🔢 mathematics

Metric results for dyadic approximation on the middle-third Cantor set

This paper resolves Velani's conjecture on the metric theory of dyadic approximation for the middle-third Cantor set by establishing new uniform Fourier decay estimates for the Cantor-Lebesgue measure, which prove that the set of well-approximable points has measure zero for $\tau > 1$ and full measure for $0 < \tau \leq 0.01$ .

Original authors: Xin-Rong Dai, Bing Li, Bo Wang, Yu-Feng Wu

Published 2026-06-25

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Xin-Rong Dai, Bing Li, Bo Wang, Yu-Feng Wu

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

The Big Picture: A Game of Hide-and-Seek on a Fractal

Imagine you have a very strange, dusty cloud of points called the Cantor Set. It's not a solid line; it's a "fractal," meaning if you zoom in, it looks the same as the whole thing. It's made by taking a line, chopping out the middle third, chopping out the middle third of what's left, and repeating this forever.

Now, imagine you are playing a game of Hide-and-Seek on this cloud.

The Hider: A specific point $x$ living inside the Cantor Set.
The Seeker: A series of "searchlights" that sweep across the number line. These searchlights are based on powers of 2 (like 2, 4, 8, 16...).
The Rule: The searchlight at step $n$ looks at the number $2^n \times x$ . If the result is very close to a whole number (like 1, 2, or 3), the hider is "caught."

The question mathematicians have been asking for decades is: How often does the hider get caught?

There is a famous guess (conjecture) by a mathematician named Velani. He suggested that the answer depends entirely on how "wide" the searchlight is.

If the searchlight is too narrow (mathematically, if the width shrinks too fast), the hider almost never gets caught.
If the searchlight is wide enough (shrinks slowly), the hider gets caught almost all the time.

The Problem: The "Static" Cloud

The difficulty in solving this problem is that the Cantor Set is "noisy." In math terms, it doesn't have a smooth, predictable pattern when you look at it through a Fourier lens (a tool that breaks waves into frequencies). It's like trying to hear a clear melody in a room full of static interference.

Previous researchers (like Baker, Allen, and others) had managed to prove Velani's guess for only a tiny slice of the "wide searchlight" scenarios and a tiny slice of the "narrow searchlight" scenarios. There was a huge gap in the middle where no one knew the answer.

The Breakthrough: Tuning the Radio

The authors of this paper (Dai, Li, Wang, and Wu) found a way to cut through the static.

The Analogy:
Imagine the Cantor Set is a radio station broadcasting a signal, but the signal is full of static. Previous researchers knew that if you listened for a short time, the signal would fade in and out unpredictably. They couldn't predict the long-term pattern.

The authors' key innovation was to prove that if you listen to the signal for a long time and add up the "loudness" of the static in a specific way, the noise actually cancels itself out in a predictable pattern. They proved a new mathematical estimate that says: "Even though the signal looks chaotic, if you sum up the squares of the noise over time, it behaves very nicely."

This allowed them to tune the radio perfectly.

The Results: Filling the Gap

By using this new "noise-canceling" technique, the authors were able to prove Velani's guess for a much larger range of scenarios than anyone else had before.

The "Never Caught" Zone (Null Part): They proved that if the searchlight is narrower than a specific threshold (roughly 1.429 times a certain standard), the hider is almost never caught. This improves on the previous best result of about 1.552.
The "Always Caught" Zone (Full Measure Part): They proved that if the searchlight is wider than a specific threshold (roughly 0.052), the hider is caught almost every time. This improves on the previous best result of 0.01.

In simple terms: They closed the gap between "never caught" and "always caught" significantly. They showed that the transition happens exactly where Velani predicted, just in a wider range than we knew before.

The "Bonus" Discovery: Other Fractals

The paper also mentions that their new "noise-canceling" technique isn't just for the Cantor Set. It works for a whole family of similar fractal shapes (called "missing-digit sets").

The Analogy:
If the Cantor Set is a specific type of dusty cloud, their new method works on any cloud made by a similar "chopping" process, as long as the chopping follows certain rules. They showed that the same "Hide-and-Seek" rules apply to these other clouds too.

Summary

The Goal: Predict how often a point in a fractal gets "caught" by a specific mathematical search pattern.
The Obstacle: The fractal is too chaotic to analyze with old tools.
The Solution: The authors developed a new way to measure the "chaos" (Fourier estimates) that reveals a hidden order.
The Result: They confirmed a long-standing guess for a much wider range of scenarios, proving that the transition from "safe" to "caught" happens exactly as predicted.

