Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Resonances and computations

Yvain Bruned, Frédéric Rousset, Katharina Schratz

TL;DR

Addresses the challenge of accurately simulating time dynamics for nonlinear dispersive PDEs with rough data, where classical schemes fail due to insufficient smoothing. Proposes resonance-based schemes that embed nonlinear frequency interactions through a decorated-tree (regularity-structure-inspired) formalism to capture dominant resonances exactly while approximating the rest. Develops a low-regularity theory via discrete Bourgain spaces to obtain $L^2$-level convergence and stability for time discretisations, with explicit schemes for KdV and extensions to nonpolynomial nonlinearities. Connects numerical methodology with structures from regularity theory and SPDE/Rough Path frameworks, offering a path toward structure-preserving, high-order methods for rough data in dispersive PDEs.

Abstract

The computation of time dynamics arising in nonlinear time-dependent partial differential equations is an ongoing challenge in numerical analysis, especially once roughness comes into play. Classical numerical schemes in general fail to resolve the oscillatory behaviour in the solution which leads to numerical instabilities and loss of convergence. Dispersive equations, e.g., nonlinear Schrödinger, Korteweg--de Vries and wave equations, thereby pose in particular a big problem as in contrast to the parabolic setting, no strong smoothing can be expected, i.e., if the initial data is rough, the solution stays rough which makes their approximation a delicate task. In this review we give an overview on a new numerical ansatz which aims to tackle the time dynamics of nonlinear dispersive partial differential equations even for very rough data. This is achieved by a resonance analysis and decorated tree formalism that draws its inpiration from the combinatorics used in the theory of regularity structures for solving singular SPDEs. One can hope to see this formalism applied in other contexts for dispersive PDEs and beyond.

Resonances and computations

TL;DR

Addresses the challenge of accurately simulating time dynamics for nonlinear dispersive PDEs with rough data, where classical schemes fail due to insufficient smoothing. Proposes resonance-based schemes that embed nonlinear frequency interactions through a decorated-tree (regularity-structure-inspired) formalism to capture dominant resonances exactly while approximating the rest. Develops a low-regularity theory via discrete Bourgain spaces to obtain -level convergence and stability for time discretisations, with explicit schemes for KdV and extensions to nonpolynomial nonlinearities. Connects numerical methodology with structures from regularity theory and SPDE/Rough Path frameworks, offering a path toward structure-preserving, high-order methods for rough data in dispersive PDEs.

Abstract

The computation of time dynamics arising in nonlinear time-dependent partial differential equations is an ongoing challenge in numerical analysis, especially once roughness comes into play. Classical numerical schemes in general fail to resolve the oscillatory behaviour in the solution which leads to numerical instabilities and loss of convergence. Dispersive equations, e.g., nonlinear Schrödinger, Korteweg--de Vries and wave equations, thereby pose in particular a big problem as in contrast to the parabolic setting, no strong smoothing can be expected, i.e., if the initial data is rough, the solution stays rough which makes their approximation a delicate task. In this review we give an overview on a new numerical ansatz which aims to tackle the time dynamics of nonlinear dispersive partial differential equations even for very rough data. This is achieved by a resonance analysis and decorated tree formalism that draws its inpiration from the combinatorics used in the theory of regularity structures for solving singular SPDEs. One can hope to see this formalism applied in other contexts for dispersive PDEs and beyond.
Paper Structure (7 sections, 6 theorems, 98 equations)

This paper contains 7 sections, 6 theorems, 98 equations.

Table of Contents

  1. Introduction
  2. Classical methods and their problem for rough data
  3. Splitting methods
  4. Exponential integrators
  5. Structure of the solution and classical methods
  6. Resonances as a computational tool: The main idea
  7. Error analysis at low regularity with discrete Bourgain spaces

Key Result

theorem 1

For every decorated tree $T$ one has, Then, one has where there exists an operator $\mathcal{L}^{r}_{ \normalfont \text{\tiny{low}}}(T,n)$ which can be explicitly computed embeds the necessary regularity of the solution.

Theorems & Definitions (9)

  • remark 1
  • theorem 1
  • corollary 1
  • definition 1
  • corollary 2
  • remark 2
  • theorem 2
  • lemma 1
  • lemma 2