The non-covered set in Dvoretzky covering is a set of multiplicity

Basic Information

• Paper ID: 2511.09726
• Title: The non-covered set in Dvoretzky covering is a set of multiplicity
• Author: Mingjie Tan (LAREMA, UMR 6093 CNRS, Université d'Angers)
• Classification: math.PR (Probability Theory)
• Submission Date: November 12, 2024
• Paper Link: https://arxiv.org/abs/2511.09726v1

Abstract

This paper proves that the non-covered set in Dvoretzky random covering is a set of multiplicity (M-set) by demonstrating that its natural multiplicative chaos measure is a Rajchman measure.

Research Background and Motivation

Research Problem

This paper investigates the harmonic analytic properties of the non-covered set in the Dvoretzky random covering problem. The core question is: Is the non-covered set $E = \mathbb{T} \setminus (\bigcup_{n=1}^{\infty} I_n)$ a set of multiplicity (M-set)?

Importance of the Problem

1. Extension of classical covering problems: Dvoretzky's random covering problem, proposed in 1956, is a classical problem in probability theory. Shepp provided necessary and sufficient conditions for complete coverage of the circle in 1972. This paper investigates the non-covered set from a harmonic analysis perspective, opening new research directions.
2. Fundamental problem in harmonic analysis: Determining whether a set is a uniqueness set (U-set) or a set of multiplicity (M-set) is a central problem in harmonic analysis. This concerns the uniqueness theory of trigonometric series and has profound theoretical significance.
3. Connection to multiplicative chaos theory: This paper links the Dvoretzky covering problem to multiplicative chaos measure theory, providing new tools for understanding the asymptotic behavior of Fourier coefficients of such random measures.

Limitations of Existing Research

1. Kahane (1985) studied the Hausdorff dimension of finite covered sets, particularly when $\ell_n = \alpha/n$, obtaining $\dim_H F = 1-\alpha$
2. The harmonic analytic properties (U-set or M-set) of the non-covered set $E$ had not been previously studied
3. Existing research on Fourier coefficients of multiplicative chaos measures has focused mainly on Gaussian Multiplicative Chaos, with limited study of Dvoretzky measures

Research Motivation

Inspired by Garban and Vargas (2024) on Gaussian Multiplicative Chaos being Rajchman measures, the author seeks to prove that Dvoretzky measures possess similar properties, thereby establishing the multiplicity of the non-covered set.

Core Contributions

1. Main Theorem: Proves that under appropriate conditions, the d-fold convolution of the Dvoretzky measure $\mu_D$ is almost surely absolutely continuous (Theorem 1.1)
2. Harmonic Analysis Characterization: Establishes that the non-covered set $E$ is an $M_0$-set (Corollary 1.2), meaning there exists a non-zero Rajchman measure supported on $E$
3. Concrete Examples: For the classical case $\ell_n = \alpha/n$ ($\alpha \in (0,1)$), proves that $E$ is almost surely an $M_0$-set; particularly, when $0 < \alpha \leq 1/2$, $\mu_D$ itself is a Rajchman measure
4. Methodological Contribution: Successfully applies the convolution method of Garban-Vargas for handling Gaussian Multiplicative Chaos to the Dvoretzky covering problem

Detailed Methodology

Problem Setup

Dvoretzky Random Covering:

• Let $\{\omega_n\}_{n\geq 1}$ be a sequence of independent identically distributed random variables, uniformly distributed on the circle $\mathbb{T} = \mathbb{R}/\mathbb{Z} \cong [0,1)$
• Let $\{\ell_n\}_{n\geq 1}$ be a positive sequence decreasing to zero
• Random intervals: $I_n := (\omega_n, \omega_n + \ell_n) \mod 1$
• Non-covered set: $E := \mathbb{T} \setminus (\bigcup_{n=1}^{\infty} I_n)$

Construction of Dvoretzky Measure: For any $t \in [0,1)$ and $n \geq 1$, define: $P_n(t) := \frac{1 - \mathbf{1}_{(0,\ell_n)}(t-\omega_n)}{1-\ell_n}, \quad M_n(t) := \prod_{k=1}^n P_k(t)$

Measure sequence: $\mu_n(dt) := M_n(t)dt$

Kahane proved that $\mu_n$ converges almost surely weakly to a random measure $\mu_D$ (Dvoretzky measure).

