Rotational solitons for the curve shortening flow on revolution surfaces (Q6579332)

Consider a surface \(\mathbb{M}^2 \subset \mathbb{R}^3\). Then a map \(\hat{\Phi}: U \times I \rightarrow \mathbb{M}^2 \subset \mathbb{R}^3\) (where \(I = \left[0,T\right)\) for some \(T < \infty\)) which is a family of curves is a {curve shortening flow (CSF)} if\N\[\N\begin{cases} \left<\partial_t \hat{\Phi}\left(s,t\right), \eta\left(s,t\right)\right> = \kappa\left(s,t\right),\\\N\Phi\left(s,0\right) = \Phi, \end{cases}\N\]\Nwhere \(\eta\left(\cdot, t\right)\) and \(\kappa\left(\cdot, t\right)\) are, respectively, the unit normal vector and geodesic curvature of \(\hat{\Phi}\left(s,t\right)\). The authors of the current paper investigate when \(\hat{\Phi}\left(\cdot, t\right)\) is a ``soliton solution''. That is, when there exists a Killing vector field with associated flow \(\Psi\left(\cdot, t\right)\) such that\N\[\N\hat{\Phi}\left(s,t\right) = \Psi\left(\hat{\Phi}\left(s, t_0\right),t\right),\N\]\Nfor some fixed \(t_0 \in I\). Such a question was investigated by \textit{H. P. Halldorsson} in \(\mathbb{R}^2\) and \(\mathbb{R}^{1,1}\) (Minkowski space) [Trans. Am. Math. Soc. 364, No. 10, 5285--5309 (2012; Zbl 1281.53066); J. Reine Angew. Math. 704, 209--243 (2015; Zbl 1319.53071)], as well as in the sphere by \textit{H. F. S. dos Reis} and \textit{K. Tenenblat} [Proc. Am. Math. Soc. 147, No. 11, 4955--4967 (2019; Zbl 1428.53105)]. In the current paper, the authors specifically aim to answer the question: does every CSF which has a subsequence converging to a geodesic converge uniquely to a geodesic? We note here that a geodesic is a static solution to CSF.\N\NThe authors are able to show that the two ends of a soliton which rotates around the \(z\)-axis on a surface of revolution are asymptotic to parallel geodesics. More specifically, their main results are the following:\N\N\begin{itemize}\N\item[1.] A standard result is that the metric tensor for a surface of revolution is given by\N\[\Ng = du^2 + \varphi^2\left(u\right) dv^2\N\]\Nfor some function \(\varphi\) and parameters \(u \in I' \subset \mathbb{R}\) and \(v \in \left(0,2\pi\right)\). The authors show that, if \(\Phi\left(s\right)\) is a regular curve parametrized by arc-length, then \(\Phi\) is a soliton in the sense of the authors if and only if\N\[\N\kappa\left(s\right) = a \varphi\left(s\right)u'\left(s\right),\N\]\Nwhere \(a \in \mathbb{R}\) and \(\Phi\) is a geodesic when \(a = 0\).\N\item[2.] If the total geodesic curvature is bounded and \(\left|\varphi\right| + \left|\dot{\varphi}\right| < \infty\), then each end of \(\Phi\) is asymptotic to a parallel geodesic.\N\item[3.] As a corollary, the authors show that if \(\Phi\) is a simple closed curve, then \(\Phi\) is a parallel geodesic.\N\end{itemize}\N\NThe authors then apply their results to the torus \(\mathbb{T}^2\), which they note does not have a sign on its Gauss curvature, satisfies the assumptions of their main theorem, and the only isometries in such a case are rotations around the \(z\)-axis. They also show the catenoid does not satisfy the assumptions of their theorem.\N\NNote for the reader: the typesetting in the journal version of the paper has caused figures (which should be in Section \(5\)) to be on pages 13--15.