On positive generalized functionals (Q1090894)

Given a Gaussian space (\({\mathcal N}^*,{\mathcal B},d\mu)\), where \({\mathcal N}^*\) is the dual of a real nuclear space \({\mathcal N}\), B its topological \(\sigma\)-algebra and \(d\mu\) a Gaussian measure on \({\mathcal B}\), we construct spaces of test functionals on \({\mathcal N}^*\) as follows. \(L^ 2({\mathcal N}^*,d\mu)\) is isometric to the symmetric Fock space \(\Gamma\) (\({\mathcal H})\) over the Hilbert space \({\mathcal H}\), whose scalar product determines \(d\mu\) [\textit{E. Nelson}: ''Probability theory and Euclidean field theory'', in Lect. Notes Phys. 25, 94-124 (1973; Zbl 0367.60108)]. If A is a linear (selfadjoint positive) operator on \({\mathcal H}\), one can construct from it a linear (selfadjoint pos.) operator \(d\Gamma\) (A) on \(\Gamma\) (\({\mathcal H})\) [loc. cit.] and therefore one obtains a corresponding operator \(d\gamma\) (A) on \(L^ 2({\mathcal N}^*,d\mu)\). If A is densely defined, then so is \(d\gamma\) (A) on the algebra \({\mathcal A}\) generated by cylindrical polynomials and exponentials on \({\mathcal N}^*\). Define a compatible system of norms on \({\mathcal A}\) by setting (A\(\geq 0)\) \({}_ A\| \cdot \|_{p,k}:=\| (1+d\gamma (A))^{k/2}\cdot \|_ p\), where \(\| \cdot \|_ p\) is the norm of \(L^ p({\mathcal N}^*,d\mu)\). The completion of A w.r.t. \({}_ A\| \cdot \|_{p,k}\) is denoted by \(L_ A^{p,k}\) and the projective limit of the resulting family of spaces is denoted by \(\Sigma_ A:\Sigma_ A:=_{p\in {\mathbb{N}},}-\lim_{k\in {\mathbb{N}}_ o}L_ A^{p,k}\Sigma\) \({}_ A\) serves as a space of test functionals on \({\mathcal N}^*\), \(\Sigma^*_ A\), its dual, is a space of generalized functionals. Let \(\Sigma_{A,+}\) denote the cone of a.e. positive test functionals and \(\Sigma^*_{A,+}\) the cone of those elements in \(\Sigma^*_ A\), which map \(\Sigma_{A,+}\) into \({\mathbb{R}}_+\). In this article two theorems on \(\Sigma^*_{A,+}\) analogous to classical results of Bochner are proven. Theorem A: Let \(F\in \Sigma^*_{A,+}\). Then there exists a unique positive measure \(d\nu_ F\) on \({\mathcal B}\), s.t. for all \(\phi\in {\mathcal A}\), \(<F,\phi >=\int_{{\mathcal N}^*}\phi (x)d\nu_ F(x)\) Theorem B: Choose \(A=id\) and let \(F\in \Sigma^*_{id}\). Then \(F\in \Sigma^*_{id,+}\) if and only if its Fourier transform \({\mathcal T}F\), \({\mathcal T}F(\xi):=<F(\cdot),\exp\) \(i<\cdot,\xi >>\), \(\xi\in {\mathcal N}\), is a positive definite functional on \({\mathcal N}\).