Gaussian fields and random packing (Q1871903)

A Poisson point process \((w_k)_1^\infty\) on the set \(A\times [0,\tau]\) \((A\subset \mathbb{R}^d\), \(\tau> 0)\) is considered. Let \(w_k\equiv (x_{w_k},t_{w_k})\), where \(x_{w_k}\in A\), \(t_{w_k}\in [0,\tau]\), and \(A\) have piecewise smooth boundary. The process is thinned out by the procedure of packing in \(A\): the point \(w_n\) is being deleted if there exists \(w_k\) such that \(t_{w_k}< t_{w_n}\), \(w_k\) is not deleted yet, and \(| x_{w_n}- x_{w_k}|< 2\). Let \(\pi_{\tau,A}\) be a projection on \(\mathbb{R}^d\) of such a rarefied Poisson process. The signed measure \(\nu_{\tau,A, \lambda}\) is investigated, where \[ \nu_{\tau,A,\lambda}(B)= (\pi_{\tau,\lambda A}(\lambda B)- E\pi_{\tau,\lambda A}(\lambda B))/ \lambda^{d/2} \quad (B\subset A). \] This signed measure is proved to converge in distribution to the Gaussian white noise. A similar result is obtained for the Poisson point process on \(\mathbb{R}^+\) indexed by the lattice points \(\mathbb{Z}^d\).