Combinatorial Calabi flow for ideal circle pattern

Shengyu Li; Zhigang Wang

REVIEW 3 major objections 5 minor 25 references

Combinatorial Calabi flow for ideal circle pattern

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The combinatorial Calabi flow exists for all time and converges exponentially fast to ideal circle patterns with prescribed discrete curvatures.

desk verdict New and correct convergence result for combinatorial Calabi flow on ideal circle patterns, with one imported technical lemma as the main dependency; deserves refereeing. read the letter →

arxiv 2501.01678 v2 pith:JRWZCLWW submitted 2025-01-03 math.DG math.CV

Shengyu Li , Zhigang Wang This is my paper · ORCID

classification math.DGmath.CV MSC 52C26

keywords idealcirclepatterncombinatorialCalabiflowdiscretecurvaturehyperbolicbackgroundgeometryEuclideanexponentialconvergencemappatterns

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

The paper proves that the combinatorial Calabi flow, a curvature-decreasing evolution for the radii of circles in a triangulated surface, works for ideal circle patterns in both hyperbolic and Euclidean background geometry. Under the angle condition (C1) and the assumption that the target curvature vector is attainable, the flow exists for all time and converges exponentially fast to a radius vector realizing the target curvatures. This turns the static existence problem for ideal circle patterns into a concrete algorithm: run the flow and read off the radii. The proof combines strict convexity of a well-defined potential with a large-radius derivative estimate, plus exponential decay of a squared-curvature energy.

What carries the argument

The engine is the curvature map written in the logarithmic variables $u_i=\ln\tanh(r_i/2)$ (hyperbolic) or $u_i=\ln r_i$ (Euclidean), together with the potential $\Psi(u)=\int_{u(0)}^u\sum_i(K_i-k_i)\,du_i$, whose Hessian is the Jacobian matrix $L$ of the curvature map. Strict convexity of $\Psi$ (from Lemma 4, where $L$ is positive definite, on a hyperplane in the Euclidean case) gives a unique critical point, the target radius vector. Long-time existence uses Lemma 5, an imported estimate that for any $\rho>1$ there is an $M$ such that $\rho\sum_{j\sim i}\partial K_j/\partial u_i+(\rho-1)\partial K_i/\partial u_i>0$ whenever $r_i>M$, to rule out a radius escaping to $+\infty$; coercivity of $\Psi$ rules out escape to $0$. Exponential convergence comes from the energy $C(u)=\sum_i(K_i-k_i)^2$, whose time derivative is $-2(K-k)^T L^2(K-k)\le -2\lambda_0 C(u)$.

What would settle it

On a genus-2 surface with any triangulation, choose an angle function $\Theta$ satisfying (C1) and a hyperbolic attainable $K$, then integrate (1.4); if the solution escapes to $0$ or $+\infty$ in finite time or does not converge exponentially, Theorem 1 is false. A cheaper algebraic check is to evaluate the Lemma 5 quantity $\rho\sum_{j\sim i}\partial K_j/\partial u_i+(\rho-1)\partial K_i/\partial u_i$ for large $r_i$ on that triangulation and see whether it is eventually positive for some $\rho>1$.

Watch

Extended reading notes

Core claim

On the paper's own terms, Theorem 1 is the central discovery: for any angle function $\Theta\colon E\to(0,\pi)$ satisfying (C1), and any hyperbolic (resp. Euclidean) attainable curvature vector $K=(k_1,\dots,k_N)$, the flow (1.4) (resp. (1.5)) has a unique solution for all time and converges exponentially fast to a radius vector $r^*$ such that the ideal circle pattern metric it determines has discrete curvatures $K$. In hyperbolic background this target is unique; in Euclidean background it is unique up to scaling, and the flow preserves the normalization $\sum_i u_i$ constant. The convergence proof bounds the squared curvature error by $C(u(0))e^{-2\lambda_0 t}$ on the compact invariant set where the flow lives, then transfers this exponential decay to the radius variables.

Load-bearing premise

The load-bearing premise is Lemma 5, an imported estimate saying that once one vertex radius is large enough, a particular weighted sum of curvature derivatives at that vertex is strictly positive; the paper does not prove this estimate for ideal circle patterns, and the hyperbolic long-time existence proof relies on it.

