Macroscopic analysis of determinantal random balls (Q1740536)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Macroscopic analysis of determinantal random balls |
scientific article |
Statements
Macroscopic analysis of determinantal random balls (English)
0 references
30 April 2019
0 references
The authors study the behavior of the collection \(\mathcal B_\rho\) of random balls depending on the parameter \(\rho>0\) asymptotically as \(\rho\to0\). The radii of the balls are multiplied by \(\rho\) i.i.d. random variables with a common probability density function \(f\) satisfying \(f(r)\sim \frac{C_\beta}{r^{\beta+1}}\) as \(r\to\infty\) for some \(\beta\in(d,2d)\). The centers of the balls are generated by a stationary determinantal point process on \(\mathbb R^d\) with intensity \(\lambda_\rho\) tending to infinity as \(\rho\to0\). It is assumed that the correlation operator of the process is bounded with spectrum included in \([0,1]\) and also is locally of trace class. Denote by \(\mathcal M_\beta\) a set of finite compactly supported Borel measures \(\mu\) on \(\mathbb R^d\) for which there exist \(p\in (0,\beta), q\in(\beta,2d]\), and \(C_\mu>0\) such that for all \(r>0\) we have \[ \int_{\mathbb R^d}\mu(B(x,r))^2 \,\mathrm dx\leq C_\mu\cdot\min(r^p,r^q),\quad\text{where}\quad B(x,r)=\{y\in\mathbb R^d:|x-y|\leq r\}. \] In particular, \(\mathcal M_\beta\) contains the set \(\mathcal M'\) of all measures having compactly supported densities from \(L^2(\mathbb R^d)\). Consider a measure-indexed random field \(M_\rho(\cdot)\) defined as the compensated total measure of all balls from \(\mathcal B_\rho\): \[ M_\rho(\mu) = \sum_{B\in \mathcal B_\rho} (\mu(B)-\mathbb E\,\mu(B)),\quad \mu\in\mathcal M_\beta. \] Depending on how \(\lambda_\rho\) and \(\rho^\beta\) are related asymptotically as \(\rho\to0\), the following three limit regimes for \(M_\rho(\cdot)\) appear. Large-grain scaling: If \(\lim_{\rho\to0}\lambda_\rho\rho^\beta=\infty\), then \(\frac{M_\rho(\cdot)}{\sqrt{\lambda_\rho\rho^\beta}}\) converges as \(\rho\to0\) in \(\mathcal M_\beta\) in the finite-dimensional distributions sense to \(X(\cdot)\) defined by \[ X(\mu)=\int_{\mathbb R^d\times\mathbb R_+}\mu(B(x,r))\,\mathrm d W,\quad \mu\in\mathcal M_\beta, \] where \(W\) is the Gaussian white noise on \(\mathbb R^d\times\mathbb R_+\) with control measure \(\frac{C_\beta}{r^{\beta+1}}\,\mathrm{d}x\,\mathrm{d}r\). Intermediate scaling: If \(\lim_{\rho\to0}\lambda_\rho\rho^\beta=a^{d-\beta}\in(0,\infty)\), then \(M_\rho(\cdot)\) converges as \(\rho\to0\) in \(\mathcal M_\beta\) in the finite-dimensional distributions sense to \(X(\cdot)\) defined by \[ X(\mu)=\int_{\mathbb R^d\times\mathbb R_+}\mu(a^{-1}B(x,r))\,\mathrm d \Pi,\quad \mu\in\mathcal M_\beta, \] where \(\Pi\) is the compensated Poisson random measure on \(\mathbb R^d\times\mathbb R_+\) with intensity measure \(\frac{C_\beta}{r^{\beta+1}}\,\mathrm{d}x\,\mathrm{d}r\). Small-grain scaling: Let \(\gamma=\frac{\beta}{d}\). If \(\lim_{\rho\to0}\lambda_\rho\rho^\beta=0\), then \(\frac{M_\rho(\cdot)}{(\lambda_\rho\rho^\beta)^{1/\gamma}}\) converges as \(\rho\to0\) in \(\mathcal M'\) in the finite-dimensional distributions sense to \(X(\cdot)\) defined by \[ X(\mu)=\int_{\mathbb R^d}\varphi(x)\,\mathrm d \Lambda,\quad \mu\in\mathcal M', \] where \(\varphi\) is the density of \(\mu\) and \(\Lambda\) is the independently scattered \(\gamma\)-stable random measure on \(\mathbb R^d\) with unit skewness and control measure \(\sigma_\gamma\,\mathrm d x\), where \[ \sigma_\gamma = C_\beta\frac{\mathrm{vol} (B(0,1))^\gamma}{d}\int_0^{\infty}\frac{1-\cos r}{r^{1+\gamma}}\,\mathrm d r. \]
0 references
determinantal point processes
0 references
generalized random fields
0 references
limit theorem
0 references
point processes
0 references
stable fields
0 references
0 references
0 references