A central limit theorem for the renormalized self-intersection local time of a stationary vector Gaussian process (Q1184079)

For a fixed \(m\geq 1\), let \(X_ 1(t),\dots,X_ m(t)\) be independent copies of a real, measurable stationary Gaussian process \(X(t)\), \(t\geq 0\), with \(EX(t)=0\), \(EX^ 2(t)=1\) and continuous covariance function \(r(t)=EX(0)X(t)\). For \(\varepsilon>0\), \(t>0\), define \[ Y(t;\varepsilon)=\varepsilon^{-m}\int_ 0^ t \int_ 0^ t \prod_{j=1}^ m \varphi(\varepsilon^{-1}(X_ j(s)-X_ j(s')))ds ds', \] where \(\varphi\) represents the standard normal density function; \(Y(t;\varepsilon)\) is an approximate self-intersection local time of \(X(s)\) on \([0,t]\). The main result shows \(Y(t;\varepsilon)\), suitably normed, has a limiting normal distribution with mean zero and variance \(8(2\pi)^{-m}\int_ 0^ \infty[(4-r^ 2(s))^{-m/2}-2^{-m}]ds\), provided only that \(r(t)\) is square-integrable and that \(t\to\infty\) and \(\varepsilon=\varepsilon(t)\to 0\) in such a way that \(t^{-1/2}\int_ 0^ 1(\varepsilon^ 2/2+1-r(s))^{-m/2}ds\to 0\).