No label defined (Q4035826)

[For the entire collection see Zbl 0752.00055.] This paper looks at weak convergence of evolving random fields: \(S_ n(x,t)=\sum_{i\leq\tau(n,t)}Y_{n,i}(x)\) for \(x\) in an index set \(X\), \(t\in[0,1]\), and \(\tau(n,t)\) perhaps random. The \(Y_{n,i}\) are dependent random variables of a semimartingale structure. For fixed \(T\), \(S_ n(x,T)\) defines a sequence of random fields on \(X\), but here the authors also allow \(t\) to vary. This models the evolution of these fields in time. The \(S_ n\) can also be viewed as a multi-parameter indexed process, but here there is a natural partial ordering on \(X\times[0,1]\) induced by time. The authors provide some practical examples and draw analogies with empirical processes (e.g. put \(Y_{n,i}=n^{-1/2}(I[X_ i\in x]-P[X_ i\in x])\) for \(x\in{\mathcal C}\subset{\mathcal B})\). The authors employ results on majorizing measures and metric entropy with bracketing to prove tightness (Theorem 2.1) and consequently weak convergence to an appropriate Gaussian field (Corollaries 2.1 to 2.6). The topology is stronger than that usually encountered for superprocesses. In addition to the semimartingale structure of the \(Y_{n,i}\), this paper also assumes uniform conditional asymptotic negligibility and centering, and a conditional bound on the weak-\(L_ 2\) norm over a collection of brackets.