Renormalization and convergence in law for the derivative of intersection local time in \(\mathbf R^2\) (Q947151)

Let \(f_{\varepsilon }\) be the Gaussian density function on \(\mathbb{R}^{2}\) with null mean and variance \(\varepsilon\), \(f'_{\varepsilon }\) its derivative with respect to the first variable, \((B_{t})_{t\geq 0}\) the brownian on \(R^{2}\), \(\alpha '_{\varepsilon }(T)= \int_{_{0}}^{^{T}}\int_{_{0}}^{^{t}}f'_{\varepsilon }(B_{t}\)-\(B_{s})dsdt\). Most of the paper is devoted to the proof of the convergence in law of \(((\log (1/\varepsilon ))^{-1}\alpha '_{\varepsilon }(T))_{T\geq 0}\) to \((5^{1/2}(8\pi )^{-1}2^{-1/4}W_{T})_{T\geq 0} \), \((W_{T})_{T\geq 0}\) being the Brownian motion on \(\mathbb{R}\). In the last part a similar proof is presented for an analogous result, in which \(B_{t}\) is replaced by a symmetric stable process with index \(\beta \in (1,2)\), \((\log(1/\varepsilon ))^{-1}\) by \(\varepsilon^{3/\beta -3/2}\) and \(W_{T}\) appears with other, explicitly expressed, coefficient \(c(\beta )\). The following, from the paragraph ``outline of the proof'', gives an idea about the first. To deal with the integral which expresses \(E (\alpha '_{\varepsilon }(T)^{n})\), we would like to factor the expectation in the integrand using the independence of the Brownian increments. This factoring will depend on the ordering of the \(s_{j}\)'s and \(t_{j}\)'s and we will therefore split the domain \(D_{T}^{n}\) (where \(D_{T}= \{(s,t);0\leq s\leq t\leq T\})\) into many regions, each corresponding to that ordering. We then proceed separately over each region. The history of the problem is presented in the introduction.