🔢 mathematics

Fourier decay of Gaussian multiplicative chaos and boundary geometry

This paper introduces a new approach to determine the Fourier dimension and decay rates of Gaussian multiplicative chaos on tori and bounded sets, demonstrating how boundary geometry and multifractal concentration jointly govern these properties and providing a simplified proof of the Garban-Vargas conjecture in the circle setting.

Original authors: Marcu-Antone Orsoni, William Verreault

Published 2026-09-11

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Marcu-Antone Orsoni, William Verreault

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

In the world of mathematics, there is a branch called harmonic analysis that studies how complex shapes and patterns can be broken down into simple waves. Imagine trying to describe a jagged mountain range not by drawing its outline, but by adding together a series of smooth, rolling ripples. The speed at which these ripples fade away as they get smaller tells mathematicians something profound about the shape itself: how rough it is, how much space it occupies, and how its mass is distributed. This measure of how quickly the waves disappear is known as the Fourier dimension. It is a way of quantifying the "smoothness" of a shape's edge, even when that shape is incredibly irregular.

For decades, mathematicians have been fascinated by a specific type of random shape called Gaussian multiplicative chaos. These are not solid objects you can hold, but rather random clouds of mass that form when you exponentiate a noisy, fluctuating field. Think of a field that jitters randomly at every point, like static on an old television screen, but with a specific kind of long-range connection where nearby points tend to move together. When you turn up the intensity of this noise, the resulting mass doesn't spread out evenly; instead, it clumps together in a wild, fractal pattern, concentrating heavily on tiny, specific points while leaving vast areas empty. The central question for researchers has been: if you try to analyze this chaotic, clumpy mass using those simple waves, how fast do the waves die out? Does the extreme irregularity of the mass stop the waves from fading, or does the mass still have a hidden order that allows the waves to disappear quickly?

A team of researchers has now answered this question with remarkable precision, revealing that the answer depends on a surprising partnership between the randomness of the mass and the geometry of the container holding it. In their study, they examined these chaotic clouds inside different shapes, ranging from a simple flat torus (which can be thought of as a flat surface where the edges wrap around to meet, like a video game screen) to solid, bounded regions in space with curved or flat edges. They discovered that the rate at which the waves fade is governed by two competing forces. The first is the internal structure of the chaos itself, which naturally limits how fast the waves can disappear based on how much the mass clumps together. The second force comes from the boundary of the shape containing the mass. If the boundary is flat, it acts as a barrier that slows down the fading of the waves, effectively making the shape appear "rougher" to the mathematical analysis than it would otherwise be. However, if the boundary is smoothly curved, it helps the waves fade even faster, sometimes overcoming the clumping of the mass entirely.

The researchers proved that on a flat, wrapping surface like a torus, the waves fade at a rate that perfectly matches the theoretical limit set by the mass's own clumping. This confirmed a long-standing guess made by other mathematicians, showing that the randomness of the chaos is strong enough to saturate the limit. But the story changes when the chaos is confined to a solid shape with a boundary. If that boundary has a flat patch, like the side of a box, the waves cannot fade any faster than a specific speed, regardless of how the mass is distributed inside. This speed is strictly slower than what the mass alone would allow. In this case, the flat edge creates a geometric obstruction that dominates the behavior of the waves.

Conversely, when the researchers looked at shapes with uniformly curved boundaries, like a perfect sphere or an egg, they found that the curvature provides an extra boost. The waves fade faster than they would on a flat surface, and in some high-dimensional cases, the curvature is so effective that it allows the waves to disappear at a rate determined by the geometry rather than the internal clumping of the mass. This means that a curved boundary can actually improve the mathematical "smoothness" of the chaotic mass, counteracting its natural tendency to be rough. The team also showed that if the boundary is curved but has points where it flattens out, the behavior becomes a complex mix of these effects, depending on how sharply the curve changes.

