[Paper Review] Effective dynamics of the nonlinear Schrödinger equation on large domains

Tristan Buckmaster, Pierre Germain+2 more|arXiv (Cornell University)|Oct 12, 2016

Advanced Mathematical Physics Problems17 references4 citations

TL;DR

This paper derives the continuous resonant (CR) equation as an effective dynamics model for the nonlinear Schrödinger equation on large periodic domains $[0,L]^d$ in the limit of small nonlinearity ($\epsilon \to 0$) and large $L$, using advanced analytic number theory tools like the Hardy-Littlewood circle method adapted to PDEs. The key result is a rigorous derivation of the CR equation that captures long-time dynamics beyond both the nonlinear and Euclidean time-scales, with error bounds decaying as $L^{-1+}$ or $L^{-1/3+}$ depending on dimension.

ABSTRACT

We consider the nonlinear Schrödinger (NLS) equation posed on the box $[0,L]^d$ with periodic boundary conditions. The aim is to describe the long-time dynamics by deriving effective equations for it when $L$ is large and the characteristic size $ε$ of the data is small. Such questions arise naturally when studying dispersive equations that are posed on large domains (like water waves in the ocean), and also in theory of statistical physics of dispersive waves, that goes by the name of "wave turbulence". Our main result is deriving a new equation, the continuous resonant (CR) equation, that describes the effective dynamics for large $L$ and small $ε$ over very large time-scales. Such time-scales are well beyond the (a) nonlinear time-scale of the equation, and (b) the Euclidean time-scale at which the effective dynamics are given by (NLS) on $\mathbb R^d$. The proof relies heavily on tools from analytic number theory, such as a relatively modern version of the Hardy-Littlewood circle method, which are modified and extended to be applicable in a PDE setting.

Motivation & Objective

• To describe the long-time dynamics of the nonlinear Schrödinger equation (NLS) on large periodic domains $[0,L]^d$ when the nonlinearity is weak ($\epsilon \to 0$) and the domain is large ($L \to \infty$).
• To identify an effective equation that governs the dynamics beyond the standard nonlinear time-scale $T_{nl} \sim \epsilon^{-2p}$ and the Euclidean time-scale $T_{\mathcal{E}} \sim L$, which are limits of the standard approximations.
• To rigorously derive the continuous resonant (CR) equation as the effective dynamics in this regime, capturing energy transfer and resonant interactions among Fourier modes.
• To establish quantitative error estimates between the true NLS solution and the CR equation, showing convergence in Sobolev norms with decay rates in $L$.

Proposed method

• Adapts the Hardy-Littlewood circle method from analytic number theory to the PDE setting to analyze lattice sums over resonant Fourier interactions in the NLS equation.
• Identifies the resonant set via the conditions $\mathcal{S}_{2p+1}(K) = 0$ and $\Omega_{2p+1}(K) = 0$, corresponding to conservation of wave vector and frequency in $2p+1$-wave interactions.
• Introduces a decomposition of the resonant interaction into even and odd mode pairs $\boldsymbol{J_e}, \boldsymbol{J_o}$, reducing the problem to analyzing $\boldsymbol{J_e} \cdot \boldsymbol{J_o} = 0$.
• Uses integration by parts on the phase function $\Omega_{2p+1}(K)$ to control oscillatory integrals and avoid growth in $K$, ensuring uniform bounds across frequency scales.
• Defines a lattice sum over resonant configurations and compares it to a continuous integral $\mathcal{P}(W)$, with correction terms for logarithmic divergences when $pn = 2$.
• Establishes error bounds via weighted Sobolev norms $X^{\ell,N}$, showing the difference between the discrete sum and the continuous integral decays as $L^{-1+}$ or $L^{-1/3+}$ depending on dimension.

Experimental results

Research questions

• RQ1 What effective equation governs the long-time dynamics of the nonlinear Schrödinger equation on large periodic domains when both the nonlinearity and domain size are scaled?
• RQ2 How do resonant interactions among Fourier modes dominate the dynamics beyond the nonlinear and Euclidean time-scales?
• RQ3 Can the Hardy-Littlewood circle method be adapted to derive rigorous asymptotics for lattice sums arising in dispersive PDEs?
• RQ4 What is the rate of convergence between the true NLS solution and the continuous resonant (CR) equation in the large-$L$ and small-$\epsilon$ limit?
• RQ5 Why does the CR equation require a logarithmic correction term when $pn = 2$, and how does this affect the error estimates?

Key findings

• The continuous resonant (CR) equation is rigorously derived as the effective dynamics for the NLS equation on large domains $[0,L]^d$ in the limit $\epsilon \to 0$, $L \to \infty$, valid over time-scales beyond both the nonlinear and Euclidean regimes.
• The error between the true NLS solution and the CR equation decays as $L^{-1+}$ in the generic case ($pn \neq 2$), with a more refined $L^{-1/3+}$ rate in the critical case $pn = 2$.
• For $pn \neq 2$, the difference between the discrete resonant sum and the continuous integral is bounded by $\|\Delta(W)\|_{X^\ell} \lesssim L^{-1+} \prod_{i=1}^{2p+1} \|f_i\|_{X^{\ell+n+2,4n+2}}$, showing uniform convergence.
• In the case $pn = 2$, a correction operator $\mathcal{C}(W)$ is required due to logarithmic divergence, and the error $\|\widetilde{\Delta}(W)\|_{X^\ell} \lesssim L^{-1/3+} \prod_{i=1}^{2p+1} \|f_i\|_{X^{\ell+n+4,4n+3}}$ is established.
• The method successfully extends the circle method to PDEs by controlling phase integrals and avoiding $K$-dependent growth, enabling uniform bounds across frequency scales.
• The derivation confirms that resonant interactions dominate the long-time dynamics, justifying the CR equation as the correct effective model in the large-$L$, small-$\epsilon$ regime.