'\).
  • Topological bookkeeping: every removed component is a spherical space form \(S^3/\Gamma\); every surgery replaces a neck \(S^2 \times I\) by two caps \(D^3\). Both are exactly the two standard moves of a prime decomposition (cf. geometrization conjecture).

The topology of the original manifold can therefore be reconstructed exactly from the removed pieces plus the long-time limit of the flow – the bridge built in Article 06.

7. What remains to show

With the surgery theorem the flow as an analytic object is rescued. Two questions remain for Act 3:

Summary

Obstacle (Act 3, Article 01) Resolution in this article
O3: remove singularities operatively surgery algorithm with \((\delta, r, h)\) parameters
O3': algorithm yields only finitely many cuts volume argument \(\ge c\, h^3\) per surgery
O3'': pinching/κ-non-collapse after surgery standard solution has positive curvature; local κ-non-collapse persists
O3''': topology stays trackable cuts = prime decomposition, removed pieces = \(S^3/\Gamma\)

Cross-references

Sources

  • Perelman, G. (2003). Ricci flow with surgery on three-manifolds. arXiv:math/0303109.
  • Morgan, J. & Tian, G. (2007). Ricci Flow and the Poincaré Conjecture. AMS, Ch. 13–17.
  • Kleiner, B. & Lott, J. (2008). Notes on Perelman's papers. Geom. Topol. 12, 2587–2855, §§57–72.
  • Cao, H.-D. & Zhu, X.-P. (2006). A complete proof of the Poincaré and geometrization conjectures. Asian J. Math. 10, §7.
  • Hamilton, R. (1997). Four-manifolds with positive isotropic curvature. Comm. Anal. Geom. 5 – original surgery idea in dim 4.