On length-preserving and area-preserving inverse curvature flow of planar curves with singularities

The curve $\gamma:I\rightarrow{\mathbb{R}^{2}}$ is referred to an Legendre curve, if there exists a unit vector field $\nu:I\rightarrow\mathbb{S}^{1}$ satisfying

for any $\theta\in I$. Here, $\gamma^{\prime}(\theta)=\frac{d\gamma}{d\theta}$ and $\langle\cdot,\cdot\rangle$ denotes the inner product. Notably, when the vector field $\nu$ is smooth, this special class of curves is termed frontal as introduced in [9]. This definition allows us to establish the Frenet frame even at singular points.

Let $\gamma$ be a frontal. In this context, the pair $({\nu(\theta),\mu(\theta)})$, satisfying $\mu(\theta)=J(\nu(\theta))$, constitutes a moving frame along the curve $\gamma(\theta)$ in $\mathbb{R}^{2}$, where $J$ represents a counterclockwise rotation of $\frac{\pi}{2}$ on $\mathbb{R}^{2}$. The Frenet formula for $\gamma$ is defined as:

Here, $\ell(\theta)=\langle\nu^{\prime}(\theta),\mu(\theta)\rangle$, and there exists a smooth function $\beta$ such that

Consequently, a Legendre curve $\gamma$ is regular if and only if $\beta$ never vanishes. The pair $(\ell,\beta)$ serves as a crucial invariant for Legendre curves, commonly referred to as the curvature pair of Legendre curves, as discussed in [10]. A straightforward computation reveals the relationship between the conventional curvature $\kappa$ and $\beta$ for Legendre curves, specifically, $\kappa=\frac{\ell}{|\beta|}$. Also, it is easy to see that the singular points on a Legendre curve satisfy $\beta=0$. Furthermore, if both $\ell$ and $\beta$ maintain consistent sign, the Legendre curve is convex, and vice versa, as detailed in [10]. In essence, Legendre curves can be seen as an extension of the convex curves. When $\ell>0$, an Legendre curve is denoted as an $\ell$-convex Legendre curve. Remarkably, as established in [33, Lemma 3.2], a Legendre curve can be transformed into one with $\ell=1$ through reparametrization. Therefore, for the sake of simplicity, we may exclusively focus on Legendre curves with $\ell=1$.

Let $\gamma:\mathbb{S}^{1}\rightarrow\mathbb{R}^{2}$ be an $\ell$-convex Legendre curve with $\ell=1$. According to [33, Lemma 3.3], the curve $\gamma$ can be expressed as

Here, $(\cos\theta,\sin\theta)$ is denoted as the unit vector field $\nu$ on $\mathbb{S}^{1}$. The function $p(\theta)$ serves the same role as the support function for a convex curve and is still called the support function for the $\ell$-convex Legendre curve $\gamma$ in [33]. Notably, $p(\theta)>0$ and $p(\theta)+p^{\prime\prime}(\theta)>0$ for any $\theta\in\mathbb{S}^{1}$, curve $\gamma$ turns into a convex curve. Moreover, in line with [33, Remark 3.5], we have

Much like convex curves, the algebraic length $L$ and the algebraic area $A$ of $\gamma$ can be defined, as detailed in [33, Definition 3.4], by the following expressions

It’s worth noting that replacing $p$ with $-p$ only alters the sign of $L$ while keeping the sign of $A$. Therefore, we may pay attention to $\ell$-convex curves with $L\geq 0$. Additionally, it’s important to highlight that the algebraic area $A$ of a Legendre curve may be positive, zero, or even negative values.

If we consider the $\ell$-convex curve $\gamma$ with a support function expressed as

then, through integration by parts and Parseval’s identity, we obtain

see e.g. [33]. Additionally, the coefficients $a_{1}$ and $b_{1}$ appeared in (1.6) own great significance in determining the position of the curve $\gamma$. Concretely,

and

The point $(a_{1},b_{1})$ is commonly referred to be the Steiner point for convex curves, see [43, p.50].

Due to the isoperimetric inequality for $\ell$-convex Legendre curves established in [33, Theorem 4.1], if its algebraic length $L$ of such a curve is equal to zero, then its algebraic area $A$ must be a negative value, unless the curve degenerates into a single point. However, when $L$ is greater than zero, the situation for $A$ becomes notably intricate. For instance, consider the support function of $\gamma$ in the form of $p(\theta)=2+\sin 2\theta$. In this case, the algebraic area $A$ equals to $\frac{5\pi}{2}$. Alternatively, if $p(\theta)=\sqrt{\frac{3}{2}}+\sin 2\theta$, the corresponding $A$ evaluates to zero. Lastly, for the support function $p(\theta)=\frac{1}{2}+\sin 2\theta$, the associated algebraic area $A$ becomes $-\frac{5\pi}{4}$; see Figure 1.

Inspired by the impressive inverse curvature flow researched by Li and Wang [33], we intend to consider the evolution problem for $\ell$-convex Legendre curves as follows

where $\lambda(t)$ is a nonlocal term.

The first purpose of this paper is to study nonlocal flows with $\lambda(t)=\frac{1}{L}\int_{0}^{2\pi}\beta^{2}d\theta$ or $\lambda(t)=\frac{L}{2\pi}$. From the evolution equations for $L$ and $A$ (see Lemma 2.1), the flow (1.9) is area-preserving when $\lambda(t)=\frac{1}{L}\int_{0}^{2\pi}\beta^{2}d\theta$, and it is length-preserving when $\lambda(t)=\frac{L}{2\pi}$.