Editorial extensions

If this is right

For every curvature vector that satisfies the Bobenko–Springborn inequalities, the flow provides a constructive way to realize an ideal circle pattern metric with exactly those curvatures.
Because the convergence rate is exponential, discretized versions of (1.4) and (1.5) give a practical numerical algorithm for finding ideal circle patterns, not just an existence proof.
In the Euclidean case the flow automatically fixes the scaling ambiguity by preserving $\sum_i u_i$, so the limiting pattern is determined up to the same homothety freedom as the static problem.
The theorem extends the combinatorial Calabi flow from the non-obtuse-circle-pattern setting of earlier work to ideal circle patterns, where only the angle sum condition (C1) is required.
After the change of variables, both the hyperbolic and Euclidean flows take the same gradient form $\dot u=-(K-k)^T L$ in the appropriate domain, so the two cases are unified.

Reading between the lines

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

Since the proof treats Lemma 5 as a black box, a natural extension is to check numerically on random triangulations whether the Lemma 5 positivity holds for ideal circle patterns; if it ever fails, the hyperbolic long-time existence argument would need a different cap on radii.
The same Lyapunov-plus-coercivity template should apply to other combinatorial curvature flows for ideal patterns, such as $p$-th Calabi flows, provided an analogous large-radius estimate can be proved.
Quantifying $\lambda_0$, the smallest eigenvalue of $L^2$ on the invariant compact set, would turn the exponential convergence statement into an explicit rate and make the algorithm practically predictable.
Because ideal circle patterns correspond to ideal hyperbolic polyhedra, this flow can be read as a discrete Calabi deformation driving an ideal polyhedron to prescribed dihedral angles, complementing Rivin's uniqueness theorem.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 bottom line: this paper proves a genuinely new convergence theorem for the combinatorial Calabi flow on ideal circle patterns, in both hyperbolic and Euclidean background geometry. Prior Calabi flow results (Ge-Hua) handled circle patterns with non-obtuse exterior angles; prior Ricci flow work (Ge-Hua-Zhou) handled ideal patterns but with a different flow. This paper gets the Calabi flow for the ideal case, which requires controlling radii near 0 and infinity. The main theorem is clean and the proof is logically solid.

What the paper does well: it constructs the right Lyapunov function, shows it is strictly convex and proper on the domain, and uses it to rule out boundary blow-up in finite time. The hyperbolic escape-to-infinity argument is the delicate part; the Euclidean case is handled more simply using the invariant hyperplane. The paper is honest about its sources and even notes a possible concurrent independent result. That is good scholarly practice.

Soft spots, in proportion: the biggest dependency is Lemma 5, imported from Ge-Hua, Glickenstein-Thomas, Wu-Xu, and Li-Luo-Xu. The paper does not prove it and gives no sketch. This lemma is the essential tool for preventing a radius from escaping to +infinity in the hyperbolic flow. I checked the application of the lemma and it is arithmetically correct: a standard choice of the parameter converts the lemma's inequality into exactly the positivity the proof needs. So the dependency is real but legitimate. A referee would still want the lemma stated in a self-contained way and the reliance flagged. The Euclidean part is terse—"by a similar argument"—and the final convergence proof there is mostly sketched. For an expert this is fine; for verification it is a bit terse.

The citation pattern is appropriate; the self-citations are to closely related technical results. No circularity: the target curvature vector K is an input, not derived from the flow.

Who is this for: researchers working on combinatorial flows, circle patterns, and discrete conformal geometry. It will be a useful result to have in the literature. My recommendation: send it to a serious referee. It is not a result that will reshape the field, but it is correct, new, and well-framed.

