Failure of Fixed-Profile Modified Scattering at the Pure $L^2$ Endpoint for the 1D Defocusing Cubic NLS

Technical Summary: Failure of Fixed-Profile Modified Scattering at the Pure $L^2$ Endpoint for the 1D Defocusing Cubic NLS

Problem Statement
The paper addresses the asymptotic behavior of solutions to the one-dimensional defocusing cubic nonlinear Schrödinger equation (NLS):
$$i q_t + q_{xx} = 2|q|^2 q, \quad q(0) = q^* \in L^2(\mathbb{R}).$$
Specifically, it investigates the validity of the fixed-profile modified-scattering ansatz at the unweighted $L^2$ endpoint. The standard conjecture posits that for any initial data $q^* \in L^2(\mathbb{R})$, there exists a time-independent profile $V \in L^2(\mathbb{R})$ such that the solution $q(t)$ behaves asymptotically as:
$$\|\mathcal{F}q(t) - e^{-i(\log t)|V|^2}V\|_{L^2_\xi} \to 0 \quad \text{as } t \to +\infty,$$
where $\mathcal{F}$ denotes the unitary Fourier transform. While this modified scattering behavior is well-established for data with higher regularity or spatial decay (e.g., weighted $L^2$ spaces, $H^1$, or $H^{1,1}$), the behavior at the pure $L^2$ endpoint remains an open question regarding the existence of such a fixed profile.

Methodology
The author employs a constructive counterexample approach, utilizing the integrability of the defocusing NLS equation via the inverse scattering transform (IST). The methodology proceeds through several key technical steps:

1. Spectral Normalization and Continuity: The paper establishes the continuity of the direct spectral map from $L^2(\mathbb{R})$ to the space of Schur functions. It reconciles different spectral conventions (Bessonov–Denisov, Deift–Zhou, and Bessonov–Gubkin) to define a transmission coefficient $a(\zeta)$ and a reflection coefficient $r(\zeta)$. A crucial conservation law is identified: the quantity $\rho_q(\xi) = \frac{1}{\pi} \log |a_q(-\xi/2 + i0)|$ is conserved in time and determines the necessary modulus of any potential scattering profile $V$ (i.e., $|V|^2 = \rho_{q^*}$).

2. Analysis of Smooth Data: For smooth, compactly supported data, the Deift–Zhou asymptotic expansion is converted into a strong $L^2$ convergence result. This establishes that for regular data, a corrected profile $V_q$ exists, with a phase determined by a one-sided logarithmic operator $T$ acting on the transmission density.

3. Far-Bump Interference Mechanism: The core of the construction relies on the exact composition of Zakharov–Shabat transfer matrices. The author demonstrates that adding a small, spatially distant "bump" (a smooth, compactly supported perturbation $p(x-X)$) to a potential $q$ induces a specific interference in the transmission coefficient.

   • The composition rule creates a term $e^{iXz}$ in the reflection coefficient.
   • The one-sided logarithmic operator $T$, which governs the phase of the asymptotic profile, acts as an amplifier. When applied to the oscillatory term $e^{iXz}$, it produces a shift of order $\log X$.
   • By choosing the amplitude of the bump $\epsilon$ and the location $X$ such that $\epsilon \log X = \kappa$ (a constant), the author can insert a uniform phase shift into the asymptotic profile while keeping the $L^1$ and $L^2$ norms of the perturbation arbitrarily small.

4. Inductive "Gliding-Hump" Construction: The paper constructs a specific initial datum $q^*$ as an infinite sum of disjoint smooth bumps. The construction is adaptive (online):

   • At each step $n$, a time $t_n$ is chosen such that the solution $q_n$ (the partial sum) is close to its profile $V_n$.
   • A new bump is added at a location $X_{n+1}$ far to the right. The parameters are chosen to ensure the $L^2$ distance between the new datum and the old one is small enough to maintain continuity at time $t_n$, yet large enough to induce a significant jump in the asymptotic profile $V_{n+1}$ relative to $V_n$.
   • This process generates a sequence of profiles $(V_n)$ that is not Cauchy in $L^2$, despite the sequence of initial data converging in $L^2$.

Key Contributions and Results
The main result is Theorem 1.1, which states:

• There exists a real-valued initial datum $q^* \in L^1(\mathbb{R}) \cap \bigcap_{k \ge 0} H^k(\mathbb{R})$ (smooth and integrable) such that $x q^* \notin L^2(\mathbb{R})$ (failure of the first moment), for which no profile $V \in L^2(\mathbb{R})$ satisfies the fixed-profile modified-scattering condition.
• The $L^2$ norm of this counterexample can be prescribed arbitrarily.
• The construction relies on the fact that while the partial data converge in $L^1 \cap L^2$, the resulting asymptotic profiles do not converge in $L^2$ due to the accumulation of phase shifts amplified by the logarithmic operator.

The paper proves that the obstruction is specific to the requirement of a single, time-independent profile. The failure arises because the phase information required to describe the solution at infinity can be pushed to arbitrarily fine spectral scales by spatially distant perturbations, which a fixed profile cannot capture.

Significance and Claims
The paper claims to resolve the question of fixed-profile modified scattering at the pure $L^2$ endpoint by demonstrating its failure.

• Negative Result: It establishes that the classical modified-scattering picture, where the solution converges to a fixed profile $V$ modulated by a logarithmic phase, is false for general $L^2$ data.
• Scope of Obstruction: The obstruction is specific to the "fixed-profile" nature of the ansatz. The paper explicitly leaves open the possibility that adaptive or scale-dependent renormalizations might still provide a valid asymptotic description for $L^2$ solutions.
• Methodological Insight: The work highlights the sensitivity of the inverse scattering map at the $L^2$ endpoint, showing that the "gliding-hump" phenomenon allows for the insertion of high-frequency oscillations into the logarithm of the transmission coefficient with negligible cost in the physical space norms.

The author notes that the counterexample was initially suggested through an interaction with an AI tool, but the rigorous proof and construction are original to the paper. The result shifts the focus of the field from proving convergence to a fixed profile to identifying the correct, potentially more complex, asymptotic object for general unweighted $L^2$ solutions.