Global dynamics above the ground state energy for the cubic NLS equation in 3D

TL;DR

This paper analyzes the long-term behavior of solutions to the focusing cubic nonlinear Schrödinger equation in 3D with energy near the ground state, classifying solutions into blow-up, scattering, or ground state convergence.

Contribution

It extends previous results to 3D focusing NLS, characterizing the global dynamics near the ground state with a detailed phase space partition and stability analysis.

Findings

Solutions split into nine regions with distinct behaviors

Existence of a smooth center-stable manifold containing ground states

Identification of unique stable/unstable manifolds from ground states

Abstract

We extend our previous result on the nonlinear Klein-Gordon equation to the nonlinear Schrodinger equation with the focusing cubic nonlinearity in three dimensions, for radial data of energy at most slightly above that of the ground state. We prove that the initial data set splits into nine nonempty, pairwise disjoint regions which are characterized by the distinct behaviors of the solution for large time: blow-up, scattering to 0, or scattering to the family of ground states generated by the phase and scaling freedom. Solutions of this latter type form a smooth center-stable manifold, which contains the ground states and separates the phase space locally into two connected regions exhibiting blow-up and scattering to 0, respectively. The special solutions found by Duyckaerts and Roudenko appear here as the unique one-dimensional unstable/stable manifolds emanating from the ground states. In analogy with the Klein-Gordon case, the proof combines the hyperbolic dynamics near the ground states with the variational structure away from them. The main technical ingredient in the proof is a "one-pass" theorem which precludes "almost homoclinic orbits", i.e., those solutions starting in, then moving away from, and finally returning to, a small neighborhood of the ground states. The main new difficulty compared with the Klein-Gordon case is the lack of finite propagation speed. We need the radial Sobolev inequality for the error estimate in the virial argument. Another major difference from the Klein-Gordon case is the need to control two modulation parameters.