On scattering for NLS: rigidity properties and numerical simulations via the lens transform

This paper introduces a novel numerical methodology based on the lens transform to efficiently compute the scattering operator for the defocusing nonlinear Schrödinger equation, combining new theoretical identities with simulations that validate known results and propose conjectures for previously unexplored regimes, including long-range scattering.

The Gist

Imagine you are watching a drop of ink fall into a glass of water. At first, it's a tight, distinct blob. But as time goes on, the water currents (dispersion) spread it out until it becomes a faint, almost invisible haze that fills the entire glass.

In the world of physics, the Nonlinear Schrödinger Equation (NLS) describes how waves (like light or quantum particles) behave. Usually, these waves spread out and fade away, just like the ink. However, sometimes the waves interact with themselves (the "nonlinear" part), which can either make them spread faster or pull them back together.

The big question this paper asks is: If we wait an infinite amount of time, can we predict exactly what the wave will look like?

This is called Scattering. It's like trying to predict the final, scattered pattern of the ink after the water has settled forever. The problem is that "forever" is impossible to simulate on a computer. If you try to run a simulation for a million years, the wave spreads so far that it leaves your computer screen, and the math breaks down.

Here is how the authors solved this puzzle, explained simply:

1. The Problem: The Wave Runs Away

Think of the wave as a runner on a track. In a normal simulation, the track is a fixed size. As the runner (the wave) gets tired and slows down, they spread out. Eventually, they run off the edge of the track. To see where they end up, you'd need an infinitely long track, which is impossible to build in a computer.

2. The Solution: The "Lens" Trick

The authors used a mathematical magic trick called the Lens Transform.

Imagine you are looking at that runner through a special pair of glasses (a lens).

• Normal View: The runner starts small, spreads out, and runs off into the distance.
• Lens View: As the runner spreads out, the lens zooms in on them. As time goes on, the lens also compresses the timeline.

In this "Lens World," the runner never actually leaves the track. Instead of running off to infinity, they slow down and settle into a specific, finite spot right in front of you. The infinite future of the real world is compressed into a short, manageable time interval (from $-\pi/2$ to $\pi/2$) in the Lens World.

This allowed the authors to simulate "infinite time" on a standard computer by just watching the wave settle down in this compressed, zoomed-in universe.

3. The Rules of the Game (Rigidity)

The authors discovered some strict "laws of physics" that these waves must obey, which they call Rigidity Properties.

• The Conservation of Identity: No matter how much the wave twists, turns, or interacts with itself, certain things never change. It's like a magical coin: no matter how many times you flip it, it always has the same total weight and spin.
• The "No-Translation" Rule: They proved that the wave cannot simply shift to the left or right as it scatters. If you see a wave pattern at the start, and a shifted version of that pattern at the end, something is wrong. The wave might change shape, but it can't just slide over like a sliding puzzle piece.

4. The Experiments: Testing the Limits

The authors used their new "Lens Glasses" to run thousands of simulations to test some deep mysteries:

• The "Rotating" Mystery: In some specific cases (when the wave interaction is perfectly balanced), the wave can settle into a state where it just spins in place, looking the same every time you check. The authors found that this only happens in very specific, delicate situations. If you change the rules even a tiny bit, this spinning stops, and the wave just scatters normally.
• The "Long-Range" Problem: In one dimension (a single line), the waves interact over very long distances. It's like a whisper that never quite dies out. The authors showed that even here, their Lens method works, provided they add a tiny "phase correction" (like adjusting the timing of a song) to account for the lingering echo.
• The "Big Data" Surprise: They tested what happens when the starting wave is huge. Analytical math says scattering should work for small waves, but they weren't sure about big ones. Their simulations suggested that for certain types of waves, if the starting wave is too big, it might not scatter cleanly. It might get stuck or behave chaotically, breaking the standard rules.

5. The Focusing Case: Solitons

Finally, they looked at a "focusing" version of the equation, where waves try to pull themselves together. This creates Solitons—stable, self-reinforcing waves that act like particles.

• They found that even if you start with a wave slightly smaller than a famous "ground state" (the most stable soliton), it might still refuse to scatter. It's as if there is a hidden threshold: if you are even a tiny bit too "heavy," the wave refuses to let go and scatter, staying trapped in a soliton shape.

