Spectral conditions for strong local nondeterminism and exact Hausdorff measure of ranges of Gaussian random fields (Q421200): Difference between revisions

Let \(X(t)=\{X_1(t), \dotsc, X_d(t)\}\), \(t\in\mathbb{R}^N\), be a Gaussian random field with values in \(\mathbb{R}^d\), where \(X_i\), \(1\leq i \leq d\), are independent copies a real-valued, centered, anisotropic Gaussian random field \(X_0(t)\) with stationary increments and the property of strong local nondeterminism. The aim of this paper is to consider the exact Hausdorff measure of the range of \(X([0,1]^N)\). The random field \(X_0\) is an isotropic in time-variable and satisfies the following conditions. There exists a constant vector \(H=(H_1,\dotsc, H_N)\in (0,1)^N\) such that (C1) there exists a constant \(C_{1,1}\geq 1\) such that \[ c_{1,1}\rho^2(s,t)\leq \mathrm{E}(X_0(s)-X_0(t))^2\leq c_{1,1}\rho^2(s-t) \] for all \(s,t\in [0,1]^N\), where \[ \rho(s,t)=\sum\limits_{j=1}^N|s_j-t_j|^{H_j} \quad \forall s, t \in\mathbb{R}^N; \] (C2) there exists a positive constant \(c_{1,2}\) such that, for all integers \(n\geq 1\) and all \(u,t^1, \dotsc, t^N\in [0,1]^N\), \[ \operatorname{Var}(X_0(u)|X_0(t^1), \dotsc, X_0(t^n))\geq c_{1,2} \min\limits_{0\leq k\leq n}\rho^2(u, t^k), \quad t^0=0. \] Condition (C1) implies that a random field \(X_0\) has continuous sample paths. If condition (C2) is satisfied, it is said that a random field \(X_0\) has the property of strong local nondeterminism in a metric \(\rho\) on \([0,1]^N\). The main statement of the paper is Theorem 1.1. Namely, if \(d>\sum\limits_{j=1}^NH^{-1}_j\), then conditions (C1) and (C2) imply \[ 0<\phi_1-m(X([0,1]^N))<\infty \quad \text{a.s.} \] where the function \(\phi_1(r)=r^{\sum_{j=1}^N H_j^{-1}}\log\log\frac{1}{r}\) and \(\phi_1-m\) is the corresponding Hausdorff measure. The paper is organized as follows. The Introduction contains the statement of the problem and short remarks on related results for Gaussian fields with stationary increments. The main statement of Section 2 is a general sufficient condition for a Gaussian random field with stationary increments to satisfy conditions (C1) and (C2). This condition is given in terms of spectral measures for Gaussian random fields which may contain either an absolutely continuous or a discrete part. It provides a way to construct a large class of such Gaussian fields. Several examples of such random fields are mentioned. The exact Hausdorff measure function for the range of \(X([0,1]^N)\) is considered in Section 3. The definition of Hausdorff measure, its basic properties and two inequalities for large and small tails of suprema of Gaussian random processes are recalled. General estimates for supremum of Gaussian random fields \(X\) are proved. The Talagrand's result for Hausdorff measure on trajectories of a multiparameter fractional Brownian motion (see [\textit{M. Talagrand}, Ann. Probab. 23, No. 2, 767--775 (1995; Zbl 0830.60034)]) is extended to anisotropic random fields and applied to obtain the upper bound in Theorem 1.1. The authors prove the law of iterated logarithm for the sojourn time of \(X\) and derive the lower bound in Theorem 1.1.