Interpolation of Gibbs measures with White Noise for Hamiltonian PDE

Abstract

We consider the family of interpolation measures of Gibbs measures and white noise given by $$dQ_{0,\b}^{(p)} = Z_\b^{-1} \ind_{{\int_{\T} u^2\le K\b^{-1/2}\}} e^{-\int_{\T} u^2 +\b \int u^p} dP_{0,\b}$$ where $P_{0, \b}$ is the Wiener measure on the circle, with variance $\beta^{-1}$, conditioned to have mean zero. It is shown that as $\beta\to 0$, $Q_0^\beta$ converges weakly to mean zero Gaussian white noise $Q_0$. As an application, we present a straightforward proof that $Q_0$ is invariant for the Kortweg-de Vries equation (KdV). This weak convergence also shows that the white noise is a weak limit of invariant measures for the modified KdV and the cubic nonlinear Schr\"odinger equations.