Singularity Analysis in Dimension 3

Summary

This article resolves the two obstacles O2 (classification of ancient \(\kappa\)-solutions) and O3 (canonical neighbourhood theorem) from Article 01. It combines three ingredients: (i) the Hamilton–Ivey pinching, which in dimension 3 forces curvature to become asymptotically non-negative; (ii) Perelman's \(\kappa\)-non-collapsing theorem, providing volume lower bounds; and (iii) the reduced \(\mathcal{L}\)-geometry, which identifies blow-up limits as \(\kappa\)-solutions. The result is the classification of ancient \(\kappa\)-solutions in dimension 3 and the canonical neighbourhood theorem: at every point of sufficiently large scalar curvature the flow is, up to \(\varepsilon\), modelled on one of three standard pieces – a neck, a cap, or a spherical space form.

1. Hamilton–Ivey pinching: positive curvature wins

On a closed 3-manifold the Riemann tensor has only three independent eigenvalues \(\lambda_1 \leq \lambda_2 \leq \lambda_3\) of the curvature operator. Hamilton 1995 / Ivey 1993 proved:

Pinching estimate. For every Ricci flow on a closed 3-manifold there is a function \(\phi:[0,\infty)\to\mathbb{R}\) with \(\phi(s)/s \to 0\) as \(s\to\infty\) such that at every space-time point $$ \lambda_1 \;\geq\; -\phi(\lambda_3). $$

In words: the most negative curvature direction is only logarithmically worse than the most positive. In particular, as curvature blows up every sectional curvature becomes non-negative in the limit. This is why singularity analysis in dimension 3 – and only there – can be carried out using the theory of ancient solutions with non-negative curvature.

2. Ancient \(\kappa\)-solutions: the list of models

An ancient solution is a Ricci flow \((M, g(t))_{t \in (-\infty, 0]}\) with bounded curvature on finite intervals. It is \(\kappa\)-non-collapsed if at every scale \(r\) $$ \frac{\mathrm{Vol}(B(p,t,r))}{r^n} \;\geq\; \kappa \quad\text{whenever}\quad |\mathrm{Rm}|\leq r^{-2}. $$

For blow-up limits Hamilton–Ivey additionally gives \(\mathrm{Rm}\geq 0\).

Classification (Perelman 0211159 §11). Every ancient \(\kappa\)-solution in dimension 3 with non-negative curvature operator is one of the following:

1. the round cylinder \(S^2 \times \mathbb{R}\) or its \(\mathbb{Z}_2\)-quotient;
2. a shrinking spherical quotient \(S^3/\Gamma\);
3. an ancient non-cylindrical model with non-compact topology \(\mathbb{R}^3\), asymptotically cylindrical (Bryant-type).

Proof idea: Hamilton–Harnack + Toponogov comparison force the sectional curvature at infinity to be locally dominated by the cylinder; the \(\mathcal{L}\)-geometry shows that every asymptotic soliton is one of these models; \(\kappa\)-non-collapsing rules out "cigars" (Hamilton's 2D soliton) and hence a fourth class.

3. Geometric building blocks: neck, cap, space form

From the classification Perelman extracts three local model geometries against which every high-curvature region is measured.

Model	Local form	Scale
\(\varepsilon\)-neck	\(\varepsilon\)-close to \(S^2 \times [-\varepsilon^{-1}, \varepsilon^{-1}]\), round \(S^2\)	\(r = R^{-1/2}\)
\(\varepsilon\)-cap	Diffeomorphic to \(D^3\) or \(\mathbb{RP}^3 \setminus \overline{B}\), with an \(\varepsilon\)-neck at the open end	\(r\)
Spherical space form	\(S^3/\Gamma\) with almost constant positive sectional curvature	whole component

Here \(R\) is the scalar curvature at the centre of the neck. The precise definition (Cao–Zhu, Morgan–Tian) requires \(\varepsilon\)-closeness in the \(C^{[1/\varepsilon]}\)-topology after parabolic rescaling.

