Non-uniqueness of self-similar shrinking curves for an anisotropic curvature flow (Q2492659): Difference between revisions

This paper deals with strictly convex simply closed smooth plane curves shrinking by the anisotropic curvature flow equation \(v=-\gamma(\theta) k\), where \(\theta\), \(v\), \(k\) are the outward normal angle, velocity and curvature, respectively, and \(\gamma(\theta)\) is a \(2\pi\)-periodic positive continuous function. It is known that for an arbitrary positive function \(\gamma\in C(S^1)\) there exists a positive function \(\sigma\in C^2(S^1)\) such that \(\left( \sigma^{\prime\prime}_{\theta\theta}+\sigma\right) \sigma =\gamma\). As a consequence, the evolution of curves governed by the equation \(v=-\gamma(\theta) k\) may be considered as the gradient flow for the anisotropic interfacial energy \(L=\int\limits_0^{l+0} \sigma (\theta(s) )ds\) with respect to some particular metric. Moreover, the family of curves represented by \[ x(\theta,t)=\sqrt{-2t}\left( \cos\theta \sigma(\theta) - \sin\theta \sigma^\prime_\theta(\theta )\right), \] \[ y(\theta,t)=\sqrt{-2t}\left( \sin\theta \sigma(\theta) + \cos\theta \sigma^\prime_\theta(\theta )\right), \] is a self-similar solution to \(v=-\gamma(\theta) k\) in \(t\in (-\infty, 0)\) which shrinks to \((0,0)\) at \(t\to 0\). The uniqueness of the self-similar solution was proved by \textit{M. Gage} and \textit{Y. Li} [Duke Math. J. 75, No.~1, 79--98 (1994; Zbl 0811.53033)], \textit{M. Gage}, [Duke Math. J. 72, No.~2, 441--466 (1993; Zbl 0798.53041)], \textit{C. Dohmen, Y. Giga} and \textit{N. Mizoguchi} [Calc. Var. Partial Differ. Equ. 4, No.~2, 103--119 (1996; Zbl 0847.34042)] for \(\pi\)-periodic positive continuous functions \(\gamma(\theta)\): For any positive \(\pi\)-periodic function \(\gamma \in C(\mathbb{R}^1)\) there exists a unique positive function \(\sigma\in C^2(S^1)\) such that \(\left( \sigma^{\prime\prime}_{\theta\theta}+\sigma\right) \sigma =\gamma\). It turns out that the uniqueness does not hold for \(2\pi\)-periodic \(\gamma\) which are not \(\pi\)-periodic. The author demonstrates that there exists two different \(2\pi\)-periodic positive functions \(\sigma\), \(\tilde\sigma \in C^\infty(\mathbb{R}^1)\) such that \(\left(\sigma^{\prime\prime}_{\theta\theta}+\sigma\right) \sigma = \left(\tilde\sigma^{\prime\prime}_{\theta\theta}+\tilde\sigma\right) \tilde\sigma>0\).