The palm measure and the Voronoi tessellation for the Ginibre process (Q2268723)

The Ginibre process is a determinantal process \(\phi\subset \mathbb R^2\), both isotropic and ergodic with respect to the translations of the plane \(\mathbb R^2={\mathcal C}\), with the kernel \[ K(z_1, z_2) =(1/\pi) e^{z_1\,\overline{z}_2}\exp\left[-(1/2)(|z_1|^2 +|z_2|^2)\right], \quad (z_1, z_2)\in {\mathcal C}^2. \] It is also pertinent to consider the full class of determinantal processes \(\phi^{*\alpha}\) related to the kernels, \(K_{*\alpha}(z_1, z_2)= (1/\pi) e^{(1/\alpha) z_1\overline{z}_2} \exp[-(1/(2\alpha))(|z_1|^2 +|z_2|^2)]\), \((z_1, z_2)\in {\mathcal C}^2\), with \(\alpha\in (0,1)\). The process \(\phi^{*\alpha}\) can be obtain by deleting, independently and with probability \(1-\alpha\), each point of the Ginibre process \(\phi\) and then applying the homothety of ratio \(\sqrt{\alpha}\) to the remaining points in order to restore the intensity of the process \(\phi\). Besides, the process \(\phi^{*\alpha}\) converges in law when \(\alpha\to 0\) to the Poisson process. In other words, the process \(\phi^{*\alpha}\) constitute an intermediate class between a Poisson process and a Ginibre process. For every determinantal process \(\Psi\), a result obtained in [see \textit{T. Shirai} and \textit{Y. Takahashi}, J. Funct. Anal. 205, No.~2, 414--463 (2003; Zbl 1051.60052)] states that \(\Psi_0=(\Psi|0\in \Psi)\setminus \{0\}\) is determinantal as well. It follows that in the Ginibre case the process \(\phi_0\) (\(\phi_0\) is nonstationary) is determinantal with the kernel \(K(z_1, z_2) =(1/\pi) (e^{z_1\overline{z}_2}-1) \exp[-(1/2) (|z_1|^2 +|z_2|^2)]\). The first main result of the paper is (Theorem 1) that \(\phi_0\) can be obtained from \(\phi\), simply by deleting one Gaussian-distributed point \(Z\). Let \(\mathbb E\) be a local compact Hausdorff space with a countable basis and \(\lambda\) be a reference Radon measure in \(\mathbb E\). Further, let \(\Psi\subset\mathbb E\) be a stationary determinantal process with kernel \(K\) defined on \(\mathbb E^2\). The author introduce the following Condition. For every bounded Borel set \(A\subset \mathbb E\), all the eigenvalues of the operator \(K_A\), acting on \(L^2(A, \lambda)\), lie in the interval \([0, 1)\). Theorem 2. Let \(\lambda\) has full support, \(\Psi\subset\mathbb E\) be a stationary determinantal process with continuous kernel \(K\) on \(\mathbb E^2\) and \(K\) satisfied the condition. Then the process \(\Psi_0\) is stochastically dominated by the process \(\Psi\). Theorem 3. Let \(\lambda\) has full support. Consider two kernels \(K\) and \(L\) satisfying the Condition above. Denote by \(\Psi\subset\mathbb E\) the determinantal process associated to the kernel \(K\) and by \(\varphi\subset\mathbb E\) the process associated to the kernel \(L\). Suppose that \(K\geq L\) in the Loewner order. Then the process \(\Psi\) dominates stochastically the process \(\varphi\). (Recall that \(K\geq L\) in the Loewner order if \(K-L\) is a positive semidefine operator). Obviously, Theorem 3 implies Theorem 2. The paper is organized as follows: Section 2 contains the necessary background. In Section 3 the author proves Theorem 3. Theorem 1 is proved in Section 4 as a consequence of Theorem 2. Unfortunately, the correlation between the process \(\phi_0\) and the random point \(Z\) is still unknown. Finally, the author mentions that it is easy to deduce the corresponding results for the Ginibre process \(\phi\) for the processes \(\phi^{*\alpha}\) with \(\alpha\in (0,1)\).