Key Assumptions

The main results require three assumptions:

(A) Non-covered set has zero measure: $\sum_{n=1}^{\infty} \ell_n = \infty$ Combined with the Borel-Cantelli lemma, this ensures that $E$ almost surely has zero Lebesgue measure.

(B) Non-covered set is non-empty: $\sum_{n=1}^{\infty} (\ell_n - \ell_{n+1}) \exp\left(\sum_{k=1}^n \ell_k\right) < \infty$ This violates Shepp's complete coverage condition (1.1), ensuring that $E$ is almost surely non-empty.

(C) Convolution absolute continuity condition: There exists an integer $d \geq 1$ such that $\int_{\mathbb{T}^d} K(t_1) \cdots K(t_d) K\left(-\sum_{i=1}^d t_i\right) dt_1 \cdots dt_d < \infty$ where $K(t) := \exp(\sum_{n=1}^{\infty} (\ell_n - |t|)_+)$.

Core Technical Strategy

Step 1: Convolution Strategy Direct estimation of Fourier coefficients $\hat{\mu}_D(n)$ is extremely difficult. The core idea is to consider the d-fold convolution of $\mu_D$: $\ast^d \mu_D(dt) = \mu_D \ast \mu_D \ast \cdots \ast \mu_D(dt)$

If one can prove that $\ast^d \mu_D$ is almost surely absolutely continuous, then by the Riemann-Lebesgue lemma, its Fourier coefficients tend to 0, making $\mu_D$ a Rajchman measure.

Step 2: Convergence of Density Functions For the d-fold convolution of $\mu_n$, the density is: $\ast^d M_n(t) = \int_{\mathbb{T}^d} M_n(t_1) \cdots M_n(t_d) M_n\left(t - \sum_{i=1}^d t_i\right) dt_1 \cdots dt_d$

Introduce notation: $F_n(\hat{t}, t) := M_n(t_1) \cdots M_n(t_d) M_n(t_{d+1}), \quad t_{d+1} = t - \sum_{i=1}^d t_i$

Step 3: Probabilistic Estimates for $L^1$ Convergence Using Lemma 2.1, one needs to prove: $\lim_{n,m\to\infty} \mathbb{E}\left[\|\ast^d M_n - \ast^d M_m\|_1^{1/2}\right] = 0$

This decomposes into two propositions:

Proposition 2.3 (Small Scale Control): For any $\delta > 0$, define $A_\delta(t) = \bigcup_{1\leq i \neq j \leq d+1} \{(\hat{t}) \in \mathbb{T}^d : \|t_i - t_j\| \leq \delta\}$

One needs to prove: $\lim_{\delta \to 0^+} \sup_{n\geq 1} \mathbb{E}\left[\left(\int_{\mathbb{T}} \int_{A_\delta(t)} F_n(\hat{t}, t) d\hat{t} dt\right)^{1/2}\right] = 0$

Proof Outline:

• Use Fan's (1995) result: $\mu_n(\mathbb{T})$ converges in all $L^p$ spaces
• Apply the key estimate (Lemma 2.2): $\frac{1}{C} K_n(t-t') \leq \mathbb{E}[M_n(t)M_n(t')] \leq C K_n(t-t')$
• When $\|t_i - t_j\| \leq \delta$, the integral can be controlled by $K(t_i - t_j)$, which can be made arbitrarily small by assumption (C)

Proposition 2.4 ($L^2$ Convergence): Prove $\lim_{n,m\to\infty} \mathbb{E}\left[\left(\int_{\mathbb{T}} \left|\int_{A_\delta(0)^c} [F_n(\hat{t}, 0) - F_m(\hat{t}, 0)] d\hat{t}\right| dt\right)^{1/2}\right] = 0$

