Book 11A

© 2026 NeckPinch · www.neckpinch.com · All rights reserved.

Course 11Book 11A: The Ricci Flow: Existence and Maximum PrinciplesChapter 8

Homogeneous Flows

A computable laboratory: the flow on Thurston’s geometries.

14 min read · Updated Oct 3, 2026

Read with Chow and Knopf's The Ricci Flow: An Introduction, chapter 1 (homogeneous solutions in dimension three), and the two papers it follows: Isenberg and Jackson, "Ricci flow of locally homogeneous geometries on closed manifolds" (Journal of Differential Geometry, 1992), and Knopf and McLeod, "Quasi-convergence of model geometries under the Ricci flow" (Communications in Analysis and Geometry, 2001).

In this chapter · 5 sections
  1. 8.1The ODE
  2. 8.2The geometries one by one
  3. 8.3Rescaled limits: collapse
  4. 8.4History
  5. 8.5Exercises

A left-invariant metric on a Lie group looks the same at every point, so its Ricci flow stays left-invariant (8A.6 Flows and the Lie Derivative): the flow reduces from a PDE to an ODE for finitely many numbers. For the three-dimensional unimodular groups of 9A.5 Computing Curvature, those numbers are the lengths of a Milnor frame, and Milnor's curvature formulas give the ODE explicitly. This chapter derives it once, in general, and solves it for each geometry: SU(2)SU(2), Nil, Sol, SL~(2,R)\widetilde{SL}(2, \mathbb{R}) and the Euclidean group. It is a laboratory in which the long-time behaviours predicted in 10A.7 Geometrization and Ricci Flow can be computed exactly: rounding and extinction, convergence to flat, and collapse with bounded or decaying curvature.

By the end of this chapter you will be able to:

  • derive the Ricci flow ODE for diagonal left-invariant metrics on a three-dimensional unimodular Lie group;
  • solve it exactly for Nil and describe its solutions for Sol, SL~(2,R)\widetilde{SL}(2, \mathbb{R}), SU(2)SU(2) and the Euclidean group;
  • read off which directions shrink and which expand, and what the rescaled limits are;
  • check the general estimates of 11A.4 Maximum Principles under Ricci Flow on these examples;
  • connect the behaviours to the roles of the geometries in geometrization.

The ODE

Let X1,X2,X3X_1, X_2, X_3 be a basis of left-invariant fields with [X2,X3]=c1X1[X_2, X_3] = c_1X_1, [X3,X1]=c2X2[X_3, X_1] = c_2X_2, [X1,X2]=c3X3[X_1, X_2] = c_3X_3 (Milnor's form, 9A.5 Computing Curvature), and consider metrics that are diagonal in this basis, g=diag⁡(A,B,C)g = \operatorname{diag}(A, B, C). The orthonormal frame e1=X1/Ae_1 = X_1/\sqrt A, and so on, has structure constants λ1=c1AABC\lambda_1 = \frac{c_1A}{\sqrt{ABC}}, λ2=c2BABC\lambda_2 = \frac{c_2B}{\sqrt{ABC}}, λ3=c3CABC\lambda_3 = \frac{c_3C}{\sqrt{ABC}}. Milnor's formula Ric⁡(e1,e1)=2μ2μ3\operatorname{Ric}(e_1, e_1) = 2\mu_2\mu_3 and A′=−2ARic⁡(e1,e1)A' = -2A\operatorname{Ric}(e_1, e_1) give (Exercise 8.1)

A′=−(c1A)2−(c2B−c3C)2BC,B′=−(c2B)2−(c3C−c1A)2CA,C′=−(c3C)2−(c1A−c2B)2AB.A' = -\frac{(c_1A)^2 - (c_2B - c_3C)^2}{BC}, \quad B' = -\frac{(c_2B)^2 - (c_3C - c_1A)^2}{CA}, \quad C' = -\frac{(c_3C)^2 - (c_1A - c_2B)^2}{AB}.

The flow keeps the metric diagonal, because the Ricci tensor is diagonal in Milnor's frame.

The geometries one by one

SU(2)SU(2), c=(2,2,2)c = (2, 2, 2): the sphere. With B=CB = C, this is the Berger flow of 11A.6 Hamilton’s 1982 Theorem: A′=−4A2B2A' = -\frac{4A^2}{B^2}, B′=−4(2−AB)B' = -4(2 - \frac AB). The metric becomes round as it shrinks to a point in finite time.

The Euclidean group, c=(1,1,0)c = (1, 1, 0). The metric with A=BA = B is flat. From any diagonal metric the flow converges to a flat one, with A−B→0A - B \to 0, while the curvature decays.

