[Paper Review] Ill-posedness issues for nonlinear dispersive equations

Nikolay Tzvetkov|arXiv (Cornell University)|Nov 20, 2004

Advanced Mathematical Physics ProblemsMathematics54 references22 citations

TL;DR

This paper investigates ill-posedness in nonlinear dispersive equations, particularly KdV and nonlinear Schrödinger equations, by constructing high-frequency approximate solutions on small time intervals. It demonstrates that the flow map fails to be uniformly continuous below a certain regularity threshold, revealing that semi-linear well-posedness is more delicate in dispersive settings than in hyperbolic ones, with critical regularity indices shifted due to concentration on curves like closed geodesics.

ABSTRACT

These notes are devoted to the notion of well-posedness of the Cauchy problem for nonlinear dispersive equations. We present recent methods for proving ill-posedness type results for dispersive PDE's. The common feature in the analysis is that the proof of such results requires the construction of high frequency approximate solutions on small time intervals (possibly depending on the frequency).

Motivation & Objective

• To analyze the well-posedness and ill-posedness of Cauchy problems for nonlinear dispersive equations in Sobolev spaces.
• To investigate why semi-linear well-posedness fails below certain regularity thresholds, even when classical well-posedness might seem plausible.
• To demonstrate that the critical regularity index for semi-linear well-posedness can be shifted from the scaling-based index due to concentration on curves such as closed geodesics.
• To provide a systematic method for proving ill-posedness using high-frequency approximate solutions on shrinking time intervals.
• To clarify the distinction between classical well-posedness and stronger semi-linear well-posedness in dispersive PDEs.

Proposed method

• Construct high-frequency approximate solutions that concentrate on small time intervals depending on the frequency, using a modified ansatz for the solution.
• Employ a bootstrap argument on a rescaled time interval to control the growth of the solution norm and validate the ansatz.
• Use the Duhamel integral formulation to reduce the PDE to a fixed-point problem in a function space, analyzing the contraction properties.
• Apply frequency-localized estimates and bounds on the nonlinear term, particularly in the context of $L^2$-based Sobolev norms.
• Introduce a rescaling $\tilde{M}_n(T)$ to control the evolution of the solution on time intervals $[0, T_n]$ with $T_n \sim n^{\frac{5s}{2} - \frac{3}{8} - \varepsilon}$.
• Leverage semi-classical reasoning and geometric concentration (e.g., on closed geodesics) to destabilize the flow map and break uniform continuity.

Experimental results

Research questions

• RQ1 Why does the flow map for nonlinear dispersive equations fail to be uniformly continuous below a certain regularity threshold, even when solutions exist?
• RQ2 How does the presence of closed geodesics or other geometric structures affect the critical regularity index for semi-linear well-posedness?
• RQ3 To what extent does the scaling-based critical index for well-posedness differ from the actual index required for semi-linear well-posedness in dispersive equations?
• RQ4 Can the failure of semi-linear well-posedness be systematically proven using high-frequency approximate solutions on shrinking time intervals?
• RQ5 Is there a geometric notion of critical exponent associated with curves, analogous to the scaling-based one for points?

Key findings

• For the KdV-type equation on the sphere, the Cauchy problem is not semi-linearly well-posed in $H^s(S^2)$ for $s < 1/8$, despite existence of solutions.
• When $1/8 < s \leq 3/20$, the flow map fails to be uniformly continuous due to $n^{-1/8 - s/2 - \varepsilon}$-type growth in high-frequency components over time intervals $T_n \sim n^{5s/2 - 3/8 - \varepsilon}$.
• The time interval $T_n$ satisfies $n^{1/2 - 2s}T_n \to \infty$ as $n \to \infty$, ensuring sufficient oscillations to break uniform continuity of the flow map.
• For $0 \leq s \leq 1/8$, the ansatz is justified up to time one, confirming ill-posedness in this regime as well, based on Banica's work.
• On $S^6$, the cubic NLS with $\alpha = 1$ is not semi-linearly well-posed in $H^1(S^6)$, showing that the failure is not limited to low regularity or specific geometries.
• The method reveals that semi-linear well-posedness is more sensitive than classical well-posedness, as it requires stronger continuity of the flow map, which can fail even when existence and uniqueness hold.