Proof Strategy:

1. Translation Invariance (Lemma 2.5): Prove that the expectation is independent of $t$, so one only needs to consider $t=0$
2. $L^2$ Boundedness (Lemma 2.7):
   • Decompose $\mathbb{T}^d$ into at most $K = [2/\delta] + 1$ subintervals of length at most $\delta/2$
   • For $(t_1, \ldots, t_d) \in A_\delta(0)^c$, there exist $d+1$ disjoint subintervals $B_{j_1}, \ldots, B_{j_{d+1}}$ such that $t_i \in B_{j_i}$ with mutual distance at least $\delta/2$
   • Use independence and estimate (2.12): $\mathbb{E}[F_n(\hat{t}, 0)F_n(\hat{t}', 0)] \leq C' K(t_1-t_1') \cdots K(t_{d+1}-t'_{d+1})$
   • Obtain $L^2$ boundedness from assumption (C)
3. $L^2$ Convergence (Lemma 2.8):
   • Define the limit function: $H(\hat{t}, \hat{t}') := \prod_{k=1}^{\infty} \mathbb{E}\left[\prod_{i=1}^{d+1} P_k(t_i)P_k(t_i')\right]$
   • Region decomposition: $S_\eta := \{(\hat{t}, \hat{t}') : \|t_i - t_i'\| \leq \eta \text{ for some } i\}$
   • On $S_\eta$: By dominated convergence theorem, the integral tends to 0
   • On $S_\eta^c$: When $\ell_k < \eta$, the intervals $(t_i - \ell_k, t_i)$ and $(t_i' - \ell_k, t_i')$ are disjoint, so the expectation is independent of $t, t'$, and thus $\mathbb{E}[F_n(\hat{t}, 0)F_n(\hat{t}', 0)]$ converges uniformly to $H(\hat{t}, \hat{t}')$

Technical Innovations

1. Adaptation of Convolution Method: Successfully applies the convolution technique of Garban-Vargas for Gaussian fields to the discrete Dvoretzky covering, overcoming the essential structural differences between the two
2. Refined Region Decomposition: Through clever decomposition of $A_\delta(t)$ and $S_\eta$, separates the integral into "points close" and "points separated" cases handled separately
3. Full Utilization of Independence: In the $S_\eta^c$ region, when $\ell_k$ is sufficiently small, different random intervals barely overlap, making the expectation value have a simple product structure
4. Establishment of Key Estimates: Lemma 2.2 establishes a two-sided estimate between $\mathbb{E}[M_n(t)M_n(t')]$ and the kernel function $K_n(t-t')$, which is the cornerstone of the entire proof

Theoretical Results

Main Results

Theorem 1.1 (Main Theorem): Assume the sequence $(\ell_n)$ satisfies conditions (A), (B), and (C). Then the d-fold convolution $\ast^d \mu_D$ of $\mu_D$ is almost surely absolutely continuous.

Corollary 1.2: Under the assumptions of Theorem 1.1, $\mu_D$ is a Rajchman measure, and the non-covered set $E$ is an $M_0$-set.

Concrete Applications

Classical Case: $\ell_n = \alpha/n$, $\alpha \in (0,1)$

1. Verification of Conditions:
   • (A): $\sum_{n=1}^{\infty} \alpha/n = \infty$ ✓
   • (B): Can be verified to hold ✓
   • (C): Holds when $d \geq \alpha/(1-\alpha)$ (see GV24, Lemma 2.3)
2. Special Cases:
   • When $0 < \alpha \leq 1/2$, one can take $d=1$
   • In this case, $\mu_D$ itself (rather than its convolution) is a Rajchman measure
   • This is because when $d=1$, condition (C) becomes: $\int_{\mathbb{T}} K(t)K(-t) dt = \int_{\mathbb{T}} K^2(t) dt < \infty$ which converges when $\alpha \leq 1/2$
3. Comparison with Hausdorff Dimension:
   • Kahane proved that the Hausdorff dimension of the finite covered set $F$ is $\dim_H F = 1-\alpha$
   • The non-covered set $E \subset F$ satisfies $\dim_H E \leq 1-\alpha$
   • This paper proves that although $E$ may have small dimension, it still supports a non-zero Rajchman measure