Summary

In short, this paper is about building a better telescope.

1. The Problem: We couldn't see the end of the wave because it ran off the edge of the universe.
2. The Tool: They built a "Lens" that compresses the infinite future into a finite view.
3. The Discovery: They proved strict rules about how these waves must behave and used their new tool to find that the universe is more complex than we thought—sometimes, big waves don't scatter at all, and stable "spinning" waves are rarer than we hoped.

It's a blend of deep math and computer power that helps us understand how waves in the universe eventually settle down, or if they ever truly do.

Technical Summary

1. Problem Statement

The paper addresses the numerical computation and theoretical analysis of the scattering operator $S$ associated with the defocusing Nonlinear Schrödinger Equation (NLS):
$$i\partial_t u + \frac{1}{2}\Delta u = |u|^{2\sigma}u, \quad (t, x) \in \mathbb{R} \times \mathbb{R}^d$$
The scattering operator maps an asymptotic state $u_-$ (as $t \to -\infty$) to an asymptotic state $u_+$ (as $t \to +\infty$).

Key Challenges:

• Infinite Time Domain: Standard numerical simulations struggle with the infinite time interval required to reach asymptotic states.
• Dispersive Effects: The linear Schrödinger group $U_0(t)$ causes solutions to spread out (disperse) over space as $t \to \infty$. In a fixed computational domain, the solution eventually leaves the box, making it difficult to capture the asymptotic state without massive domains or complex boundary treatments.
• Long-Range Scattering: For specific powers (e.g., 1D cubic, $\sigma=1/d$), standard scattering fails, requiring modified scattering theories with phase corrections.

2. Methodology: The Lens Transform

The authors overcome the computational barriers by employing the Lens Transform, a space-time compactification technique.

• Transformation: They introduce a new variable $v(t, x)$ defined for $t \in (-\pi/2, \pi/2)$:
  $$v(t, x) = \frac{1}{(\cos t)^{d/2}} u\left(\tan t, \frac{x}{\cos t}\right) e^{-i \frac{|x|^2}{2} \tan t}$$
• Transformed Equation: The function $v$ satisfies a modified NLS equation with a harmonic oscillator potential:
  $$i\partial_t v + \frac{1}{2}\Delta v = \frac{|x|^2}{2}v + (\cos t)^{d\sigma-2}|v|^{2\sigma}v$$
• Advantages:
  1. Time Compactification: The infinite time interval $t \in (-\infty, \infty)$ for $u$ maps to the finite interval $t \in (-\pi/2, \pi/2)$ for $v$.
  2. Spatial Confinement: The harmonic potential $\frac{|x|^2}{2}$ acts as a confining force. Unlike the free Schrödinger equation where solutions disperse to infinity, solutions to the transformed equation remain localized, allowing for the use of bounded computational domains without boundary artifacts.
  3. Spectral Discretization: The natural basis for the harmonic oscillator is the Hermite functions. The authors use a Hermite spectral method combined with Lie splitting (operator splitting) to solve the transformed equation. This allows for exact computation of Fourier transforms and derivatives within the spectral basis.

Numerical Algorithm:

1. Initialization: Convert the asymptotic state $u_-$ to the initial state $v(-\pi/2)$ using the inverse lens transform.
2. Propagation: Evolve $v$ from $t = -\pi/2$ to $t = \pi/2$ using the split-step method (linear harmonic part + nonlinear part).
3. Extraction: Convert the final state $v(\pi/2)$ back to the asymptotic state $u_+$ (the scattering operator output).

3. Key Theoretical Contributions

The authors establish new analytical properties of the scattering operator to validate their numerical methods and deepen theoretical understanding.

• New Conservation Laws (Rigidity Properties):
  They prove two new identities relating the asymptotic states $u_\pm$ and the initial data $u_0$:

  1. Center of Mass Conservation:
     $$\int_{\mathbb{R}^d} x |u_-(x)|^2 dx = \int_{\mathbb{R}^d} x |u_0(x)|^2 dx = \int_{\mathbb{R}^d} x |u_+(x)|^2 dx$$
     This implies the scattering operator cannot act as a non-trivial translation on asymptotic states.
  2. L2-Critical Case Identity ($\sigma = 2/d$):
     A new identity involving the variance and the Fourier transform norm:
     $$\|x u_-\|_{L^2}^2 + \frac{2d}{2+d}\|\hat{u}_-\|_{L^{2+4/d}}^{2+4/d} = \|x u_0\|_{L^2}^2 = \dots$$
     This is derived using the pseudo-conformal conservation law and the lens transform.

