Anderson localized states for the quasi-periodic nonlinear Schrödinger equation on $\mathbb Z^d$

This paper establishes the existence of large sets of Anderson localized states for the quasi-periodic nonlinear Schrödinger equation on $\mathbb Z^d$, thereby extending localization results from linear to nonlinear and from random to deterministic settings through new Diophantine estimates and Bourgain's geometric lemma.

The Gist

Imagine you are trying to keep a perfectly balanced stack of Jenga blocks. If you just leave them alone, they might wobble a little due to the wind (random noise), but they stay put. This is what physicists call Anderson Localization: a state where energy (or a wave) gets "stuck" in one spot and doesn't spread out, even in a messy environment.

For a long time, scientists knew this worked for linear systems (where the blocks don't interact with each other) and for random systems (where the wind blows in unpredictable directions).

This paper, by Yunfeng Shi and Wen-Ming Wang, tackles a much harder problem: What happens when the blocks start talking to each other (nonlinearity) and the wind blows in a very specific, repeating pattern (quasi-periodic)?

Here is the breakdown of their discovery using simple analogies:

1. The Setup: The Messy Room vs. The Patterned Room

• The Linear Case (Old Science): Imagine a room with a floor covered in random bumps. If you roll a marble, it hits a bump and stops. It gets "localized." Scientists already knew this worked.
• The Random Case (Old Science): Imagine the bumps are placed randomly by a drunk person. The marble still gets stuck.
• The New Challenge (This Paper): Now, imagine two things change:
  1. The Marble is Sticky (Nonlinearity): The marble isn't just a passive object; if it gets too big or moves too fast, it changes the shape of the floor it's rolling on. It interacts with itself.
  2. The Floor is Patterned (Quasi-Periodic): Instead of random bumps, the floor has a complex, repeating pattern (like a wallpaper design that never quite repeats itself exactly).

The big question was: If you have a sticky marble on a patterned floor, will it still get stuck, or will it eventually slide all over the room?

2. The Discovery: "Yes, it gets stuck!"

The authors proved that yes, even with the sticky interactions and the complex pattern, there are huge sets of conditions where the wave (the marble) remains trapped in a small area forever. They didn't just prove it for one specific case; they showed it works for a "large set" of scenarios, meaning it's a robust phenomenon, not a fluke.

3. The Tools: How They Solved the Puzzle

To prove this, they had to overcome two massive hurdles. Think of these as the "magic wands" they used:

A. The "Diophantine" Lockpick

In a patterned room, waves can sometimes get confused. They might bounce off a bump and hit another bump at the exact right time to amplify each other, creating a "resonance" that lets the wave escape.

• The Analogy: Imagine trying to walk through a hallway where the doors open and close in a rhythm. If your footsteps match the rhythm perfectly, you get stuck in a loop. If they don't match, you can slip through.
• The Math: The authors had to prove that the "rhythm" of the pattern (the frequencies) and the "rhythm" of the wave are so mathematically different (a concept called Diophantine estimates) that they can never sync up perfectly to let the wave escape. They developed a new way to measure this mismatch, even in high-dimensional spaces (rooms with many dimensions, not just 1 or 2).

B. The "Geometric" Net

The second problem was the sheer size of the room. In high dimensions, there are infinite ways for the wave to wiggle out.

• The Analogy: Imagine trying to catch a slippery fish in a giant, multi-dimensional ocean. You can't just throw a net everywhere; you need to know exactly where the fish can't go.
• The Math: They used a technique called Semi-Algebraic Geometry. Think of this as drawing a map of the "bad zones" (where the wave might escape). They proved that these bad zones are so thin and sparse that they take up almost no space. It's like showing that the "escape routes" in the room are so narrow that the chance of the wave finding one is practically zero.

4. The "Newton" Ladder

To build their proof, they used a method called Newton Iteration.