Theoretical Significance

1. Harmonic Analysis Perspective: First characterizes the non-covered set of Dvoretzky covering from a harmonic analysis viewpoint, complementing existing measure-theoretic and dimensional studies
2. Multiplicative Chaos Theory: Provides a complete proof of the Rajchman property of Dvoretzky measures as a class of multiplicative chaos measures
3. Universality of Methods: The proof techniques may be applicable to other types of multiplicative chaos measures

Related Work

Dvoretzky Covering Problem

1. Dvoretzky (1956): Proposed the random covering problem
2. Shepp (1972): Provided necessary and sufficient conditions for complete coverage (1.1)
3. Kahane (1985): Studied the Hausdorff dimension of finite covered sets and the construction of Dvoretzky measures
4. Fan (1989, 1991, 1995): Established orthogonality, equivalence, and $L^p$ convergence of measures

Uniqueness and Multiplicity Sets

1. Kahane & Lemarié-Rieusset (1995): Systematic treatment of uniqueness theory for trigonometric series
2. Lyons (2020): 70-year survey of research on Rajchman measures
3. Kechris & Louveau (1985): Descriptive set theory and structure of uniqueness sets

Multiplicative Chaos Measures

1. Kahane (1985): Introduced multiplicative chaos theory
2. Berestycki (2017): Elementary methods for Gaussian Multiplicative Chaos
3. Garban & Vargas (2024): Proved GMC is a Rajchman measure; this paper directly borrows their convolution method

Innovation of This Paper

• First establishes the connection between Dvoretzky covering and harmonic analysis
• Successfully transfers GMC techniques to discrete multiplicative chaos
• Provides a complete proof that the non-covered set is almost surely an $M_0$-set

Conclusions and Discussion

Main Conclusions

1. Core Theorem: Under appropriate assumptions, the d-fold convolution of the Dvoretzky measure $\mu_D$ is almost surely absolutely continuous
2. Harmonic Analysis Characterization: The non-covered set $E$ is almost surely an $M_0$-set, meaning there exists a non-zero Rajchman measure supported on $E$
3. Classical Example: For $\ell_n = \alpha/n$ ($\alpha \in (0,1)$), $E$ is almost surely an $M_0$-set; when $\alpha \leq 1/2$, $\mu_D$ itself is a Rajchman measure

Limitations

1. Restrictiveness of Conditions:
   • Assumption (C) requires the d-fold convolution of the kernel function $K$ to be integrable, which restricts the decay rate of $\ell_n$
   • For $\ell_n = \alpha/n$ with $\alpha > 1/2$, one needs $d \geq 2$, preventing direct proof that $\mu_D$ is a Rajchman measure
2. Limitations of Method:
   • Although the convolution method is effective, it requires strong integrability conditions
   • Whether more direct methods for estimating Fourier coefficients exist remains an open question
3. Generality:
   • Results mainly apply to $\{\ell_n\}$ satisfying specific decay conditions
   • Verification of condition (C) may be difficult for more general sequences

Future Directions

1. Generalization to More General Sequences: Study broader classes of $\{\ell_n\}$, relaxing assumption (C)
2. Direct Estimation of Fourier Coefficients: Seek methods to directly prove $\hat{\mu}_D(n) \to 0$ without relying on convolutions
3. Precise Asymptotic Behavior: Study the precise decay rate of $\hat{\mu}_D(n)$
4. Higher-Dimensional Generalization: Extend results to random coverings in higher-dimensional spaces
5. Connection to Other Multiplicative Chaos: Explore deeper connections between Dvoretzky measures and other multiplicative chaos measures such as GMC

