Quantitative uniqueness for mean curvature flow

Tobias Holck Colding; William P. Minicozzi II

REVIEW 4 major objections 4 minor 1 cited by

Quantitative uniqueness for mean curvature flow

T0 review · 4 major / 4 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read If a rescaled mean curvature flow starts close to a cylinder and its Gaussian area barely changes, the flow cannot drift far; the bound is independent of elapsed time.

desk verdict A short, honest extension of the authors' earlier uniqueness-of-blowups theorem to an effective statement with drifting centers; the new result is real, but the proof of a key regularity lemma is only a citation. read the letter →

arxiv 2502.03634 v1 pith:7HLDWC4P submitted 2025-02-05 math.DG math.AP

Tobias Holck Colding , William P. Minicozzi II This is my paper · ORCID

classification math.DGmath.AP MSC 53E1053C44

keywords meancurvatureflowuniquenessofblowupsLojasiewiczinequalityrescaledGaussianareacylindricalsingularitiesancientflowsentropy

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

Mean curvature flow can form singularities that, after rescaling, look like cylinders. This paper proves a quantitative version of the earlier uniqueness theorem for such cylindrical blowups: if a rescaled flow is close to a cylinder for an initial stretch and its Gaussian area $F$ changes by only a small amount between the start and an arbitrary later time, then the flow stays close to its own initial slice, with the distance controlled by a power of the change in $F$ and constants that do not depend on how long the flow runs. Using that bound, the paper shows that two tangent cylinders of the same singularity must actually be the same cylinder, and that cylindrical blow-down limits of ancient flows are unique. The point is that a qualitative uniqueness statement becomes a quantitative estimate that can be reapplied over arbitrarily long time intervals.

What carries the argument

The engine is a discrete Lojasiewicz-type inequality for rescaled MCF imported from the authors' earlier work (Theorem 3.1): for every cylinder $C=S^k\sqrt{2k}\times\mathbb{R}^{n-k}$, there are constants $C,\bar R,\epsilon,\tau\in(1/3,1)$ so that whenever a rescaled flow $\Sigma_s$ stays $\epsilon$-close to $C$ in $B_{\bar R}$ for $s\in[t-1,t+1]$, one has $|F(\Sigma_t)-F(C)|^{1+\tau}\le C(F(\Sigma_{t-1})-F(\Sigma_{t+1}))$. This inequality converts a small drop in the Gaussian area $F$ into control of the motion. A new discrete proposition (Proposition 2.1) shows that if a non-increasing sequence $x_j>0$ satisfies $x_{j+1}^{1+\tau}\le C(x_j-x_{j+1})$, then $\sum_j |x_j-x_{j+1}|^{1/2}\le c x_1^\alpha$; summing square roots of successive drops is exactly the length estimate one needs for a gradient flow. Lemma 3.3 upgrades the resulting $L^1$ bound into $C^{2,\alpha}$ graphical closeness in a fixed ball, using the entropy bound, the Brakke estimate, and parabolic regularity. The proof of the main theorem applies these three ingredients in three cases depending on whether $F$ stays above $F(C)$, below $F(C)$, or crosses it.

What would settle it

A concrete way to test the theorem is to run a numerical or analytic rescaled MCF that stays within $\epsilon_1$ of a cylinder $C$ on $[t_1,t_1+2]$ with $|F(\Sigma_{t_i})-F(C)|<\epsilon_2$ at its endpoints, and look for a later time $t\in[t_1+1,t_2]$ where $\mathrm{dist}_{R_2}(\Sigma_t,\Sigma_{t_1+1})$ exceeds $c(|F(\Sigma_{t_1})-F(C)|^\alpha+|F(\Sigma_{t_2})-F(C)|^\alpha)$; finding such an example would refute Theorem 0.5. Since the constants in Theorem 3.1 are not proved here, the same test could check the discrete inequality $|F(\Sigma_t)-F(C)|^{1+\tau}\le C(F(\Sigma_{t-1})-F(\Sigma_{t+1}))$ directly on flows that stay close to a cylinder over three consecutive unit time steps.