For the area-preserving flow, one can get

For the length-preserving flow, one has

In essence, the flow (1.9) is an inverse flow for $\ell$-convex Legendre curves. In particular, when the initial curve becomes a convex curve, this flow represents Gao-Pan-Tsai’s area-preserving model [14] when $\lambda(t)=\frac{1}{L}\int_{0}^{2\pi}\beta^{2}d\theta$, and corresponds to Pan-Yang’s length-preserving flow [38] when $\lambda(t)=\frac{L}{2\pi}$. Unlike previous work on the inverse curvature flow for convex curves, see [13, 14], the long-term existence of the length-preserving and area-preserving flow for $\ell$-convex Legendre curves does not necessitate additional curvature conditions. More insights into the inverse curvature flow for convex curves can be found in [29], and the related literature therein.

The exploration of evolution problems for planar convex curves began with the comprehensive work of Gage and Hamilton [12]. They established the renowned result that a convex curve evolves into a point at finite time under the curve shortening flow, and the normalized of evolving curve converges to a circle as time goes to infinity. As a natural extension of this research, Gage [11] introduced a nonlocal area-preserving flow. The area-preserving flows with nonlocal speed for planar convex curves have since garnered substantial attention and undergone extensive investigations [14, 34, 35, 39, 44, 50]. Distinct from Gage’s area-preserving model, Pan and Yang [38] introduced an inverse curvature flow that is a length-preserving flow with nonlocal term. Further investigations into length-preserving flows have been undertaken in [13, 44]. For immersed curves, corresponding nonlocal area-preserving and length-preserving flows have been discussed by [41, 46, 47]. The nonlocal flows for general polygons are discussed in [17, 40]. In another significant extension, Grayson [19] demonstrated that any embedded planar curve can evolve into a convex one within finite time. This extends the Gage-Hamilton’s result to embedded curves and is known as the Gage-Hamilton-Grayson theorem. This theorem has also been studied in the context of nonlocal area-preserving and length-preserving flows, as detailed in [8, 36, 37, 15]. There is another nonlocal flow that is a gradient flow for the isoperimetirc ratio of evolving curve [28]. Research on the associated parabolic equations for evolving problems can be found in works such as [2, 3, 4, 18]. For a comprehensive understanding various aspects of evolving problems, readers can refer to the monographs [1] and [7].

For an $\ell$-convex Legendre curve $\gamma$ with algebraic length $L$ and algebraic area $A$, Li and Wang [33] have established a set of geometric inequalities:

Inequality (1.10) constitutes the isoperimetric inequality for $\ell$-convex Legendre curves. Differing from the classical isoperimetric inequality, it remains valid even for non-simple curves. Notably, the equalities in (1.10) and (1.11) hold if and only if $\gamma$ is a circle, while the equalities in (1.12) and (1.13) hold if and only if the support function of $\gamma$ follows the form

The proofs of these inequalities rely on Fourier series. For additional geometric inequalities related to convex curves, we refer the readers to [20, 31] and the references therein. It should be noted that if the support function of an $\ell$-convex Legendre curve is form (1.14), it is a parallel curve of an astroid centered at $(a_{1},b_{1})$. An example of the astroid can be seen in Figure 1(D) and the associated introduction to this kind of curve can be found in [33, Example 3.3] (see also [10]).

The second objective of this paper is to derive some geometric inequalities through the exploration of evolution problems concerning $\ell$-convex Legendre curves. As no geometric inequalities are employed in the discussion of the length-preserving inverse curvature flow presented in Theorem 1.3, this particular model is more advantageous in the pursuit of deriving geometric inequalities. To be specific, the length-preserving inverse curvature flow in Theorem 1.3 allows us to derive the isoperimetric inequality (1.10) and to obtain the generalizations for inequalities (1.11), (1.12), and (1.13).

The inverse curvature flow plays a pivotal role in the derivation of geometric inequalities, and there exists a body of research dedicated to this topic, with notable contributions found in [6, 23, 27], among others. The locally constrained inverse curvature flow, introduced by Brendle et al. [5] (see also Guan and Li [21]), was designed to ensure the monotonicity of certain geometric quantities and weaken the initial condition in some cases. The earliest investigations into the inverse curvature flow for hypersurfaces in Euclidean space can be traced back to Gerhardt [16] and Urbas [45]. Building upon these foundational works, Guan and Li [22] demonstrated the monotonicity of the ratio of quermassintegrals under the associated inverse curvature flow. They also provided a proof for Alexandrov-Fenchel inequalities applicable to $k$-convex and star-shaped domains. In recent times, a lot of research efforts have been dedicated to locally constrained inverse curvature flows, as evidenced by works such as [24, 25, 26, 32, 42, 48, 49], among others. For the latest developments focused on the inverse curvature flow for convex curves, readers can explore in [13, 14, 29, 50], and the literature therein.

This paper is organized as follows. In Section 2, we present some basic concepts and results about geometric flows for $\ell$-convex Legendre curves. In Section 3, we deal with the area-preserving flow. In Section 4, we research the length-preserving inverse curvature flow. In Section 5, we can get some geometric inequalities for $\ell$-convex Legendre curves through the length-preserving inverse curvature flow.