A Li-Yau inequality for the 1-dimensional Willmore energy (Q2691826)

The authors start recalling the Willmore energy associated to an immersion \( f:\Sigma \rightarrow \mathbb{R}^{n}\) of a surface \(\Sigma \) and defined as \( \mathcal{W}(f)=\frac{1}{4}\int_{\Sigma }\left\vert H\right\vert ^{2}d\mu \), where \(H\) is the mean curvature vector and \(\mu \) the Riemannian measure induced by pulling back the Euclidean metric to \(\Sigma \). Li and Yau proved that if \(\Sigma \) is compact, then \(\mathcal{W}(f)<8\pi \) implies that \(f\) is an embedding. In their paper, the authors consider the case of planar curves. For a closed smooth curve \(\gamma :\mathbb{S}^{1}\rightarrow \mathbb{ R}^{2}\), which is immersed, that is such that \(\left\vert \gamma ^{\prime }\right\vert >0\), and has signed curvature \(\kappa \), its elastic energy is defined by \(\mathcal{E}(\gamma )=\int_{\mathbb{S}^{1}}\left\vert \kappa \right\vert ^{2}ds\), and a natural scaling invariant is the total curvature defined by \(\mathcal{K}(\gamma )=\int_{\mathbb{S}^{1}}\left\vert \kappa \right\vert ds\). The authors first prove that \(2\pi =\inf\{\mathcal{K}(\gamma ):\gamma \in C^{2}(\mathbb{S}^{1};\mathbb{R}^{2})\) is a non-embedded immersion\(\}=\inf\{\mathcal{K}(\gamma ):\gamma \in C^{2}(\mathbb{S}^{1}; \mathbb{R}^{2})\) is an immersion\(\}\). Moreover, the infimum among non-embedded immersions is not attained. The main result of the paper proves that if \(\gamma \in W^{2,2}(\mathbb{S}^{1};\mathbb{R}^{2})\) is an immersed curve with \(\mathcal{E}(\gamma )\mathcal{L}(\gamma )<c^{\ast }=\mathcal{E} (\gamma ^{\ast })\mathcal{L}(\gamma ^{\ast })\), where \(\gamma ^{\ast }\) is the figure eight elastica, then \(\gamma \) is an embedding. The one-fold cover of the elastic figure eight is defined by \(\gamma ^{\ast }:[0,4K(m^{\ast })]\rightarrow \mathbb{R}^{2}\), \(\gamma ^{\ast }(s)=\left( \begin{array}{c} 2E(am(s,m^{\ast }),m^{\ast })-s \\ -2\sqrt{m^{\ast }}cn(s,m^{\ast }) \end{array} \right) \), where \(E(m)=\int_{0}^{\pi /2}\sqrt{1-m\sin ^{2}(\theta )}d\theta \) is the complete elliptic integral of second kind, \(m^{\ast }\in (0,1)\) is the unique root of the map \((0,1)\ni m\rightarrow 2E(m)-K(m)\), \( K(m)=\int_{0}^{\pi /2}\frac{1}{\sqrt{1-m\sin ^{2}(\theta )}}d\theta \) being the complete elliptic integral of first, \(F(\cdot ,m)=am(\cdot ,m)^{-1}\), \(F\) being the incomplete elliptic integral of first kind, and \( cn(x,m):=\cos(am(x,m))\). For the proof, the authors first recall properties of the Sobolev spaces \(W^{k,p}(\mathbb{S}^{1};\mathbb{R}^{l})\), a characterization of embeddedness for immersions \(\gamma \in C^{1}(\mathbb{S} ^{1};\mathbb{R}^{2})\). They also prove the existence of a minimizer to the problem \(\inf_{\gamma \in \mathcal{A}}\mathcal{E}(\gamma )\mathcal{L}(\gamma )\), where \(\mathcal{A}=\{\gamma \in W^{2,2}(\mathbb{S}^{1};\mathbb{R}^{2}): \gamma \) is a non-injective immersion\(\}\) and they prove properties of this minimizer, first deriving the Euler-Lagrange equation associated to this problem. They finally prove properties of elastica. The paper ends with results concerning the case where \(\gamma :[0,T)\times \mathbb{S} ^{1}\rightarrow \mathbb{R}^{2}\) is a family of smooth curves which evolves with respect to the gradient flow equation \(\partial _{t}\gamma =-\nabla _{s}^{2}\overrightarrow{\kappa }-\frac{1}{2}\left\vert \overrightarrow{ \kappa }\right\vert ^{2}\overrightarrow{\kappa }+\lambda \overrightarrow{ \kappa }\), where \(\overrightarrow{\kappa }=\overrightarrow{\kappa }\left[ \gamma \right] =\partial _{s}^{2}\gamma \) is the curvature vector and \( \nabla _{s}\) is the normal part of the arc-length derivative. The authors prove that if the initial data \(\gamma _{0}\in C^{\infty }(\mathbb{S}^{1}; \mathbb{R}^{2})\) is an embedded curve such that \(\mathcal{E}(\gamma _{0}) \mathcal{L}(\gamma _{0})<c^{\ast }\), the elastic flow with fixed length and initial datum \(\gamma _{0}\) remains embedded for all times and converges, as \(t\rightarrow \infty \), after reparametrization with constant speed to a one-fold cover of a circle with radius \(\mathcal{L}(\gamma _{0})/2\pi \). They finally prove that the above bound is optimal.