Weak-strong uniqueness for volume-preserving mean curvature flow (Q6118794)
From MaRDI portal
scientific article; zbMATH DE number 7810706
Language | Label | Description | Also known as |
---|---|---|---|
English | Weak-strong uniqueness for volume-preserving mean curvature flow |
scientific article; zbMATH DE number 7810706 |
Statements
Weak-strong uniqueness for volume-preserving mean curvature flow (English)
0 references
28 February 2024
0 references
Let \(\Sigma ^{\ast }=(\Sigma ^{\ast }(t))_{t\in \lbrack 0,T^{\ast })}\), with \(\Sigma ^{\ast }(t)=\partial \Omega ^{\ast }(t)\) and \(\Omega ^{\ast }(t)\) bounded (for example \(\Omega ^{\ast }(t)\subset B_{R}(0)\) for all \(t\in \lbrack 0,T^{\ast }]\)). \(\Sigma ^{\ast }(t)\) is a regular solution of the volume-preserving mean curvature flow if \(\Sigma ^{\ast }(t)\) is of class \(C^{3,\alpha }\) with normal velocity field \(V^{\ast }\) of class \(C^{2,\alpha } \), and for all \(t\in \lbrack 0,T^{\ast }]\) \[ V^{\ast }=-H^{\ast }+\lambda ^{\ast } \] holds on \(\Sigma ^{\ast }(t)\), where \(H^{\ast }\) denotes the mean curvature of the evolving surface \(\Sigma ^{\ast }(t)\) and the function \(\lambda ^{\ast }(t)\) is the Lagrange multiplier corresponding to the volume constraint \(\left\vert \Omega (t)\right\vert =\left\vert \Omega (0)\right\vert =m^{\ast }\), and is explicitly given by \[ \lambda ^{\ast }(t)= \frac{1}{\mathcal{H}^{d-1}(\Sigma ^{\ast }(t))}\int_{\Sigma ^{\ast }(t)}H^{\ast }d\mathcal{H}^{d-1}. \] The first main result proves that if \( \Sigma ^{\ast }(t)\) is a regular solution to the volume-preserving mean curvature flow, then there exists a gradient-flow calibration \((\xi ,B,\vartheta ,\lambda ^{\ast })\) of \(\Sigma ^{\ast }\), which means \begin{itemize} \item \(\xi \in C_{c}^{1}(\mathbb{ R}^{d}\times \lbrack 0,T^{\ast }];\mathbb{R}^{d})\) satisfies \(\xi (\cdot ,t)=\nu ^{\ast }(\cdot ,t)\) on \(\Sigma ^{\ast }(t)\) and \(\left\vert \xi (\cdot ,t)\right\vert \leq \max (1-c\mathrm{dist}^{2}(\cdot ,\Sigma ^{\ast }(t)),0)\) for some positive constant \(c\), and is a solution to \[ (\partial _{t}\left\vert \xi \right\vert ^{2}+(B\cdot \nabla )\left\vert \xi \right\vert ^{2})(\cdot ,t)=O(\mathrm{dist}^{2}(\cdot ,\Sigma ^{\ast }(t))) \] and \item \( B(\cdot ,t)\in C^{1,1}(\mathbb{R}^{d};\mathbb{R}^{d})\) satisfies for each \(t\in \lbrack 0,T^{\ast }\) \[ \nabla \cdot B(\cdot ,t)=O(\mathrm{dist}(\cdot ,\Sigma ^{\ast }(t))) \] and \(\vartheta \in C^{0,1}(\mathbb{R}^{d}\times \lbrack 0,T^{\ast })\cap L^{\infty }(\mathbb{R}^{d}\times \lbrack 0,T^{\ast })\) is the solution to \[ ((\partial _{t}\vartheta +(B\cdot \nabla )\vartheta )(\cdot ,t)=O(\mathrm{dist}(\cdot ,\Sigma ^{\ast }(t))) \] \item \(\lambda ^{\ast }:[0,T^{\ast }]\rightarrow \mathbb{R}\) is given by \[ \lambda ^{\ast }(t)=\frac{1}{\mathcal{ H}^{d-1}(\Sigma ^{\ast }(t))}\int_{\Sigma ^{\ast }(t)}\nabla \cdot \xi (\cdot ,t)d\mathcal{H}^{d-1} \] \item there exist a constant \(C<1\) and a function \( f:\mathbb{R}^{d}\times \lbrack 0,T^{\ast }]\rightarrow \mathbb{R}\) with \[ \left\Vert f(\cdot ,t)\right\Vert _{L^{\infty }}\leq C \] for all \(t\in \lbrack 0,T^{\ast }\) such that the vector field \(\xi \) is almost transported by \(B\) in the sense that \[ (\partial _{t}\xi +(B\cdot \nabla )\xi +(\nabla B)^{\intercal }\xi )(\cdot ,t)=f(\cdot ,t)\xi (\cdot ,t)+O(\mathrm{dist}(\cdot ,\Sigma ^{\ast }(t))) \] \item one has \[ B(\cdot ,t)\cdot \xi (\cdot ,t)+\nabla \cdot \xi (\cdot ,t)-\lambda ^{\ast }(t)=O(\mathrm{dist}(\cdot ,\Sigma ^{\ast }(t))). \] \end{itemize} The author defines the notion of distributional solution to the volume-preserving mean curvature flow through an appropriate variational formulation and other equations or constraints. The second main result proves that any calibrated flow is unique and stable in the class of distributional solutions of the volume-preserving mean curvature flow, the author establishing an upper bound for the relative entropy defined as \[ \mathcal{E}[\chi ,\Sigma ^{\ast }](t)=\int_{\mathbb{R}^{d}\times \{t\}}(1-\nu (x,t)\cdot \xi (x,t)\left\vert \nabla \chi \right\vert \] and the volume error defined as \[ \mathcal{F}[\chi ,\Sigma ^{\ast }](t)=\int_{ \mathbb{R}^{d}}\left\vert \chi (x,t)-\chi _{\Omega ^{\ast }(t)}(x)\right\vert \left\vert \vartheta (x,t)\right\vert dx. \] To prove the existence of a gradient-flow calibration, the author builds \(\xi (x,t)=\zeta (\cdot ,t)s(x,t)\), where \(\zeta \) is a smooth cut-off function and \(s\) is the signed distance function to \(\Sigma ^{\ast }\) and \(\vartheta (x,t)=\tau (s(x,t))\), where \(\tau \) is a smooth and non-decreasing truncation of the identity. He also needs the solution to the Neumann-Laplace problem: \(\Delta \varphi =0\) in \(\Omega ^{\ast }(t)\), \(\nu ^{\ast }\cdot \nabla \varphi =V^{\ast }(\cdot ,t)\) on \(\Sigma ^{\ast }(t)\) and he defines \(B\) as the gradient of an extension of \(\varphi \). The bounds for the relative entropy and the volume error are obtained through direct computations.
0 references
volume-preserving mean curvature flow
0 references
constrained gradient flow
0 references
weak solutions
0 references
gradient-flow calibrations
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references