• Rotating Points:
  They revisit the existence of "rotating points" (eigenfunctions of the scattering operator, $S(u_-) = e^{i\theta}u_-$).

  • Analytical Proof: They provide a concise proof (using the lens transform) that in the $L^2$-critical case ($\sigma=2/d$), infinitely many such points exist for any phase $\theta$.
  • Numerical Investigation: They attempt to find such points in the $L^2$-supercritical regime ($\sigma > 2/d$).

4. Numerical Results and Findings

A. Validation

The numerical scheme was validated against known conservation laws (Mass, Energy, Momentum, and the new Center of Mass identity). The errors were found to be negligible ($10^{-14}$ to $10^{-6}$ depending on resolution), confirming the method's accuracy.

B. Rotating Points in Supercritical Regimes

• Hypothesis: Do rotating points exist for $\sigma > 2/d$?
• Result: The simulations suggest no. While the method successfully finds rotating points in the critical case ($\sigma=2$), attempts to find them in the supercritical case ($\sigma=2.01$) failed to converge to a solution, even with high resolution.
• Conjecture: The existence of rotating points is likely a unique feature of the $L^2$-critical case and does not extend to supercritical powers.

C. Scattering Thresholds (Large Data)

• Context: Analytical theory guarantees scattering in the space $\Sigma$ (weighted Sobolev space) for $\sigma \geq \sigma_0(d)$ (Strauss exponent). For $1/d < \sigma < \sigma_0(d)$, scattering is known for small data, but the large data case is open.
• Experiment: The authors simulated large random initial data for $d=1$ with $\sigma = 1.15$ (which lies in the "unknown" range $1 < \sigma < \sigma_0(1) \approx 1.28$).
• Result: The $\Sigma$-norm of the solution $v(t)$ became unbounded as $t \to \pi/2$.
• Conjecture: For large data in the range $1/d < \sigma < \sigma_0(d)$, asymptotic completeness in $\Sigma$ fails. The solution may scatter in $H^1$ but not in the weighted space $\Sigma$.

D. Long-Range Scattering (1D Cubic, $\sigma=1$)

• Challenge: In 1D cubic NLS, standard scattering fails; a phase correction is required. The lens transform introduces a singularity at $t = \pm \pi/2$.
• Approach: The authors incorporated the known phase correction into the lens-transformed variables and avoided the singular endpoints numerically.
• Result: The method successfully recovered the modified scattering operator, demonstrating that the lens transform can handle long-range effects with appropriate modifications.

E. Focusing Case and Ground States

• Context: In the focusing 1D cubic case, solitary waves (ground states) exist. It is an open question if the ground state is the threshold for the existence of modified scattering.
• Result: Simulations with initial data scaled by $\alpha \in (0, 1)$ relative to the ground state suggest that modified scattering may fail even for data smaller than the ground state (specifically for $\alpha \approx 0.75$). This challenges the idea that the ground state is the sharp threshold for scattering in the focusing regime.

5. Significance and Impact

• Methodological Breakthrough: This is the first application of the lens transform for numerical simulations of the scattering operator. It solves the long-standing problem of simulating infinite-time dispersive dynamics on finite computational domains.
• Theoretical Insight: The paper provides rigorous proofs of new rigidity properties (conservation of the center of mass) and offers strong numerical evidence for conjectures regarding the non-existence of rotating points in supercritical regimes and the failure of scattering in $\Sigma$ for large data in intermediate regimes.
• Future Directions: The authors propose that the lens transform combined with Hermite spectral methods is a powerful tool for exploring regimes where analytical proofs are currently out of reach, particularly for large data problems and long-range scattering.

In summary, the paper successfully bridges rigorous analysis and high-precision numerics to explore the global dynamics of the NLS equation, revealing new structural properties and challenging existing assumptions about scattering thresholds.