🔢 mathematics

Optimal Covariance Estimates for Schrödinger Semigroups with White Noise in $d=1,2$

This paper establishes optimal asymptotic bounds on the covariance of the trace of Schrödinger semigroups perturbed by white noise in dimensions one and two using Feynman-Kac formulas and Brownian bridge local times, thereby improving existing results for $d=1$ , providing the first such estimates for $d=2$ , and applying these findings to prove quantitative hyperuniformity and decorrelation properties.

Original authors: Youssef Djellouli, Pierre Yves Gaudreau Lamarre

Published 2026-07-21

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Youssef Djellouli, Pierre Yves Gaudreau Lamarre

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the universe as a giant, invisible trampoline. If you place a heavy bowling ball in the center, the fabric curves down, creating a smooth valley. This is how physicists often think about gravity: a smooth, predictable curve. But what if the trampoline wasn't smooth at all? What if, instead of a perfect curve, it was covered in millions of tiny, invisible pebbles, each one jiggling randomly? A marble rolling across this surface wouldn't just follow a smooth path; it would bounce, jitter, and get stuck in unpredictable ways. This is the world of "quantum particles" moving through "disorder."

In the world of quantum mechanics, particles like electrons don't just sit still; they dance to the rhythm of energy levels. Usually, we can predict these levels pretty well if the environment is calm. But in the real world, materials are messy. They have impurities, defects, and random noise. Scientists model this mess using something called "white noise." Think of white noise not as the sound of a radio static, but as a mathematical storm where every single point in space has a random, jiggling value that changes instantly. It's the ultimate chaos. The big question for mathematicians and physicists is: when you throw a quantum particle into this chaotic storm, how do its energy levels behave? Do they clump together? Do they spread out? And if you look at two different moments in time, how much do their behaviors influence each other?

This paper is a deep dive into that exact question, but with a very specific and tricky twist. The authors, Youssef Djellouli and Pierre Yves Gaudreau Lamarre, are looking at a mathematical object called a "Schrödinger operator." You can think of this as the master equation that tells us how our quantum particle moves. They are studying what happens when this particle is trapped in a potential well (like a valley) but is also being bombarded by that chaotic white noise. The challenge is that white noise is so wild and jagged that it breaks standard math tools, especially in higher dimensions. The authors focus on one and two dimensions (like a flat line or a flat sheet), because in three dimensions, the noise is so strong that the math simply explodes and becomes impossible to calculate.

The main discovery of this paper is a precise recipe for how "connected" the energy levels of this particle are at different times. Imagine you take a snapshot of the particle's energy at time $s$ and another at time $t$ . The authors calculated exactly how much these two snapshots "talk" to each other. They found that the connection (or covariance) depends on two things: how close the times $s$ and $t$ are to zero, and how steep the walls of the valley are. If the valley walls are steep (a strong "confining potential"), the particle is forced to stay in a small area, so it keeps bumping into the same random noise, making the energy levels very correlated. If the valley is wide, the particle can roam, and the connection fades faster.

The paper proves that this relationship follows a very specific mathematical rule involving the dimensions of space and the steepness of the valley. They didn't just guess this; they built a rigorous bridge between the chaotic noise and the particle's behavior using a clever tool called the "Feynman-Kac formula." This formula is like a translator that turns the wild, abstract math of quantum noise into the language of "Brownian bridges"—which are just random walks that start and end at the same point, like a drunkard stumbling out of a bar and stumbling right back into it. By studying how often these random walks cross themselves or each other (a concept called "local time"), the authors could calculate the exact strength of the connection between the energy levels.

One of the most exciting parts of their finding is that they proved their formula is the best possible answer. They didn't just find an upper limit; they found the exact rate at which the connection fades. This allows them to prove a property called "hyperuniformity." In plain English, this means that even though the particle is in a chaotic, noisy environment, its energy levels are surprisingly organized. They don't fluctuate as wildly as you would expect from pure randomness. It's as if the chaos of the noise and the order of the valley reach a perfect balance, creating a structure that is "rigid" on a large scale, even if it looks messy up close.

The authors also looked at how this changes as the dimension changes. In one dimension (a line), the connection between energy levels fades at a certain speed. In two dimensions (a flat sheet), it fades at a different, faster speed. This makes sense intuitively: in a larger space, the particle has more room to escape the specific "noise" it encountered at the first moment, so the two moments become less related. The paper confirms this with hard math, showing that the dimension of the space is a key player in how quickly the particle "forgets" its past.

Why does this matter? While this might sound like pure theory, understanding how particles behave in noisy environments is crucial for designing new materials, like better semiconductors or quantum computers. If we can predict how energy levels fluctuate and correlate, we can build devices that are more stable and less prone to errors caused by impurities. This paper gives scientists a new, sharper tool to predict these fluctuations, moving from vague estimates to precise, optimal bounds. It's a step forward in taming the chaos of the quantum world, showing us that even in a storm of white noise, there is a hidden, calculable rhythm to the dance of particles.

Technical Summary: Optimal Covariance Estimates for Schrödinger Semigroups with White Noise in $d=1, 2$

Problem Statement
This paper investigates the statistical properties of the eigenvalues of the random Schrödinger operator $H = -\frac{1}{2}\Delta + V + \xi$ acting on $L^2(\mathbb{R}^d)$ , where $d \in \{1, 2\}$ . Here, $\xi$ represents a standard Gaussian white noise (a Schwartz distribution), and $V$ is a deterministic potential exhibiting power-law growth at infinity ( $V(x) \geq a|x|^\kappa - b$ ). The primary focus is on the fluctuations and correlations of the large eigenvalues of $H$ , specifically analyzed through the covariance of the exponential trace (semigroup) $C(s, t) = \text{Cov}(\text{Tr}[e^{-sH}], \text{Tr}[e^{-tH}])$ as $s, t \to 0$ .

