Positive curvature and rational ellipticity in cohomogeneity three

Elahe Khalili Samani; Marco Radeschi

REVIEW 1 major objections 5 minor 23 references

Positive curvature and rational ellipticity in cohomogeneity three

T0 review · 1 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Positive curvature and a three-dimensional symmetry group force rational ellipticity.

desk verdict New case of Bott-Grove-Halperin with a fixable gap in the gluing lemma; worth refereeing. read the letter →

arxiv 2505.22577 v1 pith:2UI6XG3N submitted 2025-05-28 math.DG

Elahe Khalili Samani , Marco Radeschi This is my paper

classification math.DG MSC 53C2057S1555P62

keywords rationalellipticitypositivecurvaturecohomogeneitythreeAlexandrovspacesgroupactionshomotopytheorysingularstratabranchedcovers

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proves that any closed, simply connected, positively curved Riemannian manifold admitting a cohomogeneity-three action by a closed Lie group with boundary-free quotient is rationally elliptic, meaning its total rational homotopy is finite-dimensional. The result extends the known rational ellipticity of homogeneous spaces, cohomogeneity-one actions, and almost non-negatively curved cohomogeneity-two actions to cohomogeneity three under positive curvature. The proof analyzes the singular stratification of the quotient space, classifies all possible singular graphs, and shows each case is covered by either a rational cohomology computation or a double disk bundle decomposition. A reader should care because this is a general cohomogeneity-three statement for positive curvature that does not assume extra properties such as infinitesimal polarity.

What carries the argument

The labeled singular graph of the quotient $X=M/G$: vertices are zero-dimensional strata, marked hollow for orbifold points and full for non-orbifold points, and edges are one-dimensional strata weighted by integers $p$ for which the space of directions is $S^2_{pp}$. The classification of all such graphs that can occur, giving the eighteen configurations of Proposition 5.1, is the backbone of the proof. It is obtained from the at-most-three-small-points bound, from branched covers along singular cycles (which preserve positive curvature and produce new small points), and from Lemma 2.5's classification of the linear slice representations of cohomogeneity three with boundary-free quotient. Once the graph is known, each case falls into one of two mechanisms: a Mayer-Vietoris computation of rational cohomology after restricting the acting group, or a double disk bundle decomposition over rationally $\Omega$-elliptic orbits.

What would settle it

Exhibit a positively curved 3-dimensional Alexandrov space, ideally the quotient of a closed positively curved cohomogeneity-three $G$-manifold, containing four distinct small points; alternatively, construct a closed, simply connected, positively curved cohomogeneity-three manifold with boundary-free quotient whose total rational homotopy $\bigoplus_i\pi_i(M)\otimes\mathbb{Q}$ is infinite-dimensional.

Watch

Extended reading notes

Core claim

The paper's central claim is Theorem A: a closed, simply connected, positively curved Riemannian manifold with a cohomogeneity-three action by a closed Lie group, and whose quotient has no boundary, is rationally elliptic. The proof rests on the fact that the quotient is a positively curved Alexandrov space homeomorphic to $S^3$ whose singular set is a graph with at most three small points. Proposition 5.1 classifies that graph into eighteen labeled configurations, indexed by edge weights and hollow/full vertex markings. For configurations with a weight-two edge and one or two non-orbifold points, the paper first restricts the acting group to products of rank-one groups and then computes the rational cohomology ring by a Mayer-Vietoris argument, concluding rational ellipticity. For the remaining configurations, it constructs a double disk bundle decomposition of $M$ over rationally $\Omega$-elliptic orbits, which also forces rational ellipticity.

Load-bearing premise

The proof depends on the external bound that a positively curved 3-dimensional Alexandrov space has at most three small points; if a fourth small point can exist, the classification of singular graphs and the whole case analysis would collapse.