In-Depth Evaluation

Strengths

1. Theoretical Innovation:
   • First investigates the Dvoretzky covering problem from a harmonic analysis perspective, opening new research directions
   • Establishes profound connections between probability theory (random covering) and harmonic analysis (uniqueness theory)
   • Employs ingenious proof techniques, successfully adapting the Garban-Vargas method
2. Mathematical Rigor:
   • Complete and rigorous proofs with clear logical flow
   • Careful statements and proofs of key lemmas (e.g., Lemmas 2.2, 2.5, 2.7, 2.9)
   • Excellent handling of technical difficulties (e.g., region decomposition, uniform convergence)
3. Depth of Results:
   • Reveals that although the non-covered set has zero Lebesgue measure, it still possesses rich harmonic analytic structure
   • Proves that $E$ is an $M_0$-set, which is stronger than merely being an M-set
   • Provides complete characterization for the classical example $\ell_n = \alpha/n$
4. Writing Quality:
   • Clear paper structure with well-motivated introduction
   • Detailed explanation of proof strategies, facilitating reader comprehension
   • Good balance between technical details and intuitive explanations

Weaknesses

1. Non-optimality of Conditions:
   • Assumption (C) may not be the weakest condition, particularly for $\alpha > 1/2$ requiring $d \geq 2$
   • Can one also prove that $\mu_D$ itself ($d=1$) is a Rajchman measure when $\alpha > 1/2$?
2. Indirectness of Method:
   • Proves Rajchman property through absolute continuity of convolutions, unable to provide precise estimates of Fourier coefficients
   • Lacks information on the decay rate of $\hat{\mu}_D(n)$
3. Limited Examples:
   • Main example is only $\ell_n = \alpha/n$
   • Lacks concrete discussion of other sequences satisfying conditions (A)(B)(C)
4. Comparison with Existing Results:
   • Could discuss more thoroughly the relationship with Kahane's results on finite covered sets
   • Could explore more deeply the similarities and differences with GMC's Rajchman property

Impact

1. Theoretical Contribution:
   • Adds a harmonic analysis dimension to Dvoretzky covering theory
   • Enriches the theory of multiplicative chaos measures
   • Provides new probabilistic constructions for uniqueness set theory
2. Methodological Value:
   • Further validates the effectiveness of convolution methods in handling multiplicative chaos measures
   • Region decomposition and uniform convergence analysis techniques have universal applicability
3. Potential Applications:
   • May inspire harmonic analysis research on other random covering models
   • Techniques may apply to other types of random measures
4. Reproducibility:
   • Proofs are based entirely on theoretical analysis, involving no numerical computation
   • Main results can be independently verified by expert readers
   • Dependent existing results (e.g., Fan's $L^p$ convergence) have clear citations

Applicable Scenarios

1. Theoretical Research:
   • Random covering problems in probability theory
   • Uniqueness theory in harmonic analysis
   • Multiplicative chaos measure theory
2. Related Fields:
   • Fractal geometry (geometric structure of non-covered sets)
   • Measure theory (Fourier analysis of singular measures)
   • Stochastic processes (martingale convergence theory)
3. Potential Generalizations:
   • Higher-dimensional random coverings
   • Other types of random sets
   • More general multiplicative chaos systems

Key References

1. Shepp (1972): "Covering the circle with random arcs" - Provides necessary and sufficient conditions for complete coverage
2. Kahane (1985b): "Some random series of functions" - Systematic study of Dvoretzky measures
3. Garban & Vargas (2024): "Harmonic analysis of Gaussian multiplicative chaos on the circle" - Main methodological reference for this paper
4. Fan (1995): "Sur la $L^p$ convergence des martingales liées au recouvrement" - Key result on $L^p$ convergence
5. Kahane & Lemarié-Rieusset (1995): "Fourier Series and Wavelets" - Systematic treatment of uniqueness theory

Overall Assessment: This is a high-quality theoretical mathematics paper that successfully bridges Dvoretzky random covering and harmonic analysis. The main results are novel and profound, with ingenious and rigorous proof techniques. Although there is room for improvement in the optimality of conditions and directness of methods, the paper provides a fresh perspective on this classical problem with significant theoretical value and potential research impact.