An inequality for Fourier-Laplace transforms of entire functions, and the existence of exponential frames in Fock space (Q1370394): Difference between revisions

Let \(\phi\) denote a convex function on \(\mathbb{C}^n\) and put \(d\mu= c_ne^{-\phi}d\lambda\), where \(\lambda\) is the Lebesgue measure and the constant \(c_n\) is chosen so as to give \(\mu\) mass \(1\). The author considers the generalized Fock spaces \[ F^2_\phi= \{f\in H(\mathbb{C}^n);\;\int| f|^2d\mu< \infty\} \] and variants \(A^2_\phi\) in which the measure \(d\mu\) above is replaced by \[ d\mu= c_ne^{-\phi}\Biggl({i\over 2}\partial\overline\partial \phi\Biggr)^n/n!; \] i.e., the Lebesgue measure is replaced by a multiple of the Monge-Ampère measure \((i\partial\overline\partial \phi)^n\). He assumes that the real Hessian of \(\phi\) is uniformly bounded from above and below and denotes by \(\phi^*\) the Legendre transform of \(\phi\), \(\phi^*(z)= \sup_\zeta(2\text{ Re}(\zeta\cdot\overline z)- \phi(\zeta))\). It is not too hard to show that the Fourier-Laplace transform satisfies an inequality \[ \|\widehat f\|_{F^2_{\phi^*}}\leq A\| f\|_{F^2_\phi}, \] where the constant \(A\) depends on the bounds of the Hessian, and that a similar inequality holds for \(A^2_\phi\). The main theorem, however, asserts that if the Fourier-Laplace transform of \(f\in A^2_\phi\) is defined by \[ \widehat f(z)= c_n \int f(\zeta) e^{z\cdot\overline\zeta} e^{-\phi} \Biggl({i\over 2}\partial\overline\partial \phi\Biggr)^n/n!, \] then the following inequality in the converse direction holds: \[ \int| f|^2 e^{-\phi}\Biggl({i\over 2}\partial\overline\partial \phi\Biggr)^n/n!\leq \int|\widehat f|^2 e^{-\phi^*} \Biggl({i\over 2}\partial\overline\partial \phi^*\Biggr)^n/n!. \] In the last section of the article, the author proves that the problem of the existence of frames consisting of exponentials in the spaces \(A^2_\phi\) and \(F^2_\phi\) is equivalent to the sampling problem in \(A^2_{\phi^*}\) and \(F^2_{\phi^*}\), which is by now quite well understood in the one-dimensional case, due to work of Seip, Seip-Wallsten, Lyubarskii-Seip, and Berndtsson-Ortega.