Mesh-preserving and energy-stable parametric FEM for geometric flows of surfaces (Q7002660)

The author develops mesh-preserving and energy-stable parametric finite element schemes for the mean curvature flow and surface diffusion of 2D surfaces. If \(\mathcal{M}\) is a regular 2D surface in \(\mathbb{R}^{3}\) and \( \overrightarrow{X}(t):\mathcal{M}\rightarrow \Gamma (t)\) a diffeomorphism that parameterizes \(\Gamma (t)\) onto \(\mathcal{M}\), the evolution of \(\Gamma (t)\) can be expressed through \[\frac{\partial \overrightarrow{X}(t)}{ \partial t}=\overrightarrow{v}(t,\overrightarrow{x}(t))\circ \overrightarrow{ X}(t),\] where \(\overrightarrow{v}(t)\in \mathbb{R}^{3}\) is a vector field on \(\Gamma (t)\) such that \(\overrightarrow{v}\cdot \overrightarrow{\nu }=H\) in the case of the mean curvature flow, and \(\overrightarrow{v}\cdot \overrightarrow{\nu }=\Delta _{\Gamma }H\) for surface diffusion, \(H=\nabla _{\Gamma (t)}(t)\cdot \overrightarrow{\nu }\) being the mean curvature and \( \overrightarrow{\nu }\) the outward unit normal vector of \(\Gamma (t)\). \N\NIn his previous paper [SIAM J. Sci. Comput. 46, No. 1, A587--A608 (2024; Zbl 1536.53177)] with \textit{B. Li}, the author introduced the reformulation \[ \frac{\partial \overrightarrow{X}}{\partial t}=(\overrightarrow{v}\cdot \overrightarrow{\nu })\overrightarrow{\nu }\circ \overrightarrow{X}(t)+k(t, \overrightarrow{x})(\mathbb{P}_{\Gamma }\circ \overrightarrow{X})\Delta _{ \mathcal{M}}\overrightarrow{X}\] of the preceding problem, where \(\mathbb{P} _{\Gamma }=I-\overrightarrow{\nu }\otimes \overrightarrow{\nu }\) and \(k(t, \overrightarrow{x}):[0,T]\times \mathcal{M}\rightarrow \mathbb{R}_{\geq 0}\). \N\NThe purpose of the present paper is to propose a family of nonlinear schemes based on this reformulation. The author starts proposing a weak formulation of the previous problem.\ For example, in the case of the mean curvature flow, the weak formulation is: Find \(\overrightarrow{X}\in H^{1}(\mathcal{M})^{3}\) and \( \overrightarrow{\kappa }\in H^{1}(\Gamma (t))^{3}\) such that \[\int_{\Gamma (t)}\frac{\partial \overrightarrow{X}}{\partial t}\circ \overrightarrow{X} ^{-1}\cdot \overrightarrow{\xi }d_{\Gamma }+\int_{\Gamma (t)}\overrightarrow{ \kappa }\cdot \overrightarrow{\xi }d_{\Gamma }+\alpha \int_{\mathcal{M} }(\nabla _{\mathcal{M}}\overrightarrow{X}):\nabla _{\mathcal{M}}(\mathbb{P} _{\Gamma }\overrightarrow{\xi }\circ \overrightarrow{X})d_{\mathcal{M}}=0\] \[ \int_{\Gamma (t)}\nabla _{\Gamma (t)}Id:\nabla _{\Gamma (t)}\overrightarrow{ \theta }d_{\Gamma }=\int_{\Gamma (t)}\overrightarrow{\kappa }\cdot \overrightarrow{\theta }d_{\Gamma }\] holds for all \(\overrightarrow{\xi }\) and \(\overrightarrow{\theta }\) in \(H^{1}(\Gamma (t))^{3}\). \N\NLet \(\mathcal{M} _{h}\) be a polyhedral surface which consists of nondegenerate simplices \(\{ \widetilde{T}_{k}\}_{k=1}^{N_{T}}\) and interpolates the reference surface \( \mathcal{M}\) at the vertices \(\{\widetilde{p}_{k}\}_{k=1}^{N_{p}}\). For any polyhedral surface \(\mathcal{N}_{h}\), let \[S(\mathcal{N}_{h})=\{w\mid w \text{is continuous and piecewise linear on each simplex}\}\] and \[W(\mathcal{N} _{h})=\{w\mid w \, \text{is constant on each simplex}\},\] and let \[\overrightarrow{X} _{h}(t)\in S(\mathcal{M}_{h})^{3}\] be the parameterization of \(\Gamma _{h}(t)\). The author assumes that \(\Gamma _{h}(t)\) is formed by nondegenerate simplices \(\{T_{k}\}_{k=1}^{N_{T}}\) with vertices \(\{p_{k}\}_{k=1}^{N_{p}}\). He introduces a weighted outward normal vector \(\overrightarrow{\nu }_{h}\in S(\Gamma _{h}(t))^{3}\) based on the unit outward normal vector \( \overrightarrow{n}_{h}\in W(\Gamma _{h}(t))^{3}\) of \(\Gamma _{h}(t)\). For a semi-discrete scheme, he introduces a discrete curvature vector \[ \overrightarrow{\kappa }_{h}=-\Delta _{\Gamma (t)}^{h}Id\in S(\Gamma _{h}(t))^{3}\] and a modified unit outward normal \(\overrightarrow{\mu } _{h}\in S(\Gamma _{h}(t))^{3}\) that allow building the semi-discrete schemes for the mean curvature flow or surface diffusion. The author also proposes an Euler scheme, then a second-order scheme, for which he proves an energy estimate. In the last part of his paper, the author presents and discusses the results obtained with the different schemes with some examples.