High order uniform in time schemes for weakly nonlinear Schrödinger equation and wave turbulence

Quentin CHAULEUR, Antoine MOUZARD · 2025

Read High order uniform in time schemes for weakly nonlinear Schrödinger equation and wave turbulence by Quentin CHAULEUR, Antoine MOUZARD (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-02662.txt.

Introduction

Quentin Chauleur and Antoine Mouzard show in a 2025 paper that number-crunching schemes for weakly nonlinear Schrödinger equations can stay accurate over long times. The catalog record prints their surnames in capitals, while the text gives standard forms. Their two multiscale schemes break down Picard iterates of the solution step by step, and they are built for high precision in the small nonlinearity parameter, under a stated step rule.

The scattering traits of the schemes, reached through a low-frequency projected linear flow, give accuracy that holds evenly over long runs. That evenness is the key trait for wave turbulence studies, where slow motion must be tracked with care. Computer runs show the theory at work. The schemes are then used on motion inside the wave turbulence frame.

Catalog record

CREATORQuentin CHAULEUR, Antoine MOUZARD
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE193 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-02662

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

High order uniform in time schemes for weakly nonlinear Schrödinger equation and wave turbulence Quentin CHAULEUR, Antoine MOUZARD Abstract We introduce two multiscale numerical schemes for the time integration of weakly nonlinear Schrödinger equations, built upon the discretization of Picard iterates of the solution. These high-order schemes are designed to achieve high precision with respect to the small nonlinearity parameter under particular CFL condition. By exploiting the scattering properties of these schemes thanks to a low-frequency projected linear flow, we also establish its uniform accuracy over long time horizons. Numerical simulations are provided to illustrate the theoretical results, and these schemes are further applied to investigate dynamics in the framework of wave turbulence. 1 – Introduction We consider the nonlinear Schrödinger equation i\partial_{t}u+\Delta u=\varepsilon|u|^{p-1}u (NLS) with initial data u(0)=\varphi on the full space \mathbb{R}^{d} for 1\leq d\leq 3 . We restrict our attention to defocusing nonlinearity with odd exponents p\in 2\mathbb{N}+1 , with furthermore assumption p\geq 5 in dimension d=1 and only the cubic case p=3 for d=3 . In our framework, the nonlinearity strength \varepsilon>0 is considered small, putting ourselves in the so-called weakly nonlinear regime. Our goal is to investigate and design semi-discrete in time schemes which can capture the multiscale behavior of equation ( NLS ) with respect to this small nonlinear strength. Moreover, such schemes will prove to have uniform in time error, that is independent of T>0 the horizon time. The nonlinear Schrödinger equation is a well-studied fundamental model, with a lot of physical applications including Bose-Einstein condensation or nonlinear optics. Our motivation here lies in the simulation of wave turbulence phenomenon, which can be observed when a large number of nonlinearly interacting waves of varying wavelengths propagate in multiple directions, such as in oceanography for which equation ( NLS ) stands as a toy model. For the past years, this topic has been the subject of intense mathematical activites with different directions, see for instance the work of Deng and Hani [ 8 ] and references therein. The usual setting is to consider equation ( NLS ) on a large torus \mathbb{T}_{L}^{d} in the limit L\gg 1 and \varepsilon\ll 1 with appropriate scaling laws. On the other hand, our framework is motivated by the recent work of Faou and Mouzard [ 14 ] which consider the full space \mathbb{R}^{2} with small initial data in the weighted space \Sigma\coloneqq\left\{\varphi(x)\in H^{1}(\mathbb{R}^{2})\penalty\ \middle|\penalty\ \ |x|\varphi(x)\in L^{2}(\mathbb{R}^{2})\right\} (1.1) with a structure that mimics the large torus \mathbb{T}_{L}^{2} . This space is natural in the context of the nonlinear Schrödinger equation to obtain scattering results, which roughly states that the solution behaves in large times as a solution of the free equation. In the different approaches for wave turbulence problems, the main idea is to iterate the Duhamel formulation of the equation to get an expansion of the solution with respect to \varepsilon>0 , which gives the formal series u(t,x)=\sum_{n\geq 0}\varepsilon^{n}U_{n}(t,x) (1.2) where U_{n} is a (p+(p-1)n) -linear functional of the initial data. For