Editorial extensions

If this is right

For every manifold satisfying the hypotheses, only finitely many rational homotopy groups are nonzero; the total rational homotopy $\bigoplus_{i\ge 2}\pi_i(M)\otimes\mathbb{Q}$ is finite-dimensional.
The quotient's singular stratum must be one of the eighteen labeled graphs, so any future example can be checked against this list.
The theorem covers quotients that are orbifolds as well, since that case follows from the authors' earlier criterion, and adds the non-orbifold cases.
In the knotted singular-cycle case, the cycle must be a trefoil knot with weight exactly two, and that weight-two edge alone suffices to put $M$ in the rationally elliptic case.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The at-most-three-small-points bound is the load-bearing external input; a positively curved Alexandrov 3-space with four small points would invalidate the graph classification and therefore the proof as written.
The double-disk-bundle mechanism may extend to non-negative curvature cohomogeneity-three actions if a similar graph classification can be established; the paper does not make this claim.
The appendix realizes most, but not all, of the eighteen graphs by explicit actions on spheres and complex projective spaces; realizing the missing configurations would show whether the classification is sharp.
The exceptional trefoil-knot case suggests that knot types of singular cycles interact with curvature through branched covers, a phenomenon that may have analogues in other positive-curvature symmetry problems.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

First, the good news: Theorem A is a real step forward. It removes the infinitesimally polar assumption for cohomogeneity-three actions under positive curvature, and the quotient graph classification in Proposition 5.1 is genuinely new. The two-technique structure—rational homotopy/Mayer-Vietoris for the weight-two cases, double disk bundle for the rest—is coherent, and the case analysis is impressive. If the proof goes through, this is the kind of result people in the area will want to quote.

The main body is the non-orbifold case, and it is mostly self-contained. The orbifold case is delegated to the authors' earlier Theorem C; that is a legitimate division of labor, not circular.

Now the soft spot, and it is real. Lemma 6.2 is load-bearing for all the double-disk-bundle cases (graphs 1–6, 8, 9, 17, and the unknot case of graph 7). Its proof asserts that π0(Diff_D^+(S^2,S^2)) ≅ π0(Diff^+(S^2,S^2)) ≅ {0}, because any two points can be moved to any other two points by an ambient isotopy. That is the wrong comparison: diffeomorphisms that fix two disjoint disks form a group whose π0 is not trivial. The configuration space of two disks in S^2 has nontrivial π1, and there is a Dehn twist of the complementary annulus that is not isotopic to the identity through disk-fixing diffeomorphisms. So the surjectivity argument in the proof does not establish the needed isotopy.

That said, the lemma looks salvageable. The situation is an annulus, and orientation-preserving diffeomorphisms of an annulus that preserve boundary components are isotopic if you allow boundary points to move; the boundary-fixing mapping class group is Z, but the proof only needs a weaker statement. A direct annulus argument should close the gap. The stress-test note is correct that the written proof, as it stands, is incomplete at exactly this junction.

Also minor: the paper states Theorem A for a closed Lie group G, but the proof in Sections 5–6 assumes G is compact and connected; the reduction is not spelled out. I suspect it is standard, but a reader has to fill it in.

Bottom line: the main theorem is probably true, the proof is mostly in order, and the one genuine gap is localized and likely repairable. This deserves a serious referee, not a desk reject. I would send it out with a request to fix Lemma 6.2 and clarify the connected-group reduction.

Referee Report

1 major / 5 minor

Summary. The paper proves Theorem A: a closed, simply connected, positively curved Riemannian manifold admitting an effective cohomogeneity-three action by a closed Lie group, with empty quotient boundary, is rationally elliptic. The proof reduces the orbifold-quotient case to the authors' previous Theorem C (Remark 1.1), then classifies the possible singular graphs of a non-orbifold positively curved quotient 3-space (Proposition 5.1). Each resulting configuration is handled either by a rational cohomology/Mayer-Vietoris argument when the quotient has one or two non-orbifold points and a weight-two singular edge (Propositions 3.7 and 3.10), or by a double disk-bundle decomposition using a product recognition lemma (Lemma 6.2) and rational Omega-ellipticity (Proposition 4.1). An appendix supplies explicit actions realizing most of the configurations.

Significance. If correct, Theorem A represents a substantial advance, extending rational ellipticity results from homogeneous, cohomogeneity-one, and almost non-negatively curved cohomogeneity-two settings to positively curved cohomogeneity-three manifolds with no quotient boundary. The classification of quotient singular graphs in Proposition 5.1 is a valuable structural contribution, and the non-orbifold part of the proof is largely self-contained, relying only on standard external facts such as the Hsiang-Kleiner and Grove-Markvorsen bound on small points, Perelman's theorem for the homeomorphism type of the quotient, and Lytchak's results on cohomogeneity-one actions. The paper is honest about its dependence on the authors' previous Theorem C for the orbifold case. The main weakness is a genuine gap in the proof of Lemma 6.2, which is load-bearing for all the double disk-bundle cases; in my view the gap is repairable, but the manuscript as written does not fully justify the main theorem.

