Sobolev gradients for the Möbius energy (Q2065714)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Sobolev gradients for the Möbius energy |
scientific article |
Statements
Sobolev gradients for the Möbius energy (English)
0 references
12 January 2022
0 references
The Möbius energy for an embedded curve \( \gamma \) in \( \mathbb{R}^m \) is \[ \mathcal{E} ( \gamma ) = \int_{ \mathbb{T} } \int_{ \mathbb{T} } \left( \frac 1 { | \gamma (x) - \gamma (y ) |^2 } - \frac 1 { \rho_\gamma^2 ( x,y ) } \right) | \gamma^\prime (x) | | \gamma^\prime (y) | \, dx dy , \] where \( \mathbb{T} = \mathbb{R} / \mathbb{Z} \). We denote the Fréchet derivative (gradient) at \( \gamma \) in \( L^2 \)-space by \( D \mathcal{E} ( \gamma ) \). Let \( G \) be an inner product on a function space on \( \mathbb{T} \). The \( G \)-gradient \( \left. \mathrm{grad} ( \mathcal{E} ) \right|_\gamma \) is defined as \[ G ( \left. \mathrm{grad} ( \mathcal{E} ) \right|_\gamma , w ) = D \mathcal{E} ( \gamma ) w , \] where \( w \in C^\infty ( \mathbb{T} , \mathbb{R}^m ) \). The most typical one is the \( L^2 \)-gradient which is defined using the standard \( L^2 \)-inner product: \[ G ( u,v ) = \int_{ \mathbb{T} } \langle u(x) , v(x) \rangle | \gamma^\prime (x) | \, dx . \] \textit{Z.-X. He} [Commun. Pure Appl. Math. 53, No. 4, 399--431 (2000; Zbl 1042.53043)] showed that the \( L^2 \)-gradient is a pseudo-differential operator of third order with the principal term \( ( - \Delta )^{ \frac 32 } \) modulo constant multiplication. For an explicit expression including lower order terms, see \textit{A. Ishizeki} and \textit{T. Nagasawa} [Math. Ann. 363, No. 1--2, 617--635 (2015; Zbl 1330.53005)]. \par The present authors study the \( G \)-gradient with other inner products, and show several advantages of it. Firstly, they explain their approach using the inner product \[ G( u , v ) = \int_{ \mathbb{T} } ( - \Delta )^{ 3/4 } \langle u(x) , ( - \Delta )^{ 3/4 } v(x) \rangle | \gamma^\prime (x) | \, dx \] as a good one. Then, it holds at least formally that \[ \left. \mathrm{grad} ( \mathcal{E} ) \right|_\gamma = ( - \Delta )^{ - 3/2 } D \mathcal{E} ( \gamma ) . \] Due to He [loc. cit.], this is a pseudo-differential operator of order \( 0 \). Hence the \( G \)-gradient flow is an ordinary differential equation on the Banach space to which \( \gamma \) belongs. In fact, the authors choose the \( W^{ 3/2 ,2 } \)-Galgliardo inner product as \( G \), and call the \( G \)-gradient a \textit{Sobolev gradient} \par The main result is that the Sobolev gradient above is a well-defined, locally Lipschitz continuous vector field on the configuration space. This yields the local well-posedness of the gradient flow. As a second advantage, since the gradient flow is an ordinary differential equation, one does not need the Courant-Friedrichs-Lewy condition for discretization. Furthermore, corresponding results under constraints (the projected gradient) are obtained. Though the discussion is not concise, the authors present a detailed strategy in the beginning of every section. Note that they quote all Lemmas, Propositions, and Theorems as Theorems (eg., they quote Lemma 4.2 as \textit{Theorem 4.2}).
0 references
Möbius energy
0 references
Sobolev gradient
0 references
gradient flow
0 references
discrete energy
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references
0 references