Area-Preserving Curve-Shortening Flow

Updated 7 July 2026

Area-preserving curve-shortening flow is a nonlocal evolution of planar curves where the normal velocity equals the curvature minus its average, ensuring area conservation while reducing length.
The methodology involves gradient flow formulations, singularity and blow-up analyses, and classification of self-similar solutions to understand convergence and deviation from classical flow behavior.
Numerical approaches, including PFEM and dual-SAV schemes, are designed to preserve key geometric properties such as area, perimeter decrease, and mesh quality for stable discretizations.

Searching arXiv for recent and foundational work on APCSF and closely related geometric-flow results. The area-preserving curve-shortening flow (APCSF) is a nonlocal geometric evolution of planar curves in which the normal velocity is the curvature minus its spatial average, so that the enclosed area is preserved while the length decreases. In its classical closed-curve form, introduced by Gage, the flow is

$\partial_t \gamma = (\kappa-\bar\kappa)\,\nu,\qquad \bar\kappa(t)=\frac{1}{L(t)}\int_{\gamma_t}\kappa\,ds,$

where $\kappa$ is the signed curvature, $\nu$ the unit normal, and $L(t)$ the length of the evolving curve. For simple closed planar curves, $\int_{\gamma_t}\kappa\,ds=2\pi$ , hence $\bar\kappa=2\pi/L(t)$ (Gao et al., 2021). APCSF is the $L^2$ -gradient flow of length under an area constraint, and in this sense is the one-dimensional analogue of volume-preserving mean-curvature flow. Its theory spans closed and open curves, free-boundary problems with Neumann conditions, singularity formation and blow-up analysis, self-similar and homothetic solutions, and a growing numerical literature on structure-preserving discretizations (Mäder-Baumdicker, 2015, Mäder-Baumdicker, 2017, Cernomazov, 1 Aug 2025, Jiang et al., 2022, Sakakibara, 12 May 2026).

1. Definition and basic variational structure

For a smooth family of simple closed plane curves $\gamma:S^1\times[0,T)\to\mathbb{R}^2$ , the classical curve-shortening flow is

$\partial_t\gamma=\kappa\,\nu,$

whereas APCSF inserts the nonlocal correction $-\bar\kappa\,\nu$ to preserve enclosed area (Mäder-Baumdicker, 2017). In the standard planar orientation convention, $\kappa$ 0 is the average curvature,

$\kappa$ 1

For simple closed curves with turning number $\kappa$ 2, Gauss–Bonnet yields $\kappa$ 3, and therefore

$\kappa$ 4

in the notation used for Gage’s flow (Gao et al., 2021, Jiang et al., 2022).

The defining structural identities are area preservation and length dissipation. Under

$\kappa$ 5

the enclosed area satisfies

$\kappa$ 6

while the length satisfies

$\kappa$ 7

(Gao et al., 2021, Jiang et al., 2022). This identifies APCSF as a constrained steepest descent for length at fixed area. A plausible implication is that APCSF should be viewed less as a perturbation of curve-shortening flow than as an isoperimetric relaxation dynamics: it dissipates perimeter while remaining on a fixed-area manifold of immersions.

For closed curves, the curvature evolves by the nonlocal semilinear parabolic equation

$\kappa$ 8

(Gao et al., 2021). The extra quadratic term distinguishes APCSF from classical curve-shortening flow and is the source of both the global area constraint and several analytical complications in nonconvex regimes.

2. Closed-curve theory: convexity, star-shaped data, and convergence

The convex closed-curve theory is the classical core of APCSF. For convex initial data, Gage established that the flow preserves enclosed area, decreases length, exists globally, and converges after normalization to a round circle; this standard picture is recalled in later works (Gao et al., 2021, Jiang et al., 16 Apr 2026). The circle is the unique stationary solution among simple closed curves with fixed area, since stationarity forces $\kappa$ 9, hence constant curvature.

A major development concerns nonconvex initial curves. The paper “Star-shaped Centrosymmetric Curves under Gage’s Area-preserving Flow” proves that smooth embedded star-shaped curves that are also centrosymmetric evolve smoothly for all time, become convex in finite time, and converge to a circle as $\nu$ 0 (Gao et al., 2021). The argument combines a polar-graph formulation

$\nu$ 1

with a quasilinear parabolic PDE for the radial function,

$\nu$ 2

and uses preservation of star-shapedness and centrosymmetry to obtain curvature bounds and global existence (Gao et al., 2021).

