The nonlinear Schrödinger equation with a potential (Q1669638)

The authors describe the long-time behavior of small solutions to the cubic nonlinear Schrödinger equation: \(i\partial _{t}u-\partial _{x}^{2}u+Vu=\lambda \left| u\right| ^{2}u\) posed in \(\mathbb{R} \times \mathbb{R}\). The potential \(V\in W^{2,1}\cap L_{\gamma }^{1}\) (\( \gamma >6\)) is supposed to have no bound states and to be generic, that is \( \int_{\mathbb{R}}V(x)m(x)dx\neq 0\) where \(m\) is the unique solution of \( (-\partial _{x}^{2}+V)m=0\) which approaches 1 when \(x\) goes to \(+\infty \). Here \(L_{\gamma }^{1}=\{f:\left| x\right| ^{\gamma }f\in L^{1}\}\). The solution starts from an initial data \(u_{0}\in H^{3}\) which satisfies \( xu_{0}\in L^{2}\). The main result of the paper first proves the existence of a unique global solution which satisfies \(\sup_{t\in \mathbb{R}}\left\| u(t)\right\| _{L_{x}^{\infty }}\lesssim \varepsilon _{0}(1+\left| t\right| )^{-1/2}\) where \(\varepsilon _{0}=\left\| u_{0}\right\| _{H^{3}}+\left\| xu_{0}\right\| _{L^{2}}\), if \(\varepsilon _{0}\) is small enough. The authors also prove bounds on the distorted Fourier transform \(\widetilde{f}(t,k):=e^{-itk^{2}}\widetilde{u}(t,k)\) of the profile \(f(t,x):=e^{-it(\partial _{x}^{2}+V)}u(t,x)\), \(\widetilde{\phi }\) being defined through \(\widetilde{\phi }(k)=\int_{\mathbb{R}}\overline{\psi (x,k)}\phi (x)dx\) where \(\overline{\psi (x,k)}=\frac{1}{\sqrt{2\pi }} T(k)f_{+}(x,k)\) if \(k\geq 0\) and \(\overline{\psi (x,k)}=\frac{1}{\sqrt{2\pi } }T(-k)f_{-}(x,k)\) otherwise, \(T(k)\) being the reflection coefficients associated to \(V\) and \(f_{\pm }\) being the solutions of \((-\partial _{x}^{2}+V)f_{\pm }=k^{2}f_{\pm }\) in \(\mathbb{R}\) with \(f_{+}(x,k)\sim e^{ixk}\) as \(x\rightarrow \infty \) and \(f_{-}(x,k)\sim e^{-ixk}\) as \( x\rightarrow -\infty \). For the proof, the authors recall elements of the spectral theory of operators \(-\partial _{x}^{2}+V\) in \(\mathbb{R}\) and of the Weyl-Kodaira-Titschmarsh theory associated to the distorted Fourier transform. They build a representation of \(\widetilde{f}\) using a multilinear oscillatory integral that they analyze.