Watch

Extended reading notes

Core claim

The central claim is Theorem 0.5: for an $n$-dimensional rescaled mean curvature flow $\Sigma_t$ with bounded entropy, there are constants $c,\alpha,\epsilon_1,\epsilon_2,R_1,R_2$ such that if $\Sigma_t$ stays $\epsilon_1$-close to a cylinder $C=S^k\sqrt{2k}\times\mathbb{R}^{n-k}$ in a large ball for $t\in[t_1,t_1+2]$ and $|F(\Sigma_{t_i})-F(C)|<\epsilon_2$ for $i=1,2$, then $\mathrm{dist}_{R_2}(\Sigma_t,\Sigma_{t_1+1}) < c|F(\Sigma_{t_1})-F(C)|^\alpha + c|F(\Sigma_{t_2})-F(C)|^\alpha$ for every $t\in[t_1+1,t_2]$. The key feature is that the right-hand side does not grow with $t_2-t_1$. The paper derives Theorem 0.1 as a consequence: under the cylindrical singularity assumptions (A) and (B), the two rotations coincide, so $O(C)=C$, meaning the cylinder is unique. The same quantitative estimate yields uniqueness of cylindrical blow-down limits for ancient mean curvature flows. The proof is modeled on the finite-dimensional Lojasiewicz argument but split into three cases according to whether $F$ stays above or below $F(C)$; time reversal is not available for MCF, so the below case is handled by running the discrete argument backwards.

Load-bearing premise

All of the quantitative control rests on the Lojasiewicz-type inequality for rescaled flows near cylinders (Theorem 3.1), imported from the authors' earlier paper and not reproved here; if that inequality fails, or only holds with constants that degenerate with the time interval, the effective bounds do not follow.

Editorial extensions

If this is right

A rescaled mean curvature flow that is initially close to a cylinder and whose Gaussian area changes by only $\epsilon$ at the endpoints stays within $O(\epsilon^\alpha)$ of its time-$(t_1+1)$ slice for the entire interval, no matter how long the interval is.
If a cylindrical singularity satisfies assumptions (A) and (B), the rotation coming from the sequence of rescalings must fix the cylinder, so $O(C)=C$; this is Theorem 0.1.
If an ancient mean curvature flow with bounded entropy has one cylindrical blow-down, then every blow-down is the same cylinder.
The same effective bound controls the total weighted motion of the flow near the cylinder, since Lemma 3.3 yields $C^{2,\alpha}$ closeness from the summed square roots of successive drops of $F$.

Reading between the lines

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

A natural next step is to run the same effective scheme with a local entropy bound in place of the global bound $\lambda(\Sigma_t)\le\lambda_0$, which would make the estimate applicable to flows that are only locally controlled.
The independence of the constants from the time interval suggests the estimate could be used to prove stability of cylindrical singularities under perturbations of the initial data, replacing limit arguments by explicit bounds.
One could attempt to relax the Lojasiewicz-type input, for instance to a power that degenerates slowly, and see whether the summability proposition still yields finite length; this would indicate how far the method reaches beyond cylinders.

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

Short version: this paper does what its abstract says—takes the arguments from [CM2] and extracts an effective uniqueness statement for rescaled MCF, Theorem 0.5, which then gives uniqueness of cylindrical blowups with drifting centers (Theorem 0.1) and cylindrical blowdowns. That is a genuinely new result, not just repackaging. The discrete effective lemma (Prop. 2.1) is proved in detail and the model case in Section 1 is a nice warm-up. Credit where due: the paper is transparent about importing the Lojasiewicz inequality from [CM2], and the application in Theorem 0.1 is straightforward once Theorem 0.5 is in hand.