Nil, c=(1,0,0)c = (1, 0, 0). The equations are A′=−A2BCA' = -\frac{A^2}{BC}, B′=ACB' = \frac AC, C′=ABC' = \frac AB. With B=CB = C, the product ABAB is constant (Exercise 8.2), and

B(t)=(B03+3A0B0t)1/3,A(t)=A0B0B(t).B(t) = \big(B_0^3 + 3A_0B_0t\big)^{1/3}, \qquad A(t) = \frac{A_0B_0}{B(t)}.

The fibre direction X1X_1 shrinks like t−1/3t^{-1/3} and the base directions grow like t1/3t^{1/3}. The curvature decays like 1t\frac1t.

Sol, c=(1,−1,0)c = (1, -1, 0). The equations are A′=−A2−B2BCA' = -\frac{A^2 - B^2}{BC}, B′=A2−B2ACB' = \frac{A^2 - B^2}{AC}, C′=(A+B)2ABC' = \frac{(A + B)^2}{AB}. The difference A−BA - B decays (Exercise 8.3), and once A=BA = B, C=C0+4tC = C_0 + 4t: the two directions of the torus fibre stay bounded while the circle direction grows linearly.

SL~(2,R)\widetilde{SL}(2, \mathbb{R}), c=(−2,2,2)c = (-2, 2, 2). With B=CB = C: A′=−4A2B2A' = -\frac{4A^2}{B^2}, B′=8+4ABB' = 8 + \frac{4A}{B}. The base directions grow like 8t8t, while the fibre direction AA converges to a positive constant.

Figure 8.1. The metric components under the Ricci flow on Nil, Sol and SL~(2,R)\widetilde{SL}(2, \mathbb{R}), on logarithmic axes (computed with a fourth-order Runge–Kutta integration from A=B=C=1A = B = C = 1, and checked against the exact Nil solution). The slopes show the power laws: Nil t−1/3t^{-1/3} and t1/3t^{1/3}; Sol t0t^0 and t1t^1; SL~(2,R)\widetilde{SL}(2, \mathbb{R}) t0t^0 and t1t^1.

Rescaled limits: collapse

Divide each metric by tt, as Perelman does for long-lived flows (10A.7 Geometrization and Ricci Flow):

geometry behaviour of (A,B,C)(A, B, C) g(t)t\frac{g(t)}{t} as t→∞t \to \infty role in geometrization
SU(2)SU(2) rounds, extinct at finite TT (extinct) spherical pieces disappear
Euclidean group converges to flat collapses to a point flat pieces are thin
Nil t−1/3t^{-1/3}, t1/3t^{1/3}, t1/3t^{1/3} collapses to a point graph-manifold (thin) pieces
Sol 11, 11, tt collapses to a line thin, Sol torus bundles
SL~(2,R)\widetilde{SL}(2, \mathbb{R}) 11, tt, tt fibre collapses; base →\to hyperbolic plane Seifert (thin) pieces

Every geometry except SU(2)SU(2) collapses after rescaling: some directions become negligible compared with t\sqrt t, the volume of g(t)t\frac{g(t)}{t}-balls of unit radius tends to zero, and the curvature of g(t)t\frac{g(t)}{t} stays bounded. This is the thin part of 10A.7 Geometrization and Ricci Flow and 12C.4 Geometrization, computed exactly. The hyperbolic geometry, which is not a Lie group geometry, is the one that does not collapse: g(t)t→4g0\frac{g(t)}{t} \to 4g_0 (10A.6 Hyperbolic Three-Manifolds). Isenberg and Jackson classified these behaviours in 1992; Knopf and McLeod (2001) described the convergence of the normalised flows in detail, including on non-unimodular groups.

Where this goes From homogeneous to general

Homogeneous flows have no singularities other than global extinction, because curvature is the same everywhere. General flows are not homogeneous, and singularities form at points. Book 11B develops the tools to study them: solitons, the Harnack inequality, compactness of flows, and the blow-up method.

History

Milnor's curvature formulas date from 1976. James Isenberg and Martin Jackson studied the Ricci flow of locally homogeneous metrics on closed 3-manifolds in 1992, and Dan Knopf and Kevin McLeod refined the picture in 2001. Lott later studied the long-time limits of these flows as geometric objects ("On the long-time behavior of type-III Ricci flow solutions", 2007).