instance, Deng and Hani considered in [ 8 ] such arbitrary high-order expansion for the cubic equation p=3 in dimension d\geq 3 . They identify in the large number of terms the main contributions that lead to a kinetic description of the covariances of the Fourier modes for random initial data. In a different direction, Faou and Mouzard consider in [ 14 ] first and second order expansions to identify a kinetic operator for both deterministic and random initial data. In this work, we propose a numerical scheme that captures this multiscale structure with respect to \varepsilon>0 for initial data in weighted Sobolev spaces \Sigma\cap H^{\sigma} with \sigma large enough. In particular, the terms U_{n} are defined by a recursive formula and have a tree-like structure that we exploit to construct arbitrary high order schemes. Plugging the series ( 1.2 ) into equation ( NLS ), the first order iterate corresponds to the linear equation i\partial_{t}U_{0}+\Delta U_{0}=0 with initial condition U_{0}(0)=\varphi , which solution is explicitly given by the linear flow U_{0}(t)=e^{it\Delta}\varphi . Following iterates then satisfy the cascades of equations i\partial_{t}U_{n}+\Delta U_{n}=F_{n} (1.3) with U_{n}(0)=0 and F_{n}\coloneqq\sum_{n_{1}+\ldots+n_{p}=n-1}U_{n_{1}}\overline{U_{n_{2}}}U_{n_{3}}\ldots\overline{U_{n_{p-1}}}U_{n_{p}} (1.4) for all n\geq 1 . Since the sum only involves terms of strictly lower order, U_{n} is given by a linear Schrödinger equation with a forcing that depends on all lower order terms. The main contribution of this work is to design uniform in time numerical schemes for the truncated family (U_{n})_{0\leq n\leq N-1} of order \varepsilon^{N} for any N\geq 1 . This work strongly relies on the scattering result from Carles and Gallagher [ 5 ] which controls the error between the solution to ( NLS ) and the expansion ( 1.2 ) in the space \Sigma . Namely, it ensures the bound \big\|u(t)-\sum_{n=0}^{N-1}\varepsilon^{n}U_{n}(t)\big\|_{\Sigma}\lesssim\varepsilon^{N}\|\varphi\|_{\Sigma}^{p+N(p-1)} (1.5) hence a precise time discretization for each U_{n} for n<N is enough to obtain a numerical scheme with error \varepsilon^{N} for the solution u . For example, the linear flow is an approximation of the solution up to an error of order \varepsilon while the next terms improve more and more the description of the nonlinear behavior. Moreover, due to the expansion in the small parameter \varepsilon , each term U_{n} can be computed with decreasing accuracy as n increases, specifically with an error of order \varepsilon^{N-n} . Given that each U_{n} satisfies a linear Schrödinger equation with a forcing term dependent on the preceding iterates, it is essential to ensure a consistent discretization across different values of n . This leads us to introduce the concept of nested schemes, further developed in Section 2 , which ensures such coherence. The mild formulation of ( 1.3 ) gives U_{n}(t)=-i\int_{0}^{t}e^{i(t-s)\Delta}F_{n}(s)\mathrm{d}s (1.6) for n\geq 1 since U_{n}(0)=0 which we approximate with a time discretization. For fixed T>0 , we consider a discretization mesh \tau>0 such that J\tau=T for J\in\mathbb{N} and the grid t_{j}=j\tau for 0\leq j\leq J . We then write U_{n}(t_{j})=-i\sum_{\alpha=0}^{j-1}\int_{t_{\alpha}}^{t_{\alpha+1}}e^{i(t_{j}-s)\Delta}F_{n}(s)\mathrm{d}s (1.7) and use two different methods to approximate the time integral on each subinterval [t_{\alpha},t_{\alpha+1}] . Since each F_{n}(s) depends on all lower order terms U_{n^{\prime}} with n^{\prime}<n , a careful propagation of error is required. At this stage, we also observe that the small parameter \varepsilon>0 no longer appears explicitly, which in turn imposes a CFL-type condition on the time step \tau . Our objective is to ensure an error of order \tau^{N-n} for each U_{n} , which motivates the nested structure of our numerical schemes. Another significant challenge in our work is to construct uniform in time numerical scheme using scattering, as done with the recent approach of Carles and Su [ 6 ] . A key aspect is the formulation of the dispersive equation ( NLS ) in the scattering space \varphi\in\Sigma . This perspective is particularly relevant given the physical motivation of our study, which is rooted in wave turbulence theory. Indeed, in the absence of dissipation or external forcing, solutions of ( NLS ) are expected to exhibit interesting transient dynamics over the so-called kinetic time , which grows as \varepsilon tends to 0 . Additionally, [ 6 ] establishes that a first order filtered Lie splitting method satisfies such uniform estimates, whereas extending this property to higher order schemes is far from straightforward. In

Read the complete text