The soft spot is exactly where the stress-test note lands. Lemma 3.3 turns an L1 bound on Gaussian area differences into C^{2,α} closeness at a fixed radius with constants independent of interval length. Its proof here is a citation to page 268 of [CM2], plus the sketch in (3.5). That may be fine for an expert who knows [CM2] cold, but it is load-bearing for Theorem 0.5. Two details need reconciliation: the lemma is stated for unit-step sums (3.4), while the proof of Theorem 0.5 uses coarser two-step sums (3.7) and (3.11) to define N; and the footnote about the absolute-value form of the Lojasiewicz inequality is reassuring but not a proof. Neither issue looks fatal—they look like missing write-ups—but a referee cannot certify Theorem 0.5 from the present text alone. Since the abstract promises a 'stronger effective version,' the proof of Lemma 3.3 should be supplied, even in an appendix.

The citation pattern is appropriate; this is a direct continuation of [CM2], and it cites a reasonable range of related work. No code or data, obviously; this is pure geometry. The math that is actually written is correct as far as I can tell—Proposition 2.1's proof is solid.

Verdict: send to a serious referee, with instructions to focus on Lemma 3.3 and the summation-index reconciliation. If those get fixed, this is a solid contribution to the MCF regularity program. I would bring it to reading group once the arXiv version has been updated or the referee report is out.

Referee Report

4 major / 4 minor

Summary. The paper proves a quantitative (effective) uniqueness theorem for rescaled mean curvature flow near a cylindrical shrinker. Theorem 0.5 states that if a rescaled MCF is C^{2,alpha}-close to a cylinder C on a fixed ball for two units of time and the Gaussian area F at the endpoints is close to F(C), then the flow remains C^{2,alpha}-close to its time-t1+1 slice, with the distance bounded by powers of the endpoint F-deviations and with constants independent of the length of the time interval. The proof adapts the finite-dimensional Lojasiewicz gradient-flow argument of Section 1 to the discrete Lojasiewicz inequality imported as Theorem 3.1 from [CM2], and uses Lemma 3.3 to convert an L1 decrease of Gaussian area into C^{2,alpha} closeness. The paper then derives Theorem 0.1: if a cylindrical singularity admits a sequence of rescalings, centered at points approaching the origin, that converges smoothly to a rotated cylinder O(C), then O(C)=C.

Significance. If the imported estimates are valid in the stated form, the result is a genuine strengthening of [CM2]: it gives an explicit, interval-independent rate of closeness under only endpoint control of F, and it yields the clean consequence Theorem 0.1. The paper is transparent about the source of its two principal inputs, and the finite-dimensional part (Sections 1 and 2) is self-contained, with Proposition 2.1 proved in detail and without fitted parameters. There is no circular use of the main theorem. However, the two geometric inputs from [CM2] are not reproved here, and the version of Lemma 3.3 needed for Theorem 0.5 is not exactly the version stated in [CM2]. The stress-test concern about Lemma 3.3 is therefore accurate: this is where the proof needs the most work.

major comments (4)

[§3.2, Lemma 3.3] Lemma 3.3 is load-bearing and its proof is not contained in the paper: the proof is the sentence 'This follows as in (1) on page 268 of [CM2]' together with the L1 estimate (3.5). The present text does not verify that the parabolic-regularity argument from [CM2] yields the linear bound dist_{Rbar}(Sigma_t,C) <= Ctilde mu, nor that the constants are independent of N and of the interval length. Since Theorem 0.5 applies this lemma repeatedly, including after the reversal in Case 2, the exact form of the lemma with the fixed radius and the unit-step sum (3.4) must be proved, or the corresponding statement in [CM2] must be quoted with all hypotheses checked.
[§3.3, Cases 1 and 2] The sequences x_j = F(Sigma_{t1+2j-1}) - F(C) sample F at every other integer time, while Lemma 3.3's hypothesis (3.4) is a sum over unit steps. The proof defines N using the two-step sums (3.7) and (3.11) and then invokes Lemma 3.3 without explaining why (3.4) holds. At best the unit-step sum is bounded by twice the two-step sum, so the argument needs either a factor 2 absorbed into mu or a redefinition of N; as written, the invocation of Lemma 3.3 is not justified.
[Footnote 3 / Theorem 3.1] Theorem 3.1 is quoted in the absolute-value form |F(Sigma_t)-F(C)|^{1+tau} <= C(F(Sigma_{t-1})-F(Sigma_{t+1})), but the footnote concedes that [CM2] was stated only when F is above F(C) and asserts without demonstration that the absolute-value form 'is not used' there. Cases 2 and 3 of the proof of Theorem 0.5 depend on the inequality when F is below F(C). The paper needs to supply the short argument, or a precise reference to the line in [CM2], showing that the proof of Theorem 6.1 in [CM2] covers this case, rather than leaving the extension to a footnote assertion.
[§3.3, Case 2] The reversal construction x_j = -y_{N-j} is only sketched in one sentence. To apply Proposition 2.1 one must check that the reindexed sequence satisfies x_j > 0, is non-increasing, and obeys x_{j+1}^{1+tau} <= C(x_j - x_{j+1}) on the full range with the same constant C. This is likely true, but the index bookkeeping should be written out, since Proposition 2.1's conclusion is used to conclude that N reaches t2.

