The exponential map of the group of area-preserving diffeomorphisms of a surface with boundary (Q1661663): Difference between revisions

The authors consider the compact Riemannian manifold \(M=S^{1}\times [ 0,L]\) with boundary \(\partial M=S^{1}\times \{0\}\cup S^{1}\times \{L\}\) and the volumorphism \(\mathcal{D}_{\mu }^{s}(M)\) of diffeomorphisms of \(M\) which preserve the Riemannian volume \(\mu \) and are of Sobolev class \(H^{s}\). They define the exponential map of the \(L^{2}\) metric given by \(\left\langle u,v\right\rangle _{L^{2}}=\int_{M}\left\langle u(x),v(x)\right\rangle d\mu (x)\) on \(\mathcal{D}_{\mu }^{s}(M)\) from \(T_{e}\mathcal{D}_{\mu }^{s}\) to \( \mathcal{D}_{\mu }^{s}\) through \(\exp _{e}(tv_{0})=\eta (t)\), where \(\eta \) is the unique geodesic issuing from the identity with initial velocity \( v_{0}\in T_{e}\mathcal{D}_{\mu }^{s}\). This is linked to the resolution of the incompressible Euler equations \(\partial _{t}v+\nabla _{\nu } v=- \operatorname{grad} p\), \(\operatorname{div}\;v=0\), \(\left\langle v,\nu \right\rangle =0\) on \(\partial M\), the solution starting from \(v_{0}\) at \(t=0\). The main result of the paper proves that the exponential map is a nonlinear Fredholm map of index 0. The authors first recall the definition of the Jacobi field solution operator \(\phi _{t}w_{0}=d\exp _{e}(tv_{0})tw_{0}\), that of the group adjoint operator \(Ad_{\eta }=dR_{\eta ^{-1}}dL_{\eta }\) where \(R_{\eta }\) and \(L_{\eta }\) are the right and left translations by \(\eta \). They introduce the operators \(\Omega _{t}=\int_{0}^{t}Ad_{\eta (\tau )^{-1}}Ad_{\eta (\tau )^{-1}}^{\ast }d\tau \) and \(\Gamma _{t}=\int_{0}^{t}Ad_{\eta (\tau )^{-1}}K_{v(\tau )}dR_{\eta (\tau )^{-1}}\phi (\tau )d\tau \) where \(K_{v}\) is a compact operator on \(T_{e} \mathcal{D}_{\mu }^{s}\) associated to the solution of the incompressible Euler equations. The authors prove that \(\phi _{t}\) is a family of bounded operators from \(T_{e}\mathcal{D}_{\mu }^{s}\) to \(T_{\eta (t)}\mathcal{D} _{\mu }^{s}\) and they recall properties of Fredholm mappings. The key tools of the proof of the main result are estimates on the invertible part of \( d\exp _{e}\), starting with a coercivity property of \(\Omega _{t}\).