A strong invariance principle for nonconventional sums (Q1939555)

This is a continuation of the author's earlier work with S. R. S. Vardhan [``Nonconventional limit theorems in discrete and continuous time via martingales'', preprint, \url{arXiv:1012.2223}] and deals with the strong invariance principle using martingales. In [loc. cit.], they obtained a functional limit theorem for the sum \[ \xi_N(t)=N^{-\frac{1}{2}} \sum_{1 \leq n \leq Nt} \big(F(X(q_1(n),\dots,X(q_{\ell}(n)))-\overline{F}\big), \] and its continuous analogue \[ \xi_N(t)=N^{-\frac{1}{2}} \int_{0}^{Nt} \big(F(X(q_1(s)),\dots,X(q_{\ell}(s)))-\overline{F} \big)\, ds, \] where \( X(n), n \geq 0 \) is a suitable fast mixing vector-valued process, \(F\) is a suitable continuous function with polynomial growth and certain regularity properties, \(\overline{F}\) is an integrated average of \(F\) with respect to a suitable measure of the initial process \(X(0)\), and the \(q_j\) denote positive functions with suitable growth conditions. In this paper, the author shows that the sum \(\mathcal{Z}(N t)=\sqrt{N}\, \xi_N(t) \) can be represented as \(\sum_{1 \leq i \leq \ell} \mathcal{Z}_i(N t) \), where each \(\mathcal{Z}_i(N t)\) can be closely approximated by a process \( \sigma_i B_i(t)\), \(\sigma_i\geq 0\) (\( B_i(t) \) being a Brownian motion). This leads to an invariance principle for an law-of-the-iterated-logarithm-type result for the sum \(\xi_N(t)\). Several applications of the result are pointed out for stochastic processes, dynamical systems, etc.