The Onsager-Machlup functions for Gaussian measures (Q1363273)

For a Gausian Radon measure \(\mu\) on a real locally convex space \(X\) let \(X^*_{\mu}\) be the set of all \(\mu-\)measurable linear functionals. Put also for any \(r>0\), for any \(\mu-\)measurable seminorm \(q\) and any \(f\in X^*_{\mu}\): \[ V_r:=\{x:q(x)\leq r\}, \qquad J_r(f):=\frac{1}{\mu(V_r)}\int_{V_r} \exp(f)d\mu. \] Consider now the sets: \[ E_q:=\{f\in X^*_{\mu}: \lim_{r\to 0}\quad J_r(f)\text{ exists}\}, \qquad F_q:=\{f\in E_q:\lim_{r\to 0}J_r(f)=1\}. \] The author announces several results, in which the conditions on \(q\) are imposed to guarantee the validity of the equality \(F_q=X^*_{\mu}\). E.g., it is stated that this is so, if either \(q\) is a continuous seminorm, which is a norm on the vector subspace of all admissible shifts of \(\mu\); or the set \(\{x:q(x)<1\}\) is bounded.