Accessible cylindrical measures on locally convex spaces (Q1070491): Difference between revisions

The accessibility of a cylindrical measure, which is weaker than the quasi-invariance [\textit{W. Linde}, Z. Wahrscheinlichkeitstheorie verw. Gebiete 40, 91-99 (1977; Zbl 0342.28007 and Zbl 0354.28006)], the scalar quasi-invariance [\textit{S. Chevet}, Lect. Notes Math. 644, 125-158 (1978; Zbl 0387.60005)] and the partial admissibility [\textit{Y. Takahashi}, Hokkaido Math. J. 8, 150-166 (1979; Zbl 0444.28013)], is introduced and the conditions for the exitence of an accessible cylindrical measure are investigated. Let \(\mu\) be a finitely additive cylindrical measure on a locally convex space E and \(\tau_{\mu}\) be the weakest linear topology on \(E^*\) (the topological dual of E) making the characteristic functional \(\int_{E}\exp (i<x,x^*>)d\mu (x)\) continuous. The topological dual of \((E,\tau_{\mu})\) is called the kernel of \(\mu\) and denoted by \(K_{\mu}\). \(\mu\) is said to be accessible if \(E\subset K_{\mu}\). Let \(\tau\) be another linear topology on E. \(\mu\) is called of type p \((0<p<+\infty)\) with respect to \(\tau\) if \(\tau\) is finer than \(\tau_{\mu}.\) The main result is as follows. Let E and F be locally convex spaces, T be a continuous linear mapping from F into E, \(\tau\) be a linear topology on E, \(\tau_ k\) be the Mackey topology on F, and let \(0<p<2\). Then the following two statements are equivalent. (1) E admits a T(F)-accessible cylindrical measure of type p with respect to \(\tau\). (2) There exists a probability space (\(\Omega\),\(\nu)\) such that T:\((E,\tau)\to (F,\tau_ k)\) can be factored as follows; \[ (E,\tau)\to S_ p\to S_ 0\to (F,\tau_ k), \] where \(S_ p\) is a subspace of \(L^ p(\Omega,\nu)\), \(S_ 0\) is a subspace of \(L^ 0(\Omega,\nu),S_ p=S_ 0\) and \(S_ p\to S_ 0\) is the identity.