Referee Report

3 major / 5 minor

Summary. The paper studies combinatorial Calabi flow for ideal circle patterns on closed triangulated surfaces, in both hyperbolic and Euclidean background geometry. The flow is defined in terms of discrete curvature K and a prescribed attainable target curvature vector k. The main result, Theorem 1, asserts that under the angle condition (C1) and attainability of k, the flow exists for all time and converges exponentially fast to a radius vector whose associated ideal circle pattern metric has discrete curvature k. The proofs go through a change of variables u (u_i = ln tanh(r_i/2) in the hyperbolic case, u_i = ln r_i in the Euclidean case), a Lyapunov function Lambda defined as the integral of (K-k)·du, strict convexity and coercivity of Lambda, and an energy-decay argument for exponential convergence. In the hyperbolic case, long-time existence is handled separately: boundary behavior toward u_i -> -infinity is controlled by the coercivity of Lambda, while escape to +infinity is ruled out using a sign estimate (Lemma 5 imported from prior work by Ge-Hua, Glickenstein-Thomas, Wu-Xu, and Li-Luo-Xu). The Euclidean case avoids the +infinity issue by working on the invariant hyperplane sum_i u_i = constant and using the coercivity of Lambda there.

Significance. If the result holds, it is a clean and useful extension of the combinatorial Calabi flow program to ideal circle patterns, providing a variational method and an algorithmic route to construct ideal circle pattern metrics with prescribed discrete curvature. The paper is concise, and the central geometric ingredients (strict convexity of the energy, positivity of the Jacobian, exponential convergence via the energy C) are standard and correctly assembled. The paper also gives explicit credit to the prior work on which it relies, especially Lemma 5, and it records the independent related work of X. Zhang. The main mathematical risk is the correctness of the imported Lemma 5 for ideal circle patterns in the hyperbolic regime; the paper does not prove Lemma 5, so the long-time existence proof in Theorem 2 is conditional on an external estimate. That is a normal dependency in the field, not an internal flaw. Overall the paper is a solid contribution that fits the journal's scope.

major comments (3)

[Section 3.2, Theorem 2] The long-time existence proof for the hyperbolic flow depends entirely on Lemma 5 to rule out ri(t) -> +infinity. Lemma 5 is not proved in the paper; it is imported from [8], [11], [13], [23]. The use of the lemma is arithmetically consistent: with A = -sum_{j~i} dK_j/du_i > 0 and B = dK_i/du_i > 0, the condition rho*A - (rho-1)*B > 0 is exactly the positivity required in the display after (3.2). However, since Theorem 2 is the only place where escape to +infinity is excluded for the hyperbolic flow, the main theorem for hyperbolic background geometry is conditional on the correctness of that external estimate. I would like the authors to state this dependency explicitly in the text and, if possible, to include a proof or a precise reference to the exact statement of Lemma 5 in each cited source. This is a load-bearing external dependency, not a fatal flaw, but it should be transparently acknowledged.
[Section 3.3, Theorem 3] In the proof of exponential convergence, after deriving |Ki - ki| <= sqrt(C(u(0))) e^{-lambda0 t}, the authors estimate |ui - u*_i| by integrating d(ui - u*_i)/dt = sum_j (Kj - kj) dKj/du_i. The bound uses uniform boundedness of dKj/du_i on the compact set containing u(t). This step is correct, but the constant lambda in the final display is not explicitly related to lambda0 and the uniform bound; it would be helpful to spell out how the two constants combine (e.g., lambda = sqrt(C(u(0))) * sup |dKj/du_i| / lambda0). The claim as written is acceptable, but the estimate deserves a one-line justification.
[Section 4.2, Theorem 4] The proof of long-time existence in the Euclidean case avoids Lemma 5 entirely and uses the coercivity of Lambda on the invariant hyperplane. However, the claim that 'the flow never touches the boundary of Υ in any finite time interval' is not fully justified as written. Since u stays in a compact set of Υ whenever T0 is finite (by the contradiction argument), the statement follows, but the wording is loose. Also, the proof of the 'similar argument' for exponential convergence is not given; given that the Jacobian is only positive definite on Υ (with null space along (1,...,1)), it is worth noting explicitly that the energy C decays via L^2 restricted to Υ and that the same computation as in Theorem 3 applies. This is a minor expositional gap, not an error.

minor comments (5)