major comments (1)

[Lemma 6.2 (Section 6)] The final step of the proof of Lemma 6.2 is not justified. The paper asserts that pi_0(Diff_D^+(S^2,S^2)) is isomorphic to pi_0(Diff^+(S^2,S^2)) = 0 because "any two points in S^2 can be moved to any other two points via a diffeomorphism isotopic to the identity." That sentence concerns transitivity on point pairs, not the component structure of the stabilizer of the disk pair D_±, so the claimed isomorphism between these two pi_0-groups requires a real argument. I do not think the paper's conclusion is contradicted by the ordinary mapping class group of the annulus, which is Z only for isotopies fixing the boundary pointwise and is not the group used here; nevertheless, the proof as written does not establish the required vanishing. A direct argument in Diff^+(S^1 x [0,1]) without pointwise boundary fixing, or a correct component computation for the stabilizer of two disks, is needed. Because Lemma 6.2 is the mechanism by which X \ B_epsilon(A* union B*) is recognized as an orbifold product in Propositions 6.3 and 6.5, this gap is load-bearing for the proof of Theorem A.

minor comments (5)

[Lemma 6.4] "colosd" should be "closed" in the statement.
[Proposition 3.10] In Case 1, the reference "by Corollary 3.5" should be "by Lemma 3.5."
[Proposition 3.10] "spactral" should be "spectral".
[Proposition 5.1] "unlabaled" and "Furthemore" should be corrected to "unlabeled" and "Furthermore".
[Lemma 6.2] The notation Diff_D^+(S^2,S^2) and the phrase "sends D_± to itself" are ambiguous: clarify whether the diffeomorphisms preserve each disk individually or only the unordered pair.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the non-orbifold proof is self-contained, and the one self-citation delegates an orbifold subcase to an independent prior theorem.

full rationale

The derivation of Theorem A does not reduce to its own inputs by construction. The bulk of the paper, Sections 3 through 6, treats quotients with at least one non-orbifold point and relies on external results rather than on the theorem being proved: the Hsiang-Kleiner and Grove-Markvorsen small-point bounds enter Lemma 2.8(3) and drive the Proposition 5.1 classification; the structure results of Lytchak and of Grove-Wilking-Ziller are used in Lemma 2.5 and Proposition 3.2; and the double mapping cylinder criterion of Grove-Halperin is invoked in Proposition 4.1. The rational ellipticity conclusions in Propositions 3.7, 3.10, 6.1, 6.3, and 6.5 are obtained by computing rational cohomology or exhibiting double disk bundles, not by fitting parameters or renaming a known result. The only self-citation is Remark 1.1, where the orbifold-quotient case of Theorem A is referred to [13, Theorem C]; that is a prior theorem with stated assumptions narrower than Theorem A and is used as an external mathematical input, not as a restatement of the present conclusion, so it does not create a circular loop. A possible correctness gap in Lemma 6.2, concerning mapping class groups of annuli, would be a mathematical error rather than a circular reduction. No fitted-input-called-prediction or definitional equivalence appears in the paper.

Assumptions & free parameters 0 free parameters · 7 assumptions · 0 invented entities

The theorem rests on several deep external results in Alexandrov geometry, transformation groups, rational homotopy, and knot theory. The only self-authored input is the prior Theorem C for the orbifold case. No free parameters or invented entities appear.

assumptions (7)

standard math M/G is an Alexandrov space homeomorphic to S^3 (Perelman's theorem, quoted as Corollary IV.4.7 of [1]).
Invoked in Section 2.3(2); the entire quotient-space analysis and graph classification require that M/G is S^3.
standard math A positively curved Alexandrov space contains at most three small points (Hsiang-Kleiner and Grove-Markvorsen).
Used in Lemma 2.8(3) and throughout Section 5; this bound drives the combinatorial classification of singular graphs.
standard math Lytchak's theorem [14] on the existence of singular orbits for cohomogeneity-one actions on compact simply connected manifolds.
Used in Lemma 2.10 and Proposition 3.2 to restrict the group and the singular strata.
standard math Grove-Wilking-Ziller Lemma 3.5 classifying positively curved cohomogeneity-one actions with K_+ = K_- = S^1 as having group S^1 or T^2.
Used in Proposition 3.2 to conclude that the lifted group G'' is abelian.
standard math Straume's classification of compact linear groups of cohomogeneity at most three.
Basis for the slice classification in Lemma 2.5.
domain assumption Theorem C from the authors' prior paper [13], asserting rational ellipticity for cohomogeneity-three manifolds with orbifold quotient.
Remark 1.1 delegates the orbifold case to this prior theorem; it is not reproven here.
standard math Standard knot theory facts: the double branched cover of the unknot is S^3, the double cover of the trefoil is L(3,1), and the triple cover of the trefoil has fundamental group Q_8.
Used in Lemma 6.4 to rule out knots other than the unknot or the trefoil with edge weight 2.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Positive curvature and rational ellipticity in cohomogeneity three." pith.science (2026). https://pith.science/paper/2UI6XG3N