4. The canonical neighbourhood theorem

Theorem (Perelman 0211159 §12.1). Given \(\varepsilon > 0\) there exists a function \(r_0(t) > 0\) such that: every point \((x,t)\) in a Ricci flow on a closed 3-manifold with \(R(x,t) \geq r_0(t)^{-2}\) lies in a canonical neighbourhood, i.e. after rescaling by \(R(x,t)\) it is \(\varepsilon\)-close to one of the three models (neck, cap, spherical space form).

Sketch of proof:

1. Argue by contradiction: suppose there is a sequence \((x_i, t_i)\) with \(R(x_i,t_i) \to \infty\) whose rescaling fits no model.
2. Hamilton compactness (volume lower bound from \(\kappa\)-non- collapsing, curvature bound from pinching) yields a Cheeger– Gromov subsequential limit.
3. Thanks to \(\mathcal{L}\)-geometry the limit is an ancient \(\kappa\)-solution with \(\mathrm{Rm}\geq 0\) – hence by §2 one of the three models.
4. Contradiction.

The proof uses every tool from Act 2 jointly: without entropy no \(\kappa\)-non-collapsing (step 2); without reduced length no convergence to an asymptotic soliton (step 3); without Hamilton–Ivey no non-negative curvature operator in the limit.

5. Local consequence: high curvature is canonical

Two direct corollaries structure the flow shortly before a singularity:

5.1 Structural decomposition. On every time slice \(t\) just below a singular time \(T\) the region \(\{R(\cdot,t) \geq r_0(t)^{-2}\}\) splits into a disjoint union of \(\varepsilon\)-necks, \(\varepsilon\)-caps and entire spherical components.

5.2 Volume control. The global volume \(\mathrm{Vol}(M, g(t))\) remains controlled up to \(T\) by the pinching estimate; in particular necks contribute small volume.

Both statements are the geometric prerequisite for defining surgery in the next article: necks are localised, their central \(S^2\) is explicit, and what survives a cut is topologically controlled.

6. What surgery inherits from this analysis

§3 + §4 together yield the local cutting plan:

• Every \(\varepsilon\)-neck has a unique central \(S^2\).
• The two components produced by a cut along that \(S^2\) each have an \(\varepsilon\)-cap as their starting end.
• Spherical components (\(S^3/\Gamma\)) become extinct in finite time and may be discarded.

What remains open: the global choice of constants \((\delta(t), h(t), r(t))\), the definition of the standard solution (model used to fill in), and the preservation of \(\kappa\)-non-collapsing + pinching after each surgery step. That is the content of Article 03.

7. Summary: obstacle → resolution

Obstacle from Art. 01	Tool	Result
O2 (classify \(\kappa\)-solutions)	Hamilton–Ivey + \(\mathcal{L}\)-geom.	§2 (3 models)
O3 (canonical neighbourhood)	Hamilton compactness + §2	§4 (theorem)
Preparation for O4 (surgery)	§3 + §5	local cutting plan

Act 3 is therefore fully prepared for the third article: we know where to cut and what survives the cut.

Sources

• G. Perelman, The entropy formula for the Ricci flow and its geometric applications, §§11–12, arXiv:math/0211159.
• R. Hamilton, The formation of singularities in the Ricci flow, Surveys in Differential Geometry II (1995), 7–136.
• T. Ivey, Ricci solitons on compact three-manifolds, Differential Geom. Appl. 3 (1993), 301–307. (Early form of the pinching estimate.)
• J. Morgan, G. Tian, Ricci Flow and the Poincaré Conjecture, Clay Math. Monographs 3, AMS 2007, ch. 9 (standard solutions), ch. 10 (canonical neighbourhoods).
• B. Kleiner, J. Lott, Notes on Perelman's papers, §§40–53, Geom. Topol. 12 (2008).
• H.-D. Cao, X.-P. Zhu, A complete proof of the Poincaré and geometrization conjectures, Asian J. Math. 10 (2006), §§5–6.
• B. Chow, P. Lu, L. Ni, Hamilton's Ricci Flow, AMS GSM 77, 2006. (Detailed proof of Hamilton–Ivey.)