Unconditional uniqueness of solutions for nonlinear dispersive equations

Abstract: When a solution to the Cauchy problem for nonlinear dispersive equations is obtained by a fixed point argument using auxiliary function spaces, it is non-trivial to ensure uniqueness of solutions in a natural space such as the class of continuous curves in the data space. This property is called unconditional uniqueness, and proving it often requires some additional work. In the last decade, unconditional uniqueness has been shown for some canonical nonlinear dispersive equations by an integration-by-parts technique, which can be regarded as a variant of the (Poincar\'e-Dulac) normal form reduction. In this article, we aim to provide an abstract framework for establishing unconditional uniqueness as well as existence of certain weak solutions via infinite iteration of the normal form reduction. In particular, in an abstract setting we find two sets of fundamental estimates, each of which can be used repeatedly to generate all multilinear estimates of arbitrarily high degrees required in this scheme. Then, we confirm versatility of the framework by applying it to various equations, including the cubic nonlinear Schr\"odinger equation (NLS) in higher dimension, the cubic NLS with fractional Laplacians, the cubic derivative NLS, and the Zakharov system, for which new results on unconditional uniqueness are obtained under the periodic boundary condition.