Banach manifolds with bounded structure and the Gauss-Ostrogradskii formula (Q2377200)

An infinite dimensional surface is a codimension 1 submanifold of a Banach manifold. The aim of this paper is to prove a formula that generalize to infinite dimensional surfaces (maybe the terminology ``hypersurfaces'' would be more precise?) the finite dimensional Green-Ostrogradskii formula \[ \int_G\mathrm{div} X = \int_{\partial G} X. n \] where \(G\) is a domain, \(\partial G\) is its boundary and \(n\) is the normal vector of the boundary. This formula has already been investigated in various infinite dimensional approaches, that the author gives in his references. But the approach given in this paper seems to be new, and is based on the following construction: given a measure \(\mu\), differentiable with respect to a vector field \(Z\), so-called ``strictly transverse'' to \(\partial G\), one can define an integral on \(\partial G\) by \[ I_Z(f) = \frac{d}{dt}|_{t = 0} \int_{\Phi_t^Z G} f d\mu, \] where \(\Phi_t^Z\) is the flow of \(Z\). This integral is shown to have good enough properties to derive a Green-Ostrogratskii formula: \[ \int_G\mathrm{div}_\mu X d\mu = \int_{\partial G} \omega(X) d\mu_\omega, \] where \(\omega\) is a 1-form that is used to make a unifying link between two integrals \(I_{Z_1}\) and \(I_{Z_2}\) generated by two different strictly transverse vector fields \(Z_1\) and \(Z_2.\) Various corollaries and reformulations are considered, finishing with a Hopf lemma on Hilbert manifolds. Technical references are dealing with two separate areas: infinite dimensional surfaces and measure theory, and infinite dimensional manifolds. The first one which is very well documented even if many references are in Russian. The technicalities of this work require that the underlying manifold, or at least a neighborhood of the domain \(G\), carries a geometric structure where one can control the Lipschitz constants of the transition maps. This requirement is a very strong restriction of the field of application of the results of the paper, but problems of this kind are very usual when one tries to extend finite dimensional results to infinite dimensional settings.