Large and moderate deviations for intersection local times (Q1424401): Difference between revisions

The authors study the large and moderate deviations for the intersection local times generated by local times of independent Brownian motions and symmetric random walks. Let \(W_1, \ldots, W_m\) be independent 1-dimensional (standard) Brownian motions with local times \(L_1(t, x), \ldots, L_m(t, x)\) \((t \geq 0, x \in {\mathbb R})\), respectively. Their Theorem 1.1 shows that \[ \lim_{\lambda \to \infty} \lambda^{-2/(mp-1)}\, \log \text{ P} \Big\{\int_{-\infty}^\infty \prod_{j=1}^m L_j^p(1, x)\, dx \geq \lambda\Big\} = - C_2(m, p), \tag{1} \] where \(0 < C_2(m, p) < \infty\) is an explicit constant. This generalizes the large deviation principle of \textit{U. Mansmann} [Stochastics Stochastics Rep. 34, 93--125 (1991; Zbl 0726.60021)] for the \(L_2\)-norm of the Brownian local time; and coincides with a result of \textit{M. Csörgő}, \textit{Z. Shi} and \textit{M. Yor} [Bernoulli 5, 1035--1058 (1999; Zbl 0960.60023)] for the self-intersection local times of the Brownian bridges. The key idea of proving (1) is to embed the Brownian local times into the Banach space \({\mathcal L}^p({\mathbb R})\) and to use a Feynman-Kac type large deviation result for the Brownian occupation times. Consider the random walk \(\{S(n)\}_{n \geq 1}\), where \(S(n) = \sum_{k=1}^n X_k\) and \(\{X_k\}\) is a sequence of i.i.d. integer-valued symmetric random variables with \(\sigma^2 = \text{ E}(X_1^2) < \infty\). Define the local time \(\ell (n, x)\) of \(\{S(n)\}_{n \geq 1}\) by \(\ell (n, x) = \sum_{k=1}^n I_{\{S(k) = x\}}. \) Let \(\{S_1(n)\}_{n \geq 1}, \ldots, \{S_m(n)\}_{n \geq 1}\) be \(m\) independent copies of \(\{S(n)\}_{n \geq 1}\) with local times \(\ell_1(n, x), \ldots,\) \(\ell_m(n, x)\). The authors prove that for any integer \(m \geq 1\) and \(p > 0\) such that \(mp>1\), \[ n^{-(mp+1)/2}\sum_{x \in {\mathbb Z}} \prod_{j=1}^m \ell_j^p(n, x)\;{\rightarrow}^d \;\sigma^{(mp-1)} \int_{-\infty}^\infty \prod_{j=1}^m L_j^p(1, x) \,dx \tag{2} \] and for any sequence \(\{b_n\}\) satisfying \(b_n \to \infty\) and \(b_n/n \to 0\), \[ \lim_{n \to \infty} {1 \over {b_n}} \log \text{ P} \Big\{ \sum_{x \in {\mathbb Z}} \prod_{j=1}^m \ell_j^p(n, x)\geq n^{(mp+1)/2} b_n^{(mp-1)/2} \Big\} = - \sigma^2 C_2(m, p). \tag{3} \] As applications of (1) and (3), the authors establish laws of the iterated logarithm for \(\int_{-\infty}^\infty \prod_{j=1}^m L_j^p(1, x) \,dx\) and \(\sum_{x \in {\mathbb Z}} \prod_{j=1}^m \ell_j^p(n, x)\).