minor comments (4)

[§1, Eq. (1.6)] The displayed inequality has a missing closing parenthesis: 'F(gamma(i + 1)' should be 'F(gamma(i + 1))'.
[§3.4, Eq. (3.18)] The displayed estimate is garbled in the text; it should read |F(Sigma^i_t) - F(C)| < delta_1.
[Definition 0.4] The definition of dist_R(Sigma, Gamma) should specify the domain of the graph and the convention for the C^{2,alpha} norm, so that 'norm less than epsilon' is unambiguous.
[Lemma 3.3] The integer N appears in the hypothesis before being quantified; the lemma should state that N is a nonnegative integer and that the flow is defined on the corresponding time interval.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the new effective uniqueness proof builds on the authors' prior [CM2] estimates, but the target theorem is not an input of those estimates and no parameter is fitted.

full rationale

The paper's central claim, Theorem 0.5, is derived from two imported ingredients: the Lojasiewicz-type inequality of Theorem 3.1 (quoted from [CM2]) and the parabolic-regularity Lemma 3.3, whose proof is sketched and referenced to page 268 of [CM2]. These are self-citations and they are load-bearing, but they are not circular reductions. Theorem 3.1 is a published, parameter-free theorem whose assumptions (entropy bound and closeness to a cylinder) do not include Theorem 0.5 or Theorem 0.1, and Lemma 3.3 is a regularity estimate obtained from Brakke's estimate and parabolic estimates, not a restatement of the desired distance bound. The discrete Proposition 2.1 and the model case are proved in the paper itself. In the proof of Theorem 0.5, the L1 sum of square-root drops is bounded by endpoint F-differences via Proposition 2.1, and Lemma 3.3 converts that L1 bound into C^{2,alpha} closeness; no constants are fitted and no quantity is defined in terms of the conclusion. The footnote extending Theorem 3.1 to the case F below F(C) is an assertion about [CM2]'s proof, but it is a possible gap or correctness concern, not circularity. The score of 2 reflects the heavy reliance on the authors' own prior work, not a finding that the derivation is circular.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The theorem rests on a published Lojasiewicz inequality and standard parabolic regularity theory; no free parameters are fitted and no new entities are introduced.

assumptions (4)

domain assumption Lojasiewicz-type inequality for rescaled MCF near a cylinder (Theorem 3.1 / Theorem 6.1 of [CM2]): |F(Sigma_t) - F(C)|^{1+tau} <= C (F(Sigma_{t-1}) - F(Sigma_{t+1})).
Imported from the authors' prior paper [CM2]; the main quantitative engine of Theorem 0.5. If the constants or exponent tau fail near every cylindrical shrinker, the effective uniqueness theorem collapses.
domain assumption Brakke estimate / local regularity theorem of White [W1] gives local curvature bounds for the MCF from an entropy bound and closeness to a cylinder.
Used in Lemma 3.3 to convert L1 gaussian-area decay into C^{2,alpha} graph closeness; stated but not proved.
standard math Rescaled MCF is the gradient flow of F, so F is non-increasing in time.
Facts used throughout (Sections 3.3 and model Section 1).
standard math Parabolic Schauder estimates upgrade the L1 bound to higher regularity.
Invoked in the proof sketch of Lemma 3.3.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Quantitative uniqueness for mean curvature flow." pith.science (2026). https://pith.science/paper/7HLDWC4P