These findings are significant because they resolve a major open problem in the field and provide a unified framework for understanding how geometry and randomness interact. The researchers developed a new method to study these problems, breaking the chaotic mass into a coarse, smooth part and a fine, fluctuating part. By analyzing how these two parts behave separately and then combining them, they were able to calculate the exact rate of decay for the waves in every scenario they considered. They proved that for any smooth, curved boundary, the waves fade at a rate that is the best possible given the constraints of the mass and the shape. For flat boundaries, they proved that the waves are stuck at a slower rate, a limitation imposed by the flatness itself.

The study also clarified the behavior of these chaotic measures when they are restricted to surfaces, such as the skin of a sphere. In these cases, the mass lives only on the surface, and the researchers found that the rate of wave decay is determined by the dimension of that surface and the intensity of the noise. This result connects the behavior of the chaos on a surface to the behavior of the chaos in the space surrounding it, showing that the geometry of the surface dictates the mathematical properties of the mass living on it.

Ultimately, this work demonstrates that the Fourier dimension of a random, chaotic mass is not just a property of the mass itself, but a joint property of the mass and the shape that contains it. The boundary is not a passive container; it actively shapes the mathematical signature of the chaos. A flat edge imposes a strict limit, while a curved edge can enhance the fading of waves. This insight changes how mathematicians understand the relationship between randomness and geometry, showing that even in the most irregular, clumpy systems, the shape of the world around them plays a decisive role in how they behave. The results are rigorous and proven, offering a complete picture of how these chaotic clouds interact with the boundaries of their world, from the simplest flat surfaces to the most complex curved shapes.

Technical Summary: Fourier Decay of Gaussian Multiplicative Chaos and Boundary Geometry

Problem Statement
This paper addresses the quantitative harmonic analysis of Gaussian Multiplicative Chaos (GMC), specifically focusing on the decay of the Fourier transform of these random measures at high frequencies. The central object of study is the subcritical GMC measure $M_\gamma$ , associated with a log-correlated Gaussian field $X$ on a domain $U \subset \mathbb{R}^d$ or the torus $\mathbb{T}^d$ . The primary question is whether $M_\gamma$ is a Rajchman measure (i.e., its Fourier transform vanishes at infinity) and, more precisely, determining its Fourier dimension ( $\dim_F$ ).

The Fourier dimension quantifies the strongest uniform polynomial decay rate of the Fourier transform. A fundamental tension exists in this problem: the local mass of $M_\gamma$ is governed by the "thick points" of the underlying field, leading to a multifractal concentration that imposes an upper bound on the decay rate, known as the correlation dimension $D_{\gamma,d}$ . However, the global oscillatory nature of the measure may induce additional cancellation. The paper investigates how the geometry of the domain's boundary interacts with this multifractal concentration to determine the exact Fourier dimension.

Methodology
The authors develop a new approach based on a finite-range reduction of the Gaussian field. Key methodological components include:

Scale Decomposition: The Gaussian field $X$ is decomposed into a "coarse" component ( $X_{>\delta}$ ) and a "fine" component ( $X_{\le\delta}$ ) using a smooth, compactly supported, positive-definite kernel. This construction ensures that the fine field has independent increments over disjoint spatial regions separated by a distance greater than the scale parameter $\delta$ .
Coarse-Fine Splitting: The chaos measure is decomposed as $M_\gamma = M_{>\delta} + \tilde{M}_{\le\delta}$ $M_{γ} = M_{> δ} + \tilde{M}_{\leq δ}$ .
- The coarse term ( $M_{>\delta}$ ) is smooth. Its contribution to the Fourier transform is analyzed via repeated integration by parts. In Euclidean settings, this converts bulk oscillation into boundary integrals.
- The fine term ( $\tilde{M}_{\le\delta}$ ) is treated conditionally on the coarse field. Due to the finite-range property, the fine field restricted to a partition of space into cubes of side $\delta$ yields conditionally independent families.
Probabilistic Estimates: The authors employ conditional martingale inequalities (specifically Lemma 2.7) to bound the sums of these independent families in weighted Fourier norms. They utilize local moment bounds for GMC (Proposition 2.2) and Kahane's convexity inequality to control the moments of the fine fluctuations.
Geometric Analysis: For domains with boundaries, the integration by parts of the coarse term generates oscillatory surface integrals. The decay of these integrals depends on the curvature of the boundary. The authors use stationary phase arguments and the geometry of supporting caps to derive sharp upper and lower bounds.

