Laplace's method for iterated complex Brownian integrals (Q2570832): Difference between revisions

The author proves the following theorem (Laplace's method for iterated stochastic integrals): Let \(d\) denote a positive integer, \(S\) the set \(\{1,\dots,d\}\) and \(\beta^1,\dots,\beta^d\) complex independent Brownian motions, defined on the same filtered probability space. Let \(g_1,\dots,g_d: [0, 1]\to{\mathbf R}\) be continuous functions and \(r\in N^*\). For \(p\in\{1,\dots,r\}\), \(\alpha\in S^p\), let \((u_n^{p, \alpha} )_{n\in N^*}\) be a sequence of nonzero complex numbers such that \(\lim_{n\to\infty}\frac 1n\log |u_n^{p, \alpha}|= 0\), and \((\lambda_\alpha^p)_{\alpha\in S^p}\) a family of complex numbers. Assume that one of the \(\lambda^r_\alpha\)'s is nonzero and put \[ X_n= \sum^r_{p=1}\sum_{\alpha\in S^p}\lambda_\alpha^p u_n^{p,\alpha} \int^1_0 \left( \int^{t_1}_0\cdots\left(\int_0^{t_{p-1}} e^{ng_{\alpha_p}(t_p)}d \beta_{t_p}^{\alpha_p} \right)\dots e^{n g_{\alpha_2}(t_2)}d\beta_{t_2}^{\alpha_2} \right) e^{n g_{\alpha_1}(t_1)}d\beta^{\alpha_1}_{t_1}. \] Then we have, almost surely, \[ \lim_{n\to\infty}\frac 1n\log|X_n|=\max_{1\leq p\leq r} \max_{\substack{ \alpha\in S^p\\ \lambda^p_\alpha\neq 0}} \max_{\Delta_p} g_{\alpha_1}\otimes\cdots\otimes g_{\alpha_p}. \] This theorem is used to extend properties of Bougerol and Jeulin's path transform in the random case when simple representations of complex semisimple Lie algebras are not supposed to be minuscule. The author gives also the example of Lie algebra, such that its fundamental representations are not minuscule.