That centrosymmetry assumption was later removed. The paper “Star-shaped Curves under Gage's Area-preserving Flow and the CSF” proves the folklore conjecture that smooth embedded star-shaped initial curves evolve globally under Gage’s area-preserving flow and converge to a circle (Gao et al., 2024). Its key mechanism is Dittberner’s singularity analysis, mediated through the Chow–Liou–Tsai turning-angle invariant $\nu$ 3. The paper shows that star-shaped embedded curves satisfy $\nu$ 4, which is sufficient to exclude finite-time singularities under APCSF (Gao et al., 2024). This result substantially enlarges the known nonconvex global-existence class.

At the same time, star-shapedness itself is not preserved in full generality. The same work constructs a star-shaped “flying wing” example showing that Gage’s area-preserving flow may lose star-shapedness during the evolution (Gao et al., 2024). This also yields a negative answer to Mantegazza’s open problem on whether curve-shortening flow always preserves star-shapedness (Gao et al., 2024). Thus, for APCSF, global regularity and preservation of a given geometric class can diverge sharply.

3. Free-boundary APCSF with Neumann conditions

A substantial branch of the theory concerns open curves whose endpoints lie on a support curve $\nu$ 5 and satisfy a Neumann free-boundary condition. In this setting, the evolution is

$\nu$ 6

with endpoints constrained to $\nu$ 7 and tangent orthogonality at the boundary (Mäder-Baumdicker, 2015, Mäder-Baumdicker, 2017). For a convex closed support curve $\nu$ 8, the boundary condition is

$\nu$ 9

and for the line support $L(t)$ 0, it becomes

$L(t)$ 1

(Mäder-Baumdicker, 2017).

Since the evolving curve is not closed, “area preservation” is defined relative to a boundary arc of $L(t)$ 2. One attaches the short piece $L(t)$ 3 joining the endpoints and defines the assembled closed curve $L(t)$ 4. Its oriented area is

$L(t)$ 5

and APCSF preserves this quantity while still decreasing the length of $L(t)$ 6 (Mäder-Baumdicker, 2015, Mäder-Baumdicker, 2017).

The 2015 paper on APCSF with Neumann free boundary conditions establishes a global-existence and subconvergence theory under geometric smallness assumptions on the initial curve. For strictly convex, embedded initial curves lying outside a convex support domain and satisfying a suitable length restriction, the flow exists for all time and subconverges smoothly to an arc of a circle meeting $L(t)$ 7 orthogonally and enclosing the same area with $L(t)$ 8 as the initial curve (Mäder-Baumdicker, 2015). The limiting stationary states are exactly circular arcs with orthogonal contact to the support.

This free-boundary theory already displays a dichotomy that later becomes central: either the flow exists globally and approaches a circular equilibrium, or it develops a finite-time singularity. The 2017 singularity paper sharpened this dichotomy considerably (Mäder-Baumdicker, 2017).

4. Singularity formation, type II blow-up, and grim-reaper limits

The paper “Singularities of the area preserving curve shortening flow with a Neumann free boundary condition” gives explicit geometric criteria guaranteeing finite-time singularity formation for APCSF outside a convex support curve or at a line (Mäder-Baumdicker, 2017). For initial curves satisfying $L(t)$ 9, with $\int_{\gamma_t}\kappa\,ds=2\pi$ 0 the minimum width of the support curve, and with total curvature in a specified interval

$\int_{\gamma_t}\kappa\,ds=2\pi$ 1

the flow must develop a finite-time singularity if either

$\int_{\gamma_t}\kappa\,ds=2\pi$ 2

$\int_{\gamma_t}\kappa\,ds=2\pi$ 3

(Mäder-Baumdicker, 2017).

A central result is the classification of any such finite-time singularity as type II, not type I. In the paper’s formulation, if $\int_{\gamma_t}\kappa\,ds=2\pi$ 4, then

$\int_{\gamma_t}\kappa\,ds=2\pi$ 5

and moreover

$\int_{\gamma_t}\kappa\,ds=2\pi$ 6

is unbounded as $\int_{\gamma_t}\kappa\,ds=2\pi$ 7 (Mäder-Baumdicker, 2017). This is a notable contrast with classical curve-shortening flow, where finite-time singularities of embedded convex curves are type I shrinkers. Here the area-preserving constraint shifts the blow-up mechanism toward translators.