Recall Book 11A in one paragraph

The Ricci flow ∂tg=−2Ric⁡\partial_tg = -2\operatorname{Ric} is a heat equation for the metric, invariant under scaling and diffeomorphisms; spheres shrink, hyperbolic metrics expand, cylinders pinch (11A.1 The Equation and Its First Solutions). Its curvature evolves by reaction–diffusion equations, ∂tR=ΔR+2∣Ric⁡∣2\partial_tR = \Delta R + 2|\operatorname{Ric}|^2 and ∂tM=ΔM+M2+M#\partial_t\mathcal M = \Delta\mathcal M + \mathcal M^2 + \mathcal M^\#, whose 3D reaction is the ODE λ′=λ2+μν\lambda' = \lambda^2 + \mu\nu (11A.2 How Curvature Evolves). DeTurck's trick gives short-time existence and uniqueness, Shi's estimates control derivatives, and the flow lasts while curvature is bounded (11A.3 Short-Time Existence and Uniqueness). Maximum principles, scalar, tensor and ODE–PDE, preserve convex ODE-invariant sets (11A.4 Maximum Principles under Ricci Flow); Hamilton–Ivey makes 3D singularities nonnegatively curved (11A.5 Hamilton–Ivey Pinching). Positive Ricci curvature in 3D flows to constant curvature (11A.6 Hamilton’s 1982 Theorem). On surfaces the flow proves uniformization, with Hamilton's entropy and Harnack inequality (11A.7 Ricci Flow on Surfaces). Homogeneous flows reduce to ODEs: spheres round out, the other geometries collapse (this chapter).

Where this goes Into Book 11B

Hamilton's 1982 theorem worked because positive Ricci curvature allowed only one kind of singularity, a round point. For general initial metrics, necks pinch and more complicated singularities form. Book 11B develops Hamilton's tools for studying them: solitons as the self-similar models (11B.1 Ricci Solitons), the Harnack inequality (11B.2 Ancient Solutions and the Harnack Inequality), compactness of sequences of flows (11B.3 Compactness of Ricci Flows), and the blow-up analysis of singularities (11B.4 Singularities). It ends where Hamilton's program stood in 2002, with one missing ingredient (11B.5 Hamilton’s Program in 2002).

Exercises

Exercise 8.1 Deriving the ODE

From μ2=λ1−λ2+λ32\mu_2 = \frac{\lambda_1 - \lambda_2 + \lambda_3}{2} and μ3=λ1+λ2−λ32\mu_3 = \frac{\lambda_1 + \lambda_2 - \lambda_3}{2} with λi\lambda_i as in the text, show Ric⁡(e1,e1)=2μ2μ3=(c1A)2−(c2B−c3C)22ABC\operatorname{Ric}(e_1, e_1) = 2\mu_2\mu_3 = \frac{(c_1A)^2 - (c_2B - c_3C)^2}{2ABC}, and deduce the equation for A′A'. Check it on the round S3S^3, c=(2,2,2)c = (2, 2, 2), A=B=C=1−4tA = B = C = 1 - 4t.

Solution

2μ2μ3=12(λ1−(λ2−λ3))(λ1+(λ2−λ3))=12(λ12−(λ2−λ3)2)=(c1A)2−(c2B−c3C)22ABC2\mu_2\mu_3 = \frac12(\lambda_1 - (\lambda_2 - \lambda_3))(\lambda_1 + (\lambda_2 - \lambda_3)) = \frac12(\lambda_1^2 - (\lambda_2 - \lambda_3)^2) = \frac{(c_1A)^2 - (c_2B - c_3C)^2}{2ABC}. Then A′=−2Ric⁡(X1,X1)=−2A⋅(c1A)2−(c2B−c3C)22ABCA' = -2\operatorname{Ric}(X_1, X_1) = -2A\cdot\frac{(c_1A)^2 - (c_2B - c_3C)^2}{2ABC}. Round: A′=−4A2A2=−4A' = -\frac{4A^2}{A^2} = -4.

Exercise 8.2 Nil exactly

For c=(1,0,0)c = (1, 0, 0) and B=CB = C, show (AB)′=0(AB)' = 0 and (B3)′=3AB(B^3)' = 3AB, and derive the solution in the text. Compute the scalar curvature R=−A2B2R = -\frac{A}{2B^2} (from Milnor's formulas) and show R(t)∼−16tR(t) \sim -\frac{1}{6t} as t→∞t \to \infty.

