Probabilistic estimates for the two-dimensional stochastic Navier-Stokes equations (Q1592972): Difference between revisions

A stochastic Navier-Stokes equation \[ dU + \bigl((U\cdot\nabla)U - \nu \nabla^2 U + \nabla p\bigr) dt = dF, \quad \nabla\cdot U = 0, \] on a two-dimensional torus \(\mathbb T_{L} \equiv (\mathbb R/2\pi L\mathbb Z)^2\) is studied. It is supposed that \(F\) is a Wiener process with covariance \(\mathbf E F_\alpha(s,x)F_\beta(t,y) = (s\land t)C_{\alpha\beta}(L^{-1}(x-y))\), \(C_{\alpha\beta}\) being a smooth function on \(\mathbb T_1\) such that \(\partial_\alpha C_{\alpha\beta} = 0\). Assume further that \(\int_{\mathbb T_{L}} U(0,x) dx = \int_{\mathbb T_{L}} F(t,x) dx = 0\). Define a function \(\omega\) on \(\mathbb R_{+}\times\mathbb T_1\) by \(\Omega(t,x) = \nu L^{-2}\omega(\nu L^{-2}t,L^{-1}x)\), where \(\Omega\) is the vorticity, \(\Omega = \partial_1 U_2 - \partial_2 U_1\). Denote by \(\omega_{k}\) the Fourier coefficients of \(\omega\), \(\omega_{k}(t) = (2\pi)^{-2}\int _{\mathbb T_1} e^{\roman ik\cdot x}\omega(t,x) dx\), \(k\in\mathbb Z^2\), and by \(\Phi\) the enstrophy, \(\Phi = \frac 12 \sum_{k} |\omega_{k}|^2\). Fix \(r>1\), \(\alpha >1+r\) and for any \(D>0\) define a norm \(\|\omega\|_{D} = \sup_{k} |\omega_{k}||k|^{r}\exp( D^{-\alpha}|k|)\). Suppose that \(\|\omega(0)\|_{D(0)} \leq D(0)^\alpha<\infty\) and \(\Phi(0) = K<\infty\). Then it is shown that there exists a random function \(D\), \(D(t)<\infty\) for all \(t\), such that \(\|\omega(t)\|_{D(t)} < D(t)^\alpha\) almost surely. Moreover, \[ \mathbf P\bigl\{ \|\omega(t)\|_{D} \leq D^\alpha, \;\Phi(t)\leq D^2\bigl\} \geq 1 - C\exp(-cR^{-1}D^2), \] whenever \(t> C(\log D(0) + \log K)\) and \(D>CR\log R\) with \(R>1\). This theorem implies, in particular, that \(\omega(t,\cdot)\) is an analytic function for every \(t\) almost surely.