The study is motivated by the physical interpretation of $H$ as the energy levels of a quantum particle in a confining potential $V$ subject to unstructured disorder $\xi$ . While white noise is the universal scaling limit of lattice noises, its high singularity makes the rigorous definition of $H$ and the analysis of its spectral properties technically challenging, particularly for $d \geq 2$ . Previous literature established bounds on the variance of the trace for $d=1$ , but these were suboptimal for $s \neq t$ , and no such estimates existed for $d=2$ with general confining potentials.

Methodology
The authors employ a probabilistic approach centered on the Feynman-Kac formula to represent the trace and its covariance. The methodology involves several key technical steps:

Construction of the Trace:
- For $d=1$ , the operator $H$ is constructed via its quadratic form, interpreting integrals against $\xi$ as stochastic integrals. The trace is well-defined as a sum of exponentials of eigenvalues.
- For $d=2$ , the operator requires renormalization due to the singularity of white noise. Instead of constructing the full operator via regularity structures or paracontrolled calculus, the authors construct the trace observable $\text{Tr}[e^{-tH}]$ directly as an $L^2$ limit of renormalized traces of smoothed operators $H_\epsilon = -\frac{1}{2}\Delta + V + \xi_\epsilon - c_\epsilon$ , where $\xi_\epsilon$ is a mollified noise and $c_\epsilon$ is a renormalization constant.
Feynman-Kac Representation:
The expectation and covariance of the trace are expressed in terms of Brownian bridges ( $B^{x,x}_t$ ).
- The expectation involves the self-intersection local time (SILT) of a single bridge.
- The covariance involves the mutual intersection local time (MILT) between two independent bridges, $B^{x,x}_t$ and $B^{y,y}_s$ , as well as their individual SILTs.
- For $d=2$ , the SILT and MILT are defined via limits of approximate local times involving the Gaussian kernel, requiring careful handling of divergences (subtracting the mean for SILT).
Asymptotic Analysis:
The core of the analysis involves deriving precise asymptotic upper and lower bounds for the covariance $C(s, t)$ as $s, t \to 0$ . This is achieved by:
- Estimating the moments of the MILT and SILT using scaling properties of Brownian bridges.
- Controlling the interaction between the potential $V$ (via the term $e^{-\int V}$ ) and the noise terms (via the intersection local times).
- Utilizing Hölder's inequality and uniform integrability estimates to separate the dominant terms in the covariance integral.

Key Contributions and Results

Optimal Covariance Bounds:
The main result (Theorem 1.4) establishes that under the growth assumption $V(x) \geq a|x|^\kappa - b$ , the covariance satisfies:
$C(s, t) \lesssim \min(s, t)^{1 - d/2} \max(s, t)^{1 - d/2 - d/\kappa}$
as $s, t \to 0$ . Furthermore, if $V$ also satisfies an upper bound $V(x) \leq c|x|^\kappa + f$ , this estimate is shown to be optimal (i.e., a matching lower bound holds).
- For $d=1$ : This significantly improves upon previous bounds (e.g., from [LGL20] and [GL25]), particularly for the case $s \neq t$ , where earlier results provided weaker powers.
- For $d=2$ : This provides the first estimates of their kind for Schrödinger operators with white noise and general power-law potentials.
Semigroup Hyperuniformity:
Applying these bounds, the authors prove a "semigroup hyperuniformity" property (Corollary 1.10). They define a hyperuniformity ratio $R_s(t) = \text{Var}(\text{Tr}[e^{-tH}]) / \mathbb{E}[\text{Tr}[e^{-tH}]]$ and show that if the potential satisfies both growth bounds, $R_s(t) \asymp t^{2 - d/2}$ as $t \to 0$ . This implies that the fluctuations of the trace vanish faster than those of a Poisson process with comparable intensity, quantifying a form of "global rigidity" in the spectrum.
Decorrelation Rates:
The paper derives a precise rate for the decorrelation of the trace at different times (Corollary 1.12). The correlation coefficient $\rho(s, t)$ decays as:
$\rho(s, t) \lesssim \left( \frac{\min(s, t)}{\max(s, t)} \right)^{d/2\kappa}$
This quantifies how eigenvalue correlations depend on the dimension $d$ and the potential growth rate $\kappa$ .

Significance and Claims
The paper claims that its results represent a significant improvement in the understanding of random Schrödinger operators with white noise:

Technical Advancement: It provides the first optimal covariance estimates for $d=2$ and improves the precision of estimates for $d=1$ where $s \neq t$ .
Methodological Innovation: By circumventing the need for full operator construction in $d=2$ and focusing directly on the trace via $L^2$ limits and Feynman-Kac formulas, the authors bypass the heavy machinery of regularity structures for this specific spectral problem.
Physical Insight: The results quantify the competition between the spreading of eigenfunctions (driven by the Laplacian) and their concentration in low-potential/noise regions (driven by $V$ and $\xi$ ). The derived decorrelation rates reflect how higher dimensions allow eigenfunctions to separate more easily, reducing correlations, while stronger confining potentials force them to "reuse" the same noise, increasing correlations.
Foundational for Future Work: The authors note that these matching upper and lower bounds are necessary to prove conjectures regarding half-line hyperuniformity ( $R_h(\lambda) \to 0$ ) and eigenvalue counting correlations, which remain open problems for $d=1, 2, 3$ .

The paper does not propose new experimental setups or applications beyond the mathematical analysis of the spectral statistics of these operators, nor does it claim to solve the construction of the operator in $d \geq 4$ . Its scope is strictly limited to establishing rigorous asymptotic bounds on the covariance of the semigroup trace in dimensions 1 and 2.

More like this