Key Contributions and Results

Torus Case (Theorem 1.1):
The paper provides a new proof for the Fourier dimension of GMC on the torus $\mathbb{T}^d$ for every smooth log-correlated covariance. It establishes that almost surely:
$\dim_F(M_\gamma) = D_{\gamma,d}$
where $D_{\gamma,d}$ is the correlation dimension. This confirms the Garban–Vargas conjecture for the circle and extends it to all dimensions and smooth covariances. The proof is noted to be substantially simpler than previous vector-valued martingale approaches.
Euclidean Restrictions and Finite Perimeter (Theorem 1.2 & Corollary 1.3):
For a bounded set $A \subset \mathbb{R}^d$ with finite perimeter, the Fourier dimension satisfies:
$\dim_F(M_{\gamma,A}) \ge \min\{2, D_{\gamma,d}\}$
This lower bound is sharp when the boundary contains a flat open patch (Theorem 1.5). The paper also proves that smooth localization (multiplication by a smooth compactly supported function) preserves the full dimension $D_{\gamma,d}$ , and that any measure absolutely continuous with respect to $M_\gamma$ is Rajchman.
Boundary Geometry Effects (Theorems 1.5, 1.6, 1.7):
The most significant contribution is the demonstration that boundary geometry can strictly reduce the Fourier dimension below the correlation dimension $D_{\gamma,d}$ , even though the boundary itself carries zero mass almost surely.
- Flat Boundaries: If the boundary contains a flat open patch, the Fourier dimension is exactly $\min\{2, D_{\gamma,d}\}$ . The exponent 2 arises from the planar nature of the boundary, which limits the cancellation of the coarse term.
- Strictly Convex Boundaries: For uniformly strictly convex $C^\infty$ domains, the curvature induces additional tangential cancellation. The Fourier dimension is given by:
  $\dim_F(M_{\gamma,D}) = \min\{D_{\gamma,d}, 2 + \alpha_{\gamma, d-1}\}$
  where $\alpha_{\gamma, p} = (\sqrt{p} - \gamma\sqrt{2})^2$ . This geometric term is active only in high dimensions ( $d \ge 5$ ) or specific parameter regimes, showing that curvature can further lower the dimension.
- Surface Chaos: For GMC constructed directly on a uniformly strictly convex hypersurface $S$ , the Fourier dimension is exactly $\alpha_{\gamma, d-1}$ . This highlights a discrepancy between the "angular" Fourier dimension (on the manifold) and the "Euclidean" Fourier dimension (in the ambient space).

Significance and Claims
The paper claims to resolve the Fourier dimension problem for GMC on tori and to provide the first sharp formulas for GMC restricted to Euclidean domains with specific boundary geometries.

Resolution of Conjectures: It offers a simpler proof of the Garban–Vargas conjecture for the circle and generalizes it to all smooth covariances on tori.
Geometric Obstruction: The work highlights a novel phenomenon where the geometry of the boundary acts as an obstruction to Fourier decay. The authors show that the Fourier dimension is governed by the interplay between the multifractal concentration of the mass (bulk obstruction) and the oscillatory geometry of the support (boundary obstruction).
Sharpness: The results are sharp. The paper demonstrates that for domains with flat patches, the dimension is strictly limited to 2 (if $D_{\gamma,d} > 2$ ), and for strictly convex domains, the dimension is further reduced by a term dependent on the curvature and the parameter $\gamma$ .
Methodological Robustness: The authors note that their finite-range reduction and conditional independence arguments are robust and can be adapted to other settings, such as critical or imaginary GMC, though these are left for future work.

The paper does not claim to solve the problem for all possible boundary geometries (e.g., those with vanishing curvature at isolated points) but provides a conjectural framework for such cases based on the order of contact with supporting hyperplanes. It emphasizes that the Fourier dimension depends not just on the distribution of mass but on the oscillatory geometry of the support.

Technical Summary: Fourier Decay of Gaussian Multiplicative Chaos and Boundary Geometry

More like this