@misc{pith2026250522577,
  author       = {Pith},
  title        = {Pith review of: Positive curvature and rational ellipticity in cohomogeneity three},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2UI6XG3N}},
  note         = {Machine review of arXiv:2505.22577}
}

read the original abstract

We prove that a closed, simply connected, positively curved, cohomogeneity-three manifold whose quotient space has no boundary is rationally elliptic, thus providing a generalization of similar results regarding rational ellipticity of homogeneous, cohomogeneity-one, and almost non-negatively curved cohomogeneity-two manifolds.

Figures

Figures reproduced from arXiv: 2505.22577 by the authors.

**Figure 1.** Double disk bundle decompositions. Dashed lines (resp. the hollow point in the first picture) represent curves (resp. point) chosen in the regular part. small points, a contradiction. Therefore, c is the unknot. Then by letting A∗, B∗ denote the two vertices, we can find a homeomorphism X → S 3 sending Bǫ(A∗) to a neighborhood of the north pole, Bǫ(B∗) to a neighborhood of the south pole, and the two edges to meridi… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

23 extracted references · 23 canonical work pages

[1]

G. E. Bredon, Introduction to Compact Transformation Groups , Academic Press, New York, 1972

work page 1972
[2]

Burago, Y

D. Burago, Y. Burago, S. Ivanov, A Course in Metric Geometry , American Mathematical Society, Prov- idence, 2001

work page 2001
[3]

Burde, H

G. Burde, H. Zieschang, Knots, De Gruyter Stud. Math., vol 5, New York, 2003

work page 2003
[4]

Galaz-Garcia, C

F. Galaz-Garcia, C. Searle, Nonnegatively curved 5-manifolds with almost maximal symm etry rank, Geom. Topol. 18 (2014), 1397–1435

work page 2014
[5]

Grove, Geometry of, and via, symmetries , Conformal, Papers from the 30th John H

K. Grove, Geometry of, and via, symmetries , Conformal, Papers from the 30th John H. Barrett Memo- rial Lectures delivered at the University of Tennessee, Knoxville, T N. Univ. Lecture Ser. 27, American Mathematical Society, Providence, RI (2002)

work page 2002
[6]

K. Grove. S. Halperin, Dupin hypersurfaces, group actions and the double mapping c ylinder, J. Diﬀerential Geom. 26 (1987), 429–459

work page 1987
[7]

Grove, S

K. Grove, S. Markvorsen, New extremal problems for the Riemannian recognition progr am via Alexandrov geometry, J. Amer. Math. Soc. 8 (1995), 1–28

work page 1995
[8]

Grove, B

K. Grove, B. Wilking, A knot characterization and 1-connected non-negatively curved 4-manifolds with circle symmetry , Geometry and Topology 18 (2014), no. 5, 3091–3110

work page 2014

Show all 23 references

[9]

Grove, B

K. Grove, B. Wilking, J. Yeager, Almost non-negative curvature and rational ellipticity in cohomogeneity two, Ann. Inst. Fourier 69 (2019), no. 7, 2921–2939

2019
[10]

Grove, B

K. Grove, B. Wilking, W. Ziller, Positively curved cohomogeneity one manifolds and 3-Sasakian geom- etry, J. Diﬀer. Geom. 78 (2008), no. 1, 33–111

2008
[11]

Grove, W

K. Grove, W. Ziller, Polar manifolds and actions , J. Fixed Point Theory Appl. 11 (2012), 279–313

2012
[12]

W. Y. Hsiang, B. Kleiner, On the topology of positively curved 4-manifolds with symmetry , J. Diﬀer. Geom. 29 (1989), no. 3, 615–621

