Exact small ball constants for some Gaussian processes under the \(L^2\)-norm (Q2567779)

The authors examine the behavior of the probabilities \(P(\|X\|_2<\varepsilon)\) as \(\varepsilon\to 0\) (the so-called ``small-ball deviations'') for some type of Gaussian processes \(X\) defined on \([0; 1].\) The technique applied is based on the Kac-Siegert expansion and the evaluation of the eigenvalues of a related integral equation. The processes analyzed are the centered Wiener process and the centered Brownian bridge, as well as their integrated versions. In all these cases the constants figuring in the expansion of the small ball probabilities are evaluated (sometimes the logarithmic and, in some cases, the exact constants). In particular for the integrated Brownian bridge \(\overline{B}(t) = \int_0^t B(s)ds\) the authors obtain the following result \[ P(\| \overline{B}\| _2\leq\varepsilon)\sim{{8}\over{\sqrt{3\pi}}}e^{-{{3}\over{8}}\varepsilon^{-2/3}}, \] while, for the integrated Brownian motion \(\overline{W}(t) = \int_0^t W(s)ds,\) the corresponding asymptotic reads \[ P(\| \overline{W}\| _2\leq\varepsilon)\sim{{2^{5/2}}\over{\sqrt{3\pi}}\varepsilon^{1/3}}e^{-{{3}\over{8}}\varepsilon^{-2/3}}. \] Much effort is devoted to work out the explicit or approximate expression of the eigenvalues for the different cases examined.