• The Analogy: Imagine you are trying to balance a tower of blocks. You start with a rough guess. Then you look at where it's wobbling, make a tiny correction, and look again. You do this over and over, getting closer and closer to perfect balance.
• The Math: They started with a simple solution and kept refining it, proving at every step that the "wobble" (the error) gets smaller and smaller, eventually vanishing. This allowed them to construct a perfect, stable solution from a messy starting point.

Why Does This Matter?

This isn't just about math puzzles. It helps us understand:

• Quantum Computers: How to keep quantum information from leaking out (decoherence) in complex, interacting systems.
• Light in Crystals: How to trap light in special materials (photonic crystals) that have complex patterns, which is crucial for new types of lasers and fiber optics.
• The Universe: It bridges the gap between "random" chaos and "ordered" patterns, showing that order can survive even when things get messy and interactive.

In a nutshell: The authors showed that even in a complex, repeating world where things interact with each other, nature still has a way to "lock" energy in place, preventing it from spreading out. They did this by proving that the patterns are too complex to sync up with the waves, and the "escape routes" are too thin to find.

Technical Summary

1. Problem Statement

The paper addresses the persistence of Anderson localized states in the nonlinear discrete Schrödinger equation (DNLS) with a quasi-periodic potential on a $d$-dimensional lattice ($\mathbb{Z}^d$).

The equation under study is:
$$i u_t + (\varepsilon \Delta + V(\theta + n\alpha)\delta_{n,n'})u + \delta |u|^{2p}u = 0$$
where:

• $u(t, n)$ is the wave function.
• $\varepsilon$ and $\delta$ are small parameters representing the hopping strength and nonlinearity, respectively.
• $\Delta$ is the discrete Laplacian.
• $V(\theta)$ is a trigonometric polynomial potential (quasi-periodic).
• $\alpha \in [0, 1]^d$ is the frequency vector, and $\theta \in [0, 1]^d$ is the phase.

Context:

• Linear Case ($\delta=0$): It is well-established (Bourgain [Bou07], Jitomirskaya-Liu-Shi [JLS20]) that for small $\varepsilon$, the linear operator exhibits Anderson localization (pure point spectrum with exponentially decaying eigenfunctions) for a large set of $(\alpha, \theta)$.
• Random Case: For random potentials, Anderson localization persists in the nonlinear regime (Bourgain-Wang [BW08], Liu-Wang [LW24]).
• The Gap: Extending these results from the linear to the nonlinear setting, and from random to deterministic quasi-periodic potentials in arbitrary dimensions ($d \geq 1$, specifically challenging for $d \geq 3$) has been an open problem. The primary difficulty lies in controlling resonances in high-dimensional phase spaces where standard Diophantine conditions are insufficient.

2. Methodology

The authors employ a Craig-Wayne-Bourgain (CWB) approach combined with Lyapunov-Schmidt decomposition and multi-scale analysis (MSA). The proof is structured into three main components:

A. Linear Analysis and Large Deviation Theorems (LDT)

The core of the proof relies on establishing Large Deviation Theorems (LDT) for the Green's functions of the linearized operator. The authors perform a multi-scale analysis in three distinct regimes:

1. Small Scales: Utilizing Neumann series arguments and standard Diophantine estimates.
2. Intermediate Scales: This is a novel contribution. Due to the small variation of frequencies ($O(\delta)$), there is a "gap" where standard Bourgain analysis is not yet applicable. The authors use Diophantine estimates derived from the potential's values at specific lattice sites to control resonances directly in this gap.
3. Large Scales: Applying Bourgain's geometric lemma and semi-algebraic geometry (specifically Cartan's Lemma and Yomdin-Gromov triangulation) to handle resonances in high dimensions.

B. Diophantine Estimates via Generalized Wronskians

A critical technical hurdle is proving that the eigenvalues of the linear operator (which depend on $V(n\alpha + \theta)$) satisfy necessary Diophantine conditions in high dimensions.