[Section 1.2] The notation T h is used both for the curvature map and for the hyperbolic background geometry; the superscript is h, which is fine, but the map from R^N_+ to R^N and the map from R^N_- (or R^N) in u-coordinates could be given different names or a remark that they are the same map in different coordinates.
[Section 2, Lemma 5] Lemma 5 is stated without specifying the precise quantitative dependence of M on rho and on the triangulation/angle function. Since the lemma plays a central role, a phrase such as 'for some M = M(rho, T, Theta, ...)' would clarify the uniformity in the subsequent argument.
[Section 3.2, equation (3.2)] The condition Ki > 2πη > kmax is used, but the parameter η is introduced via Lemma 3. It would be clearer to define η as a fixed number in (0,1) such that kmax < 2πη, and to state explicitly that Lemma 3 gives M1 depending on η. The current text is correct but a bit implicit.
[Section 4.1] The proof of Lemma 8 is omitted with a reference to the hyperbolic argument. Since the Euclidean domain is R^N rather than R^N_-, and the invariance of sum_i u_i is used to restrict to Υ, a sentence explaining that Lemma 7 still applies on the unbounded convex set Υ would improve readability.
[General] The paper has several typographical issues: 'P A TTERN' in the title header, 'W ang' for the author name, and a space before the comma in 'j ∼ i' in the display after (3.3). These should be corrected in the final version.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the flow convergence proof is self-contained given independent external estimates, and the target curvature vector is a genuine input.

full rationale

The paper's central claim is that the combinatorial Calabi flow (1.4)/(1.5) converges to the unique ideal circle pattern metric whose discrete curvature vector is a prescribed attainable K. The target K is an input, not an output of the flow: the flow is driven by the error K - k, and the Lyapunov function is built from that error. Existence of the target radius vector comes from Theorem A, an external result of Bobenko-Springborn characterizing the image of the curvature map; this is not a restatement of the flow theorem. The positive-definiteness of the Jacobian (Lemma 4) and the large-radius estimate (Lemma 5) are cited from prior work. Although Lemma 5 is attributed in part to the authors' own paper [13], the paper states that Glickenstein-Thomas [11], Wu-Xu [23], and Ge-Hua [8] also provided proofs, so the estimate is independently supported and not a self-citation chain. The proof of Theorem 2 uses Lemma 5 only to rule out escape to r_i = +∞ in the hyperbolic case; it is a technical curvature estimate, not an assumption equivalent to the theorem's conclusion. The Euclidean case avoids Lemma 5 and uses coercivity on the invariant hyperplane. No fitted parameters are renamed as predictions, and no known result is repackaged as new. The derivation chain is therefore not circular.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The central claim rests on the cited analytic lemmas (especially Lemma 5) and the Bobenko-Springborn characterization; these are established results from the same research line, but the paper does not reprove them. No free parameters or invented entities are used.

assumptions (6)

domain assumption Lemma 4: curvature Jacobian L is symmetric and positive definite (Euclidean case: on any hyperplane sum u_i = constant).
Cited from Ge-Hua-Zhou [10, Lemmas 2.1 and 3.1]. Used in Sections 3.1 and 4.1 to make Psi strictly convex and C(u) decay.
domain assumption Lemma 5: for any rho > 1 there exists M > 0 with rho * sum_{j ~ i} dK_j/du_i + (rho - 1) * dK_i/du_i > 0 when r_i > M.
Cited from Ge-Hua [8, Lemmas 2.2 and 2.3], Glickenstein-Thomas [11, Proposition 9], Wu-Xu [23], Li-Luo-Xu [13]. Load-bearing in Theorem 2 to rule out r_i going to +infinity.
domain assumption Lemma 3: in hyperbolic geometry, every angle at vertex i is less than epsilon once r_i > M.
Cited from Ge-Hua-Zhou [10, Lemma 1.5]. Used to make K_i > k_max for large r_i in Theorem 2.
domain assumption Theorem A (Bobenko-Springborn): characterization of attainable curvature vectors and injectivity of the curvature map.
Cited in Section 1.2. Guarantees a unique u* with K(u*) = k, the critical point of Psi.
standard math Lemma 7: a strictly convex smooth function on an unbounded convex set with a critical point is coercive along the boundary.
Cited from Ge-Hua-Zhou [9, Lemma 2.8]; standard convex analysis fact used to prove Lemmas 6 and 8.
standard math ODE existence and uniqueness theory plus Lyapunov stability theorems.
Used in Section 3.2 to get local existence and in Theorem 3 for exponential decay.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Combinatorial Calabi flow for ideal circle pattern." pith.science (2026). https://pith.science/paper/JRWZCLWW