The blow-up analysis uses Hamilton’s rescaling. Choosing spacetime points $\int_{\gamma_t}\kappa\,ds=2\pi$ 8 where

$\int_{\gamma_t}\kappa\,ds=2\pi$ 9

one rescales by

$\bar\kappa=2\pi/L(t)$ 0

Because the average-curvature term scales as $\bar\kappa=2\pi/L(t)$ 1, it disappears in the limit, and the blow-up limit solves the pure curve-shortening flow

$\bar\kappa=2\pi/L(t)$ 2

(Mäder-Baumdicker, 2017). For convex initial data satisfying the singularity criterion, the limit is either a full grim reaper or, at a boundary singularity, a half grim reaper meeting a line orthogonally (Mäder-Baumdicker, 2017).

The line-supported case is handled by reflection. Reflecting the free-boundary curve across the line $\bar\kappa=2\pi/L(t)$ 3 produces a closed curve $\bar\kappa=2\pi/L(t)$ 4, and Escher–Ito’s theory for closed APCSF can then be invoked. In this reflected setting, the turning index is always an odd integer, and finite-time singularity follows if either the reflected enclosed area is negative or an isoperimetric condition involving the index holds (Mäder-Baumdicker, 2017). This makes the free-boundary APCSF theory structurally parallel to the closed immersed-curve theory.

5. Self-similar solutions and shrinkers of APCSF

Although APCSF is area-preserving for embedded simple closed curves of turning number $\bar\kappa=2\pi/L(t)$ 5, homothetic solutions can still exist in the broader immersed setting because the enclosed signed area may vanish. The paper “Shrinkers of the area-preserving curve-shortening flow: Existence and saddle-point property” studies such homothetic evolutions and identifies their governing equation (Cernomazov, 1 Aug 2025).

$\bar\kappa=2\pi/L(t)$ 6

is a homothetic APCSF evolution, then after normalization the initial curve satisfies

$\bar\kappa=2\pi/L(t)$ 7

(Cernomazov, 1 Aug 2025). This differs from the classical self-shrinker equation for curve-shortening flow by the nonlocal additive term $\bar\kappa=2\pi/L(t)$ 8. A direct consequence is that APCSF shrinkers must have zero signed enclosed area: $\bar\kappa=2\pi/L(t)$ 9 (Cernomazov, 1 Aug 2025).

The paper connects APCSF shrinkers to the theory of $L^2$ 0-curves, namely solutions of

$L^2$ 1

In tangential polar coordinates, the curvature $L^2$ 2 of a $L^2$ 3-curve satisfies the ODE

$L^2$ 4

with first integral

$L^2$ 5

(Cernomazov, 1 Aug 2025). Closedness is controlled by a semi-period map $L^2$ 6; rational values of $L^2$ 7 correspond to closed $L^2$ 8-symmetric curves with tangent turning index $L^2$ 9 (Cernomazov, 1 Aug 2025).

Using this framework, the paper proves existence of non-circular APCSF shrinkers for every coprime pair $\gamma:S^1\times[0,T)\to\mathbb{R}^2$ 0 satisfying

$\gamma:S^1\times[0,T)\to\mathbb{R}^2$ 1

For each such rational $\gamma:S^1\times[0,T)\to\mathbb{R}^2$ 2, there exists a non-circular $\gamma:S^1\times[0,T)\to\mathbb{R}^2$ 3-symmetric APCSF shrinker with tangent turning index $\gamma:S^1\times[0,T)\to\mathbb{R}^2$ 4 (Cernomazov, 1 Aug 2025). This is an APCSF analogue of the Abresch–Langer theory for classical curve-shortening flow, although the classification is presently partial rather than complete.

