Long time behavior of small solutions of NLS with non-generic potentials in one dimension

Technical Summary: Long Time Behavior of Small Solutions of NLS with Non-Generic Potentials in One Dimension

Problem Statement
This paper investigates the long-time behavior of small solutions to the one-dimensional cubic nonlinear Schrödinger equation (NLS) with a real-valued external potential $V(x)$:
$$i\partial_t u - \partial_{xx} u + V(x)u \pm |u|^2 u = 0, \quad u(0, x) = u_0(x).$$
The primary focus is on the non-generic case, where the Schrödinger operator $H = -\partial_{xx} + V$ admits a zero-energy resonance (a non-trivial bounded solution to $H\phi = 0$). Unlike the generic case or symmetric non-generic cases (e.g., even potentials), the non-generic setting without symmetry assumptions presents significant analytical challenges. Specifically, the standard distorted Fourier transform (dFT) exhibits a discontinuity at zero frequency, and the nonlinear interactions generate "dangerous" principal value (p.v.) terms that lack low-frequency improvements. The goal is to establish quantitative decay bounds for small initial data in weighted Sobolev spaces over an almost global time scale.

Methodology
The analysis relies on the method of space-time resonances adapted to the distorted Fourier setting, combined with a bootstrap argument. The core technical innovations include:

1. Modified Distorted Fourier Transform (dFT):
   In the non-generic case, the standard dFT kernel $K(x, k)$ is discontinuous at $k=0$. The author introduces a modified kernel $K^\sharp(x, k)$ via a unitary transformation of the standard basis. This modification ensures continuity at zero frequency, preserving Plancherel identities and allowing for a consistent spectral analysis. The kernel is decomposed into a "singular" part $K^\sharp_S$ (capturing the resonance effects) and a "regular" part $K^\sharp_R$ (localized in space).

2. Decomposition of the Nonlinear Spectral Distribution (NSD):
   The nonlinear interactions in the distorted frequency space are encoded in the NSD, $\mu^\sharp(k, \ell, m, n)$. The paper provides a refined decomposition of this distribution into four components:

   • $\mu^\sharp_0$: A linear combination of Dirac deltas, analogous to the flat case ($V=0$).
   • $\mu^\sharp_L$: An "improved low-frequency" part where coefficients vanish at zero frequency, allowing the use of standard smoothing estimates.
   • $\mu^\sharp_{p.v.}$: A "dangerous" principal value part. Crucially, in the general non-generic case (without symmetry), this term does not vanish and lacks the low-frequency improvement found in previous works (e.g., Chen and Pusateri [5]).
   • $\mu^\sharp_R$: A regular part consisting of localized functions and smooth distributions.

3. Estimates and Bootstrap Scheme:

   • Linear Estimates: The paper establishes dispersive decay ($t^{-1/2}$ in $L^\infty$), smoothing estimates, and local decay estimates tailored to the modified kernel and the specific low-frequency structures of the singular parts.
   • Nonlinear $L^2$ Estimates: For the singular parts ($N_0, N_L, N_{p.v.}$), the author employs a commutation identity between $\partial_k$ and the trilinear forms. While $N_0$ and $N_L$ are handled via standard smoothing estimates, the treatment of $N_{p.v.}$ requires a novel Fourier restriction-type inequality (Lemma 5.5). This inequality, $\left(\int |\hat{F}(k^2)|^2 dk\right)^{1/2} \lesssim \|F\|_{L^{4/3}}$, is central to controlling the growth of the $L^2_k$ norm of the profile derivative without low-frequency cancellation.
   • Nonlinear $L^\infty$ Estimates: The regular part contributes only to higher-order remainders. The singular parts are analyzed using stationary phase lemmas for oscillatory integrals with distributional symbols.

4. Bootstrap Argument:
   The proof utilizes a bootstrap hypothesis on the profile $f(t, x) = e^{-itH}u(t, x)$ in the distorted frequency space, assuming bounds of the form $\|f^\sharp\|_{L^\infty_k} + \langle t \rangle^{-1/4} \|\partial_k f^\sharp\|_{L^2_k} \leq 2C_0 \varepsilon$. The analysis shows that the growth of the $L^2_k$ norm of $\partial_k f^\sharp$ is at most $O(\varepsilon^3 \langle t \rangle^{1/4})$, which is sufficient to close the bootstrap argument on an almost global time interval.

Key Results
The main theorem (Theorem 1.1) establishes that for a non-generic potential $V$ satisfying specific decay and spectral conditions (no eigenvalues, $V \in L^1(\langle x \rangle^\gamma dx)$ with $\gamma > 5/2$), there exists $\varepsilon_0 > 0$ such that for initial data $u_0$ with $\|u_0\|_{H^1} + \|xu_0\|_{L^2} \leq \varepsilon \leq \varepsilon_0$:

• The equation admits a unique global solution $u \in C(\mathbb{R}, H^1(\mathbb{R}))$.
• The solution satisfies the sharp linear decay rate: $\|u(t)\|_{L^\infty_x} \lesssim \varepsilon \langle t \rangle^{-1/2}$.
• These bounds hold for all $t \in [-T_\varepsilon, T_\varepsilon]$ where $T_\varepsilon = \exp(1/c\varepsilon^2)$ (almost global existence).
• The profile $f^\sharp$ satisfies uniform bounds in the weighted norm defined above.

Significance and Claims
The paper claims its main novelty lies in removing additional symmetry assumptions on the potential $V$ that were previously required to handle the zero-energy resonance.

• Removal of Symmetry: Previous results (e.g., Chen and Pusateri [5]) established modified scattering or global bounds only under assumptions that the zero-energy resonance was either odd or even (or had limits at $\pm \infty$ that were equal or opposite). This work proves that such symmetries are not necessary for obtaining almost global quantitative bounds.
• Handling the "Dangerous" p.v. Term: The paper identifies that in the absence of symmetry, a "dangerous" principal value term ($\mu^\sharp_{p.v.}$) appears in the nonlinear spectral distribution. This term prevents the use of standard smoothing estimates that rely on low-frequency vanishing. The author resolves this by introducing a new Fourier restriction inequality, which allows for the control of the $L^2$ norm of the profile derivative despite the lack of low-frequency improvement.
• Optimality of Growth: The analysis reveals that the growth of $\|\partial_k f^\sharp\|_{L^2_k}$ is of order $t^{1/4}$. The paper argues that this growth rate is likely optimal for asymmetric non-generic potentials. It discusses that closing the bootstrap scheme for a global-in-time result with a slower growth rate (e.g., $t^\alpha$ for $\alpha < 1/4$) would fail due to the specific asymptotic behavior of the solution at $x=0$ driven by the non-symmetric resonance. Consequently, the result is presented as "almost global" rather than global, reflecting the limitations imposed by the lack of symmetry.

In summary, the paper extends the understanding of long-time dynamics for NLS with potentials to the general non-generic case, demonstrating that while global modified scattering may not hold without symmetry, sharp decay and quantitative bounds can still be established for exponentially long times using refined spectral and harmonic analysis techniques.