@misc{pith2026250101678,
  author       = {Pith},
  title        = {Pith review of: Combinatorial Calabi flow for ideal circle pattern},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JRWZCLWW}},
  note         = {Machine review of arXiv:2501.01678}
}

read the original abstract

We study the combinatorial Calabi flow for ideal circle patterns in both hyperbolic and Euclidean background geometry. We prove that the flow exists for all time and converges exponentially fast to an ideal circle pattern metric on surfaces with prescribed attainable curvatures. As a consequence, we provide an algorithm to find the desired ideal circle patterns.

Figures

Figures reproduced from arXiv: 2501.01678 by the authors.

**Figure 1.** An ideal circle pattern. Let Θ : E → (0, π) be an angle function. A natural question asks whether there exists a T -type ideal circle pattern with a given exterior intersection function Θ. Thurston [20, Chap.13] observed the relation between circle patterns and hyperbolic polyhedra. Let B 3 denote the hyperbolic 3-space. Given a convex hyperbolic polyhedron Q in B 3 , the plane of each polyhedron face will cut a ci… view at source ↗

**Figure 2.** A two-circle configuration. Let △vivjvk be the triangle by connecting the centers vi ,vj and one of the intersection points of two circles (see [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 23 canonical work pages

[8]

H. Ge, B. Hua, On combinatorial Calabi flow with hyperbolic circle patterns , Adv. Math. 333 (2018), 523–538

work page 2018
[11]

Glickenstein, J

D. Glickenstein, J. Thomas, Duality structures and discrete conformal variations of piecewise constant curvature surfaces, Adv. Math. 320 (2017), 250–278

work page 2017
[13]

S. Li, Q. Luo, Y. Xu, Combinatorial Calabi flow on surfaces of finite topological type , Proc. Amer. Math. Soc. 152 (2024), 4035–4047

work page 2024
[23]

T. Wu, X. Xu, Fractional combinatorial Calabi flow on surfaces , arXiv:2107.14102, 2021

arXiv 2021
[1]

E. A. Barbaˇ shin, N. N. Krasovskii, On the existence of Lyapunov functions in the case of asymptotic stability in the large , (Russian) Akad. Nauk SSSR. Prikl. Mat. Meh. 18 (1954), 345–350

work page 1954
[2]

A. I. Bobenko, B. A. Springborn, Variational principles for circle patterns and Koebe’s the- orem, Trans. Amer. Math. Soc. 356 (2004), 659–689

work page 2004
[3]

B. Chow, F. Luo, Combinatorial Ricci flows on surfaces , J. Differential Geom. 63 (2003), 97–129

work page 2003
[4]

F. H. Clarke, Y. S. Ledyaev, R. J. Stern, Asymptotic stability and smooth Lyapunov functions, J. Differential Equations 149 (1998), 69–114

work page 1998

Show all 25 references

[5]

Colin de Verdi` ere, Un principe variationnel pour les empilements de cercles , (French) Invent

Y. Colin de Verdi` ere, Un principe variationnel pour les empilements de cercles , (French) Invent. Math. 104 (1991), 655–669

1991
[6]

Ge, Combinatorial methods and geometric equations , Thesis (Ph.D.), Peking University, Beijing, 2012

H. Ge, Combinatorial methods and geometric equations , Thesis (Ph.D.), Peking University, Beijing, 2012

2012
[7]

Ge, Combinatorial Calabi flows on surfaces , Trans

H. Ge, Combinatorial Calabi flows on surfaces , Trans. Amer. Math. Soc. 370 (2018), 1377– 1391

2018
[9]

H. Ge, B. Hua, Z. Zhou, Circle patterns on surfaces of finite topological type , Amer. J. Math. 143 (2021), 1397–1430

2021
[10]

H. Ge, B. Hua, Z. Zhou, Combinatorial Ricci flows for ideal circle patterns , Adv. Math. 383 (2021), Paper No. 107698, 26 pp

2021
[12]