The same paper also establishes a saddle-point property. Small normal perturbations of a non-circular APCSF shrinker bifurcate into two distinct dynamical regimes: one side evolves toward an $\gamma:S^1\times[0,T)\to\mathbb{R}^2$ 5-fold cover of a circle, while the other develops a singular curve with $\gamma:S^1\times[0,T)\to\mathbb{R}^2$ 6 cusps in finite time (Cernomazov, 1 Aug 2025). This suggests that APCSF shrinkers organize nearby dynamics in much the same way as Abresch–Langer curves do for curve-shortening flow.

6. Variants, nonstandard area-preserving flows, and affine analogues

The term “area-preserving curve-shortening flow” is most often reserved for Gage’s nonlocal flow with Lagrange multiplier $\gamma:S^1\times[0,T)\to\mathbb{R}^2$ 7, but nearby constructions also appear in the literature and illuminate what is specific to APCSF.

A 2025 paper studies a different nonlocal area-preserving curvature flow with normal velocity

$\gamma:S^1\times[0,T)\to\mathbb{R}^2$ 8

This flow also preserves area, decreases length, preserves convexity, and converges smoothly to a circle for strictly convex initial data (Sun et al., 23 Feb 2025). It is not the classical APCSF because the Lagrange multiplier depends on $\gamma:S^1\times[0,T)\to\mathbb{R}^2$ 9 rather than average curvature. The comparison is useful because it isolates which analytical features are robust under nonlocal perimeter-constrained flows and which depend specifically on the average-curvature structure.

There are also non-Euclidean analogues. The paper “A fourth-order area-preserving curve flow in centro-equiaffine geometry” constructs a local, fourth-order, area-preserving flow in which the velocity is

$\partial_t\gamma=\kappa\,\nu,$ 0

using the affine Minkowski identity $\partial_t\gamma=\kappa\,\nu,$ 1 (Jiang et al., 16 Apr 2026). This flow preserves Euclidean enclosed area, exists globally for smooth strictly convex origin-symmetric initial curves, and converges smoothly to a round circle up to $\partial_t\gamma=\kappa\,\nu,$ 2 (Jiang et al., 16 Apr 2026). It is not Euclidean APCSF: it is fourth-order, local rather than nonlocal, and affine rather than Euclidean in structure. Nonetheless, it clarifies a broader theme: area preservation can be enforced either through a nonlocal Lagrange multiplier, as in APCSF, or through a local geometric identity built into the velocity law.

7. Numerical discretization and structure preservation

Numerical work on APCSF focuses on preserving the flow’s defining geometry at the discrete level: area conservation, perimeter decrease, convexity preservation, and mesh quality.

The 2022 paper “A convexity-preserving and perimeter-decreasing parametric finite element method for the area-preserving curve shortening flow” develops a semi-discrete PFEM based on Dziuk’s mass-lumped formulation (Jiang et al., 2022). In parametric form, the continuous APCSF is written as

$\partial_t\gamma=\kappa\,\nu,$ 3

and the semi-discrete scheme is proved to preserve two fundamental geometric properties for initially convex curves: discrete convexity and perimeter decrease (Jiang et al., 2022). The paper also establishes an $\partial_t\gamma=\kappa\,\nu,$ 4 $\partial_t\gamma=\kappa\,\nu,$ 5-error estimate and reports approximately first-order convergence in the $\partial_t\gamma=\kappa\,\nu,$ 6-seminorm and close to second-order behavior in $\partial_t\gamma=\kappa\,\nu,$ 7 and velocity errors in numerical tests (Jiang et al., 2022).

More recent work emphasizes higher-order time integration and constraint handling. The 2026 dual-SAV PFEM framework treats APCSF as a constrained $\partial_t\gamma=\kappa\,\nu,$ 8-gradient flow of length with an area constraint enforced by a Lagrange multiplier, while using separate auxiliary variables for geometric energy and mesh regularization (Sakakibara, 12 May 2026). For one global area constraint, the nonlinear algebraic system per time step is reduced to dimension $\partial_t\gamma=\kappa\,\nu,$ 9, independently of the number of mesh vertices (Sakakibara, 12 May 2026). Numerical experiments for APCSF show near machine-precision area conservation together with significant mesh-quality improvement under tangential regularization (Sakakibara, 12 May 2026).

