Optimal integrability threshold for Gibbs measures associated with focusing NLS on the torus (Q2118060): Difference between revisions

Consider random fields \[ u_0(\omega,x)=\sum_{n\ne 0}\frac{g_n(\omega)}{2\pi n}e^{2\pi i nx},\qquad u_1(\omega,y,z)=\sum_{n=1}^\infty\frac{g_n(\omega)}{z_n}e_n(r),\quad r^2=y^2+z^2 \] for \(x\) in the one-dimensional torus \(\mathbb T\) and \((y,z)\) in the open unit ball \(\mathbb D\) in \(\mathbb R^2\) where \(\{z_n\}\) are the successive positive zeros of the zero-order Bessel function \[ J_0(v)=\sum_{j=0}^\infty\frac{(-1)^j}{(j!)^2}\left(\frac v2\right)^j, \] let \(\{e_n\}\) be the orthonormal system in \(L^2_{\operatorname{rad}}(\mathbb D)\) defined via \[ e_n(r)=\|J_0(z_n\cdot)\|^{-1}_{L^2(\mathbb D)}J_0(z_n r),\qquad 0\le r\le 1, \] assume that \(\{\operatorname {Re}\,g_n,\operatorname {Im}\,g_n\}_n\) are independent centered Gaussian random variables with mean \(1/2\) and define \[ Z_{p,K,i}=\mathbb E\left[e^{\frac 1p\|u_i\|_{L^p(\mathbb O_i)}^p}\mathbf 1_{[\|u_i\|_{L^2(\mathbb O_i)}\le K]}\right] \] for \(\mathbb O_0=\mathbb T\) and \(\mathbb O_1=\mathbb D\). It has been known that \(Z_{p,K,0}<\infty\) for every \(2<p<6\) and every \(K>0\), and that \(Z_{6,K,0}<\infty\) for small positive \(K\). The authors prove that \(Z_{6,K,0}<\infty\) for \(K\le\|Q\|_{L^2(\mathbb R)}\) and \(Z_{6,K,0}=\infty\) for \(K>\|Q\|_{L^2(\mathbb R)}\) where \(Q\) is the optimizer for the Gagliardo-Nirenberg-Sobolev inequality on \(\mathbb R\) such that \(\|Q\|^6_{L^6(\mathbb R)}=3\|Q^\prime\|^2_{L^2(\mathbb R)}\). In the radial \(\mathbb R^2\) case, it has been known that \(Z_{p,K,1}<\infty\) for every \(2<p<4\) and every \(K>0\), and \(Z_{4,K,1}<\infty\) for small positive \(K\) and \(Z_{4,K,1}=\infty\) for \(K>0\) near \(+\infty\). The authors prove that \(Z_{4,K,1}<\infty\) for \(K<\|Q\|_{L^2(\mathbb R^2)}\) and \(Z_{4,K,1}=\infty\) for \(K>\|Q\|_{L^2(\mathbb R^2)}\) where \(Q\) is the optimizer for the Gagliardo-Nirenberg-Sobolev inequality on \(\mathbb R^2\) such that \(\|Q\|^4_{L^4(\mathbb R^2)}=2\|\nabla Q\|^2_{L^2(\mathbb R^2)}\). Finally, if \[ u(\omega,x)=\sum_{n\in\mathbb Z}\frac{g_n(\omega)}{(1+4\pi^2n^2)^\frac 12}e^{2\pi i nx},\quad x\in\mathbb T, \] \(\mu\) is the law of \(u\), \[ Z=\mathbb E\left[e^{\frac 16\|u\|_{L^6(\mathbb T)}^6}\mathbf 1_{[\|u\|_{L^2(\mathbb T)}\le K]}\right] \] and \(\rho\) denotes the the Gibbs measure with the formal density \[ \frac{d\rho}{d\mu}=Z^{-1}e^{\frac 16\|u\|_{L^6(\mathbb T)}^6}\mathbf 1_{[\|u\|_{L^2(\mathbb T)}\le K]} \] for \(K\le\|Q\|_{L^2(\mathbb R)}\) then the non-linear Schödinger equation (NLS) \[ i\partial_tu+\Delta u+|u|^4u=0 \] on \(\mathbb T\) is almost surely globally well-posed with respect to \(\rho\) and \(\rho\) is invariant under the NLS dynamics. An analogous result also holds for the focusing quintic generalized KdV equation on \(\mathbb T\).