1989
[13]

Khalili Samani, M

E. Khalili Samani, M. Radeschi, Rational ellipticity of G-manifolds from their quotients , Compositio Mathematica (2024), to appear

2024
[14]

Lytchak, Geometric resolution of singular Riemannian foliations , Geom

A. Lytchak, Geometric resolution of singular Riemannian foliations , Geom. Dedicata 149 (2010), 379– 395

2010
[15]

Lytchak, G

A. Lytchak, G. Thorbergsson, Curvature explosion in quotients and applications , J. Diﬀer. Geom. 85 (2010), no. 1, 117–140

2010
[16]

McGowan, C

J. McGowan, C. Searle, How Tightly Can You Fold a Sphere? , Diﬀerential Geometry and its Applications 22 (2005), 81–104

2005
[17]

Mendes, Lifting isometries of orbit spaces , Bull

R.A.E. Mendes, Lifting isometries of orbit spaces , Bull. Lond. Math. Soc. 53 (2021), no. 6, 1621–1626. 28 ELAHE KHALILI SAMANI AND MARCO RADESCHI

2021
[18]

Mendes, M

R. Mendes, M. Radeschi, Laplacian algebras, manifolds submetries and their invers e invariant theory problem, Geom. Func. Anal. 30 (2020), 536–573

2020
[19]

Petersen, Riemannian geometry, Third Edition , Graduate Texts in Mathematics, Springer, 2016

P. Petersen, Riemannian geometry, Third Edition , Graduate Texts in Mathematics, Springer, 2016

2016
[20]

Rolfsen, Knots and Links , United Kingdom: AMS Chelsea Pub., 2003

D. Rolfsen, Knots and Links , United Kingdom: AMS Chelsea Pub., 2003

2003
[21]

J. P. Serre, Groupes d’homotopie et classes de grupes ab´ eliens, Ann. of Math. 58 (1953), 258–294

1953
[22]

Straume, On the invariant theory and geometry of compact linear group s of cohomogeneity ≤ 3, Diﬀerential Geometry and its Applications 4 (1994) 1–23

E. Straume, On the invariant theory and geometry of compact linear group s of cohomogeneity ≤ 3, Diﬀerential Geometry and its Applications 4 (1994) 1–23

1994
[23]

Wilking, Positively curved manifolds with symmetry , Ann

B. Wilking, Positively curved manifolds with symmetry , Ann. of Math. 163 (2006), no. 2, 607–668. (E. Khalili Samani) Department of Mathematics, Clark University, Worcester, MA 01610, USA Email address : ekhalilisamani@clarku.edu (M. Radeschi) Universit´a Degli Studi Di Torino...

2006

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

G. E. Bredon, Introduction to Compact Transformation Groups , Academic Press, New York, 1972

work page 1972

[2] [2]

Burago, Y

D. Burago, Y. Burago, S. Ivanov, A Course in Metric Geometry , American Mathematical Society, Prov- idence, 2001

work page 2001

[3] [3]

Burde, H

G. Burde, H. Zieschang, Knots, De Gruyter Stud. Math., vol 5, New York, 2003

work page 2003

[4] [4]

Galaz-Garcia, C

F. Galaz-Garcia, C. Searle, Nonnegatively curved 5-manifolds with almost maximal symm etry rank, Geom. Topol. 18 (2014), 1397–1435

work page 2014

[5] [5]

Grove, Geometry of, and via, symmetries , Conformal, Papers from the 30th John H

K. Grove, Geometry of, and via, symmetries , Conformal, Papers from the 30th John H. Barrett Memo- rial Lectures delivered at the University of Tennessee, Knoxville, T N. Univ. Lecture Ser. 27, American Mathematical Society, Providence, RI (2002)

work page 2002

[6] [6]

K. Grove. S. Halperin, Dupin hypersurfaces, group actions and the double mapping c ylinder, J. Diﬀerential Geom. 26 (1987), 429–459

work page 1987

[7] [7]

Grove, S

K. Grove, S. Markvorsen, New extremal problems for the Riemannian recognition progr am via Alexandrov geometry, J. Amer. Math. Soc. 8 (1995), 1–28

work page 1995

[8] [8]

Grove, B

K. Grove, B. Wilking, A knot characterization and 1-connected non-negatively curved 4-manifolds with circle symmetry , Geometry and Topology 18 (2014), no. 5, 3091–3110

work page 2014