Solution

(AB)′=A′B+AB′=−A2B+A2B=0(AB)' = A'B + AB' = -\frac{A^2}{B} + \frac{A^2}{B} = 0, and (B3)′=3B2⋅AB=3AB(B^3)' = 3B^2\cdot\frac AB = 3AB, constant, so B3=B03+3A0B0tB^3 = B_0^3 + 3A_0B_0t. Ricci in the orthonormal frame is (A2B2,−A2B2,−A2B2)(\frac{A}{2B^2}, -\frac{A}{2B^2}, -\frac{A}{2B^2}), so R=−A2B2=−A0B02B3∼−A0B06A0B0t=−16tR = -\frac{A}{2B^2} = -\frac{A_0B_0}{2B^3} \sim -\frac{A_0B_0}{6A_0B_0t} = -\frac{1}{6t}.

Exercise 8.3 Sol

Show (A−B)′=−(A−B)(A+B)2ABC(A - B)' = -\frac{(A - B)(A + B)^2}{ABC}, so ∣A−B∣|A - B| decreases. With A=BA = B, show C′=4C' = 4 and R=−2CR = -\frac2C, and check the lower bound R≥−32tR \geq -\frac{3}{2t} of 11A.4 Maximum Principles under Ricci Flow.

Solution

A′−B′=−A2−B2BC−A2−B2AC=−(A2−B2)(A+B)ABCA' - B' = -\frac{A^2 - B^2}{BC} - \frac{A^2 - B^2}{AC} = -\frac{(A^2 - B^2)(A + B)}{ABC}. With A=BA = B: C′=4A2A2=4C' = \frac{4A^2}{A^2} = 4. Ricci: Ric⁡(e1)=Ric⁡(e2)=0\operatorname{Ric}(e_1) = \operatorname{Ric}(e_2) = 0 and Ric⁡(e3)=−(A+B)22ABC=−2C\operatorname{Ric}(e_3) = -\frac{(A + B)^2}{2ABC} = -\frac2C, so R=−2C0+4t≥−24t=−12t>−32tR = -\frac{2}{C_0 + 4t} \geq -\frac{2}{4t} = -\frac{1}{2t} > -\frac{3}{2t}.

Exercise 8.4 The fibre of SL~(2,R)\widetilde{SL}(2, \mathbb{R})

With c=(−2,2,2)c = (-2, 2, 2) and B=CB = C, show B′≥8B' \geq 8, so B≥B0+8tB \geq B_0 + 8t, and A′≥−4A2(B0+8t)2A' \geq -\frac{4A^2}{(B_0 + 8t)^2}. Deduce that A(t)A(t) stays bounded below by a positive constant. What does g(t)t\frac{g(t)}{t} converge to on the base directions?

Solution

B′=8+4AB≥8B' = 8 + \frac{4A}{B} \geq 8. Then ddt(1A)=−A′A2=4B2≤4(B0+8t)2\frac{d}{dt}\big(\frac1A\big) = -\frac{A'}{A^2} = \frac{4}{B^2} \leq \frac{4}{(B_0 + 8t)^2}, which is integrable, so 1A\frac1A stays bounded and AA is bounded below. The base metric Bt(ω22+ω32)→8(ω22+ω32)\frac{B}{t}(\omega_2^2 + \omega_3^2) \to 8(\omega_2^2 + \omega_3^2), a hyperbolic metric on the base, while At→0\frac At \to 0: the circle fibres collapse.

Exercise 8.5 Rehearsal: decay rates

For each of Nil, Sol and SL~(2,R)\widetilde{SL}(2, \mathbb{R}), find how ∣R∣|R| decays as t→∞t \to \infty, and check in each case that it is consistent with the universal bound R≥−32tR \geq -\frac{3}{2t} of 11A.4 Maximum Principles under Ricci Flow. Flows that exist for all time with ∣Rm⁡∣≤Ct|\operatorname{Rm}| \leq \frac Ct are called Type III; these are examples, and their rescalings g(t)t\frac{g(t)}{t} are the collapsing limits of the table above.

Solution

Nil: R=−A0B02B3∼−16tR = -\frac{A_0B_0}{2B^3} \sim -\frac{1}{6t}. Sol: R=−2C0+4t∼−12tR = -\frac{2}{C_0 + 4t} \sim -\frac{1}{2t}. SL~(2,R)\widetilde{SL}(2, \mathbb{R}): with B=CB = C, Milnor's formulas give R=−8B−2AB2R = -\frac{8}{B} - \frac{2A}{B^2}, which is ∼−1t\sim -\frac{1}{t} since B∼8tB \sim 8t and AA is bounded. Each is ≥−32t\geq -\frac{3}{2t} for large tt, as the bound requires; the full curvature tensor is of the same order, so the flows are Type III.

© 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.