@misc{pith2026250203634,
  author       = {Pith},
  title        = {Pith review of: Quantitative uniqueness for mean curvature flow},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7HLDWC4P}},
  note         = {Machine review of arXiv:2502.03634}
}

read the original abstract

We show how to use the arguments of [CM2] to get a stronger effective version of uniqueness of blowups that has a number of consequences.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Regularity of cylindrical singular sets of mean curvature flow
math.DG 2025-09 conditional novelty 7.0 of 10

Degenerate cylindrical singular sets of mean curvature flow are locally contained in C^{2,α} submanifolds, with curvature determined by the flow's asymptotic profile.

Reference graph

Works this paper leans on

21 extracted references · 20 canonical work pages · cited by 1 Pith paper

[1]

Carson, J

T. Carson, J. Isenberg, D. Knopf and N. Sesum, Singularity formation of complete Ricci flow solutions , Adv. Math. 403 (2022), Paper No. 108326, 52 pp

work page 2022
[2]

Chodosh, K

O. Chodosh, K. Choi, C. Mantoulidis and F. Schulze, Mean curvature flow with generic low-entropy initial data , Duke Math. J. 173 (2024), no. 7, 1269--1290

work page 2024
[3]

Chodosh and F

O. Chodosh and F. Schulze, Uniqueness of asymptotically conical tangent flows , Duke Math. J. 170 (2021), no. 16, 3601--3657

work page 2021
[4]

K. Choi, R. Haslhofer and O. Hershkovits, Ancient low-entropy flows, mean-convex neighborhoods, and uniqueness , Acta Math. 228 (2022), no. 2, 217--301

work page 2022
[5]

Colding T

T.H. Colding T. Ilmanen and W.P. Minicozzi II, Rigidity of generic singularities of mean curvature flow. Publ. Math. IHES 121 (2015), 363--382

work page 2015
[6]

Colding and W.P

T.H. Colding and W.P. Minicozzi II, Generic mean curvature flow I; generic singularities, Annals of Math., Volume 175 (2012), Issue 2, 755--833

work page 2012
[7]

Colding and W.P

T.H. Colding and W.P. Minicozzi II, Uniqueness of blowups and Lojasiewicz inequalities. Ann. of Math. (2) 182 (2015), no. 1, 221--285

work page 2015
[8]

Colding and W.P

T.H. Colding and W.P. Minicozzi II, Lojasiewicz inequalities and applications , Surveys in differential geometry 2014. Regularity and evolution of nonlinear equations, 63--82, Surv. Differ. Geom., 19, Int. Press, Somerville, MA, 2015

work page 2014

Show all 21 references

[9]

Colding and W.P

T.H. Colding and W.P. Minicozzi II, The singular set of mean curvature flow with generic singularities , Invent. Math. 204 (2016), no. 2, 443--471

2016
[10]

Colding and W.P

T.H. Colding and W.P. Minicozzi II, Regularity of elliptic and parabolic systems , Ann. Sci. Ec. Norm. Super. (4) 56 (2023), no. 6, 1883--1921

2023
[11]

Gang, On the non-degenerate and degenerate generic singularities formed by mean curvature flow , Adv

Z. Gang, On the non-degenerate and degenerate generic singularities formed by mean curvature flow , Adv. Math. 457 (2024), Paper No. 109937

2024
[12]

Haslhofer and B

R. Haslhofer and B. Kleiner, Mean curvature flow with surgery , Duke Math. J. 166 (2017), no. 9, 1591--1626

2017
[13]