G. Hu, Z. Lei, Y. Qi, P. Zhou, Combinatorial p-th Calabi flows for total geodesic curvatures in hyperbolic background geometry, J. Geom. Anal. 35 (2025), Paper No. 18

2025
[14]

A. Lin, X. Zhang, Combinatorial p-th Calabi flows on surfaces , Adv. Math. 346 (2019), 1067–1090. 14 SHENGYU LI AND ZHIGANG W ANG

2019
[15]

J. Liu, Z. Zhou, How many cages midscribe an egg , Invent. Math. 203 (2016), 655–673

2016
[16]

Rivin, A characterization of ideal polyhedra in hyperbolic 3-space, Ann

I. Rivin, A characterization of ideal polyhedra in hyperbolic 3-space, Ann. of Math. (2) 143 (1996), 51–70

1996
[17]

Rodin, D

B. Rodin, D. Sullivan, The convergence of circle packings to the Riemann mapping , J. Dif- ferential Geom. 26 (1987), 349–360

1987
[18]

Schramm, How to cage an egg , Invent

O. Schramm, How to cage an egg , Invent. Math. 107 (1992), 543–560

1992
[19]

Stephenson, Introduction to circle packing

K. Stephenson, Introduction to circle packing. The theory of discrete analytic functions, Cam- bridge University Press, Cambridge, 2005

2005
[20]

Thurston, Geometry and topology of 3-manifolds, Princeton Lecture Notes, 1976

W. Thurston, Geometry and topology of 3-manifolds, Princeton Lecture Notes, 1976

1976
[21]

Thurston, The finite Riemann mapping theorem

W. Thurston, The finite Riemann mapping theorem. Invited talk, an international symposium at Purdue University on the occasional of the proof of the Bieberbach Conjecture, 1985

1985
[22]

Walter, Ordinary Differential Equations , Springer, New York, 1998

W. Walter, Ordinary Differential Equations , Springer, New York, 1998

1998
[24]

X. Xu, C. Zheng, Combinatorial curvature flows for generalized circle packings on surfaces with boundary, Int. Math. Res. Not. IMRN 2023 (2023), 17704–17728

2023
[25]

Zhou, Circle patterns with obtuse exterior intersection angles , arXiv:1703.01768, 2017

Z. Zhou, Circle patterns with obtuse exterior intersection angles , arXiv:1703.01768, 2017. Shengyu Li School of Mathematics and Statistics, Hunan First Normal University, Changsha 410205, Hunan, P. R. China. Email address : lishengyu@hnu.edu.cn Zhigang W ang School of Mathema...

2017 arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.

[1] [8]

H. Ge, B. Hua, On combinatorial Calabi flow with hyperbolic circle patterns , Adv. Math. 333 (2018), 523–538

work page 2018

[2] [11]

Glickenstein, J

D. Glickenstein, J. Thomas, Duality structures and discrete conformal variations of piecewise constant curvature surfaces, Adv. Math. 320 (2017), 250–278

work page 2017

[3] [13]

S. Li, Q. Luo, Y. Xu, Combinatorial Calabi flow on surfaces of finite topological type , Proc. Amer. Math. Soc. 152 (2024), 4035–4047

work page 2024

[4] [23]

T. Wu, X. Xu, Fractional combinatorial Calabi flow on surfaces , arXiv:2107.14102, 2021

arXiv 2021

[5] [1]

E. A. Barbaˇ shin, N. N. Krasovskii, On the existence of Lyapunov functions in the case of asymptotic stability in the large , (Russian) Akad. Nauk SSSR. Prikl. Mat. Meh. 18 (1954), 345–350

work page 1954

[6] [2]

A. I. Bobenko, B. A. Springborn, Variational principles for circle patterns and Koebe’s the- orem, Trans. Amer. Math. Soc. 356 (2004), 659–689

work page 2004

[7] [3]

B. Chow, F. Luo, Combinatorial Ricci flows on surfaces , J. Differential Geom. 63 (2003), 97–129

work page 2003

[8] [4]

F. H. Clarke, Y. S. Ledyaev, R. J. Stern, Asymptotic stability and smooth Lyapunov functions, J. Differential Equations 149 (1998), 69–114

work page 1998