• The authors develop a generalized Wronskian approach to prove that the values of the trigonometric polynomial $V$ at distinct lattice sites are linearly independent in a Diophantine sense.
• This involves bounding determinants of matrices formed by derivatives of $V$ and using Lojasiewicz-type inequalities to estimate the measure of resonant sets.
• This allows them to remove a set of $(\alpha, \theta)$ with small measure (logarithmic in $\varepsilon$) where resonances occur.

C. Nonlinear Iteration (Newton Scheme)

The existence of quasi-periodic solutions is constructed via a Newton iteration scheme:

• Lyapunov-Schmidt Decomposition: The equation is split into $P$-equations (infinite-dimensional, solved via Newton iteration) and $Q$-equations (finite-dimensional, solved via the Implicit Function Theorem to adjust frequencies).
• Green's Function Estimates: The convergence of the Newton scheme relies on the exponential decay of the Green's function estimates established in the linear analysis.
• Semi-Algebraic Projection: To handle the dependence of the Green's function on parameters $(\alpha, \theta)$, the authors use semi-algebraic projection lemmas to ensure that the set of "bad" parameters removed at each iteration step has small measure.

3. Key Contributions

1. Extension to Nonlinearity: The paper successfully extends Anderson localization from the linear quasi-periodic setting to the nonlinear setting for the DNLS on $\mathbb{Z}^d$.
2. Extension to Deterministic High Dimensions: It generalizes previous results (which were often limited to $d=1$ or random potentials) to arbitrary dimensions ($d \geq 1$) for deterministic quasi-periodic potentials.
3. New Diophantine Estimates: The authors introduce a generalized Wronskian method to establish Diophantine properties for trigonometric polynomials in arbitrary dimensions, a result of independent interest in number theory and dynamical systems.
4. Intermediate Scale Analysis: The paper identifies and resolves the "intermediate scale" problem in multi-scale analysis where the frequency variation is small ($O(\delta)$), utilizing specific clustering properties of the spectrum.
5. Refined Resonance Control: The use of semi-algebraic geometry and Bourgain's geometric lemma is adapted to handle the specific resonance structures arising in the nonlinear quasi-periodic context, overcoming the lack of eigenvalue separation present in higher-dimensional tori.

4. Main Results

Theorem 1.1 (Main Result):
For any distinct lattice sites $n_1, \dots, n_b \in \mathbb{Z}^d$, there exists a large set of parameters $(\alpha, \theta) \in [0, 1]^{2d}$ (with measure close to 1, specifically $1 - O(\log^{-c}(\varepsilon+\delta)^{-1})$) such that:

• For small $\varepsilon, \delta$ (with $\varepsilon \leq \delta \leq \log^{-1}(1/\varepsilon)$), there exist quasi-periodic in time solutions to the nonlinear equation.
• These solutions are Anderson localized: they are concentrated around the initial sites $n_\ell$ and decay exponentially in space.
• The solutions are of the form $u(t, n) = \sum_{k \in \mathbb{Z}^b} \hat{u}(k, n) e^{i k \cdot \omega t}$, where the frequencies $\omega$ are close to the linear frequencies $V(n_\ell \alpha + \theta)$.
• The set of amplitudes $a$ for which these solutions exist has measure close to full in the amplitude space.

5. Significance

• Theoretical Breakthrough: This work bridges a significant gap in the theory of Anderson localization, demonstrating that the phenomenon is robust against nonlinearity even in deterministic, high-dimensional quasi-periodic environments.
• Methodological Advancement: The combination of semi-algebraic geometry, generalized Wronskians, and refined multi-scale analysis provides a powerful toolkit for tackling other high-dimensional quasi-periodic problems where standard KAM or Diophantine methods fail.
• Physical Relevance: The results support the stability of localized states (solitons) in physical systems modeled by quasi-periodic potentials, such as optical lattices with incommensurate frequencies, suggesting that disorder-induced localization can persist even in the presence of nonlinear interactions.

In summary, Shi and Wang provide a rigorous proof that Anderson localization persists in the nonlinear quasi-periodic DNLS on $\mathbb{Z}^d$, overcoming the formidable challenges of high-dimensional resonances through innovative analytical techniques.