Huisken, Asymptotic behavior for singularities of the mean curvature flow

G. Huisken, Asymptotic behavior for singularities of the mean curvature flow. J. Differential Geom. 31 (1990), no. 1, 285--299

1990
[14]

Ilmanen, Singularities of Mean Curvature Flow of Surfaces, preprint, 1995, \\ http://www.math.ethz.ch/\ /papers/pub.html

T. Ilmanen, Singularities of Mean Curvature Flow of Surfaces, preprint, 1995, \\ http://www.math.ethz.ch/\ /papers/pub.html

1995
[15]

Lojasiewicz, Ensembles semi-analytiques, IHES notes (1965)

S. Lojasiewicz, Ensembles semi-analytiques, IHES notes (1965)

1965
[16]

Sun, Local entropy and generic multiplicity one singularities of mean curvature flow of surfaces , J

A. Sun, Local entropy and generic multiplicity one singularities of mean curvature flow of surfaces , J. Differential Geom. 124 (2023), no. 1, 169--198

2023
[17]

White, A local regularity theorem for mean curvature flow

B. White, A local regularity theorem for mean curvature flow. Ann. of Math. 161 (2005), 1487--1519

2005
[18]

White, Partial regularity of mean-convex hypersurfaces flowing by mean curvature, Int

B. White, Partial regularity of mean-convex hypersurfaces flowing by mean curvature, Int. Math. Res. Notices (1994) 185--192

1994
[19]

White, The size of the singular set in mean curvature flow of mean-convex sets , J

B. White, The size of the singular set in mean curvature flow of mean-convex sets , J. Amer. Math. Soc. 13 (2000), no. 3, 665--695

2000
[20]

White, The nature of singularities in mean curvature flow of mean-convex sets , J

B. White, The nature of singularities in mean curvature flow of mean-convex sets , J. Amer. Math. Soc. 16 (2003), no. 1, 123--138

2003
[21]

Zhu, Lojasiewicz inequalities for mean convex self-shrinkers , Int

J. Zhu, Lojasiewicz inequalities for mean convex self-shrinkers , Int. Math. Res. Not. IMRN 2023, no. 2, 1236--1254

2023

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.

[1] [1]

Carson, J

T. Carson, J. Isenberg, D. Knopf and N. Sesum, Singularity formation of complete Ricci flow solutions , Adv. Math. 403 (2022), Paper No. 108326, 52 pp

work page 2022

[2] [2]

Chodosh, K

O. Chodosh, K. Choi, C. Mantoulidis and F. Schulze, Mean curvature flow with generic low-entropy initial data , Duke Math. J. 173 (2024), no. 7, 1269--1290

work page 2024

[3] [3]

Chodosh and F

O. Chodosh and F. Schulze, Uniqueness of asymptotically conical tangent flows , Duke Math. J. 170 (2021), no. 16, 3601--3657

work page 2021

[4] [4]

K. Choi, R. Haslhofer and O. Hershkovits, Ancient low-entropy flows, mean-convex neighborhoods, and uniqueness , Acta Math. 228 (2022), no. 2, 217--301

work page 2022

[5] [5]

Colding T

T.H. Colding T. Ilmanen and W.P. Minicozzi II, Rigidity of generic singularities of mean curvature flow. Publ. Math. IHES 121 (2015), 363--382

work page 2015

[6] [6]

Colding and W.P

T.H. Colding and W.P. Minicozzi II, Generic mean curvature flow I; generic singularities, Annals of Math., Volume 175 (2012), Issue 2, 755--833

work page 2012

[7] [7]

Colding and W.P

T.H. Colding and W.P. Minicozzi II, Uniqueness of blowups and Lojasiewicz inequalities. Ann. of Math. (2) 182 (2015), no. 1, 221--285

work page 2015

[8] [8]

Colding and W.P

T.H. Colding and W.P. Minicozzi II, Lojasiewicz inequalities and applications , Surveys in differential geometry 2014. Regularity and evolution of nonlinear equations, 63--82, Surv. Differ. Geom., 19, Int. Press, Somerville, MA, 2015

work page 2014