A complementary 2026 paper on arbitrary-order structure-preserving discretizations does not treat APCSF explicitly, but it develops a continuous Petrov–Galerkin framework that preserves area or volume laws for curvature-driven flows at arbitrary order in space and time (Zhang et al., 19 May 2026). It shows that conservation laws can be replicated discretely by introducing suitable auxiliary variables and test functions, and it explicitly notes that the same machinery can be adapted to APCSF by introducing a scalar Lagrange multiplier and an area constraint tested with $-\bar\kappa\,\nu$ 0 (Zhang et al., 19 May 2026). This suggests a route to genuinely high-order, structure-preserving APCSF solvers.

8. Relation to curve-shortening flow and broader geometric-flow theory

APCSF is best understood in comparison with classical curve-shortening flow. Under curve-shortening flow, convex embedded curves shrink to round points in finite time and type I self-similar shrinkers govern singularity models. Under APCSF, the area constraint prevents extinction and instead drives the evolution toward stationary constant-curvature states when global existence holds (Gao et al., 2021, Gao et al., 2024). For convex or suitably controlled nonconvex closed curves, this means convergence to a circle; for free-boundary problems, to a circular arc (Mäder-Baumdicker, 2015).

Yet the area constraint also creates new singular behavior. In free-boundary APCSF, finite-time singularities are type II and have grim-reaper rather than shrinking-soliton blow-up models (Mäder-Baumdicker, 2017). In immersed closed settings, APCSF admits non-circular homothetic shrinkers with zero signed area (Cernomazov, 1 Aug 2025). These two facts together show that APCSF blends elliptic equilibrium selection with singular translator dynamics in a way that has no direct analogue in the simplest embedded-curve theory for classical curve-shortening flow.

A recurring misconception is that “area-preserving” implies benign long-time behavior for all smooth initial data. The literature does not support that view. For convex closed data, the behavior is global and circular; for star-shaped embedded data, global existence still holds (Gao et al., 2024); but for general immersed or free-boundary data, finite-time singularities can occur, and explicit criteria are known (Mäder-Baumdicker, 2017, Cernomazov, 1 Aug 2025). Another misconception is that APCSF is merely curve-shortening flow with a harmless lower-order correction. The classification of type II blow-ups and the existence of non-circular APCSF shrinkers indicate that the nonlocal term qualitatively changes both singularity theory and the global phase portrait.

A plausible implication of the current literature is that APCSF occupies an intermediate position between unconstrained second-order flows and higher-order area-preserving flows such as curve diffusion. It retains a strong maximum-principle component, especially for curvature and convexity arguments, but already exhibits the rich dynamical structure typical of constrained geometric evolution: nonlocality, multiple equilibria in immersed classes, and delicate interactions between topology, turning number, and singularity formation.

9. Outlook

Several directions remain active. The first is classification. For closed embedded star-shaped curves, global existence is now well understood (Gao et al., 2024), but for general immersed curves the complete APCSF singularity landscape remains only partially charted. The recent theory of APCSF shrinkers suggests an Abresch–Langer-type classification indexed by rational turning-number data, but uniqueness up to similarity is still conjectural (Cernomazov, 1 Aug 2025).

A second direction concerns free boundaries and support geometry. The convex-support Neumann theory is comparatively mature (Mäder-Baumdicker, 2015, Mäder-Baumdicker, 2017), but more general support manifolds, anisotropies, and higher-dimensional analogues remain open. The reflection argument at a straight line suggests that some closed-curve theory can be imported to boundary problems, yet the boundary singularity structure is genuinely richer.

A third direction is discretization. Recent PFEM and dual-SAV frameworks show that discrete area preservation, perimeter decrease, convexity preservation, and mesh regularization can be achieved simultaneously in practical schemes (Jiang et al., 2022, Sakakibara, 12 May 2026). The arbitrary-order structure-preserving methodology developed for related curvature flows indicates that high-order APCSF solvers with exact discrete conservation laws are now technically plausible (Zhang et al., 19 May 2026).

In aggregate, APCSF has evolved from a constrained variant of curve-shortening flow into a distinct subject within geometric analysis: one in which nonlocality, isoperimetric optimization, singularity formation, and discrete structure preservation all play essential roles. Its central equations are simple, but the theory they generate is not; it includes global convergence theorems, translator blow-ups, free-boundary grim reapers, non-circular homothetic shrinkers, and numerically nontrivial conservation laws, all tied together by the single principle of preserving area while shortening curve length.