They didn't invent a new application for this (like medical imaging or cryptography); they simply solved a deep, abstract puzzle about the nature of numbers and shapes.

Technical Summary: Metric Results for Dyadic Approximation on the Middle-Third Cantor Set

Problem Statement
This paper addresses the problem of Diophantine approximation on fractals, specifically focusing on the middle-third Cantor set $C$ . The central object of study is the set $W_2(\tau)$ , defined as the collection of points $x \in [0, 1]$ such that $\|2^n x\| < n^{-\tau}$ for infinitely many $n \in \mathbb{N}$ , where $\|\cdot\|$ denotes the distance to the nearest integer. The authors investigate the measure of this set with respect to the Cantor-Lebesgue measure $\mu$ (the $\gamma$ -dimensional Hausdorff measure restricted to $C$ , where $\gamma = \log 2 / \log 3$ ).

The work centers on a conjecture attributed to Velani, which posits a zero-one law for $\mu(W_2(\tau))$ : the measure should be 0 if $\tau > 1$ and 1 if $0 < \tau \le 1$ . While this conjecture remains open in its full generality, previous research had established partial results. Notably, Allen, Baker, Chow, and Yu [1] proved the null case for $\tau > \approx 1.552$ , and Baker [3] proved the full measure case for $0 < \tau \le 0.01$ . The difficulty of the problem stems from the fact that the base-2 and base-3 expansions of numbers in $C$ are not both structured, a principle related to Furstenberg's $\times 2, \times 3$ conjecture, but current quantitative understanding of this independence is limited.

Methodology
The authors employ a Fourier-analytic approach, building upon the techniques developed in [1] and [3]. A primary obstruction in this field is that the Cantor-Lebesgue measure $\mu$ lacks global Fourier decay (i.e., $|\hat{\mu}(\xi)|$ does not tend to 0 as $|\xi| \to \infty$ ). Previous works relied on Baker's Lemma [3], which established that $|\hat{\mu}(h2^n)|$ decays polynomially for all but a small set of $n$ .

The key methodological innovation in this paper is the establishment of precise estimates for the partial sums of the Fourier coefficients of $\mu$ along the sequence $h2^n$ . Specifically, the authors prove uniform estimates for the sums of $|\hat{\mu}(h2^n)|$ and $|\hat{\mu}(h2^n)|^2$ . These estimates are derived using number-theoretic properties of the powers of 2 modulo powers of 3, specifically leveraging the 3-adic valuation and the distribution of residues.

The core technical tools include:

New Fourier Estimates: Proving that $\sum_{n=1}^N |\hat{\mu}(h2^n)|^2 \ll N^{1-\gamma}$ and $\sum_{n=1}^N |\hat{\mu}(h2^n)| \ll N^{(1-\gamma)/2}$ uniformly in $h \in \mathbb{Z} \setminus \{0\}$ .
Smoothing and Integral Estimates: Utilizing Jackson's inequality to construct smooth approximations of indicator functions for the shrinking targets. These are combined with the new Fourier estimates to derive bounds on the first and second moments of the counting function.
Borel-Cantelli Arguments: Applying the moment bounds to establish the zero-one law via the Borel-Cantelli lemma.

Key Contributions and Results
The paper provides improved bounds for the convergence and divergence parts of Velani's conjecture:

Null Part (Convergence): The authors prove that $\mu(W_2(\tau)) = 0$ for $\tau > \frac{1}{\gamma} - \frac{1-\gamma}{3-\gamma} \approx 1.429$ . This improves upon the previous best result of $\tau > \approx 1.552$ by Allen, Baker, Chow, and Yu [1].
Full Measure Part (Divergence): The authors prove that $\mu(W_2(\tau)) = 1$ for $0 < \tau < \frac{\gamma}{12} \approx 0.052$ . This improves upon the previous best result of $\tau \le 0.01$ by Baker [3].
General Counting Result: For the full measure part, the authors establish a stronger counting result (Theorem 1.4). For any sequence $(x_n)$ in $[0, 1]$ and $0 < \tau < \gamma/12$ , for $\mu$ -almost every $x \in C$ :
$\lim_{N \to \infty} \frac{\#\{1 \le n \le N : \|2^n x - x_n\| < n^{-\tau}\}}{2 \sum_{n=1}^N n^{-\tau}} = 1.$
This generalizes Baker's result to inhomogeneous settings and arbitrary sequences.
Extension to Missing-Digit Sets: The methodology is extended to a broader class of self-similar measures on missing-digit sets $K_{b,D}$ (where $b$ is a prime and $D$ is a digit set). Theorem 1.7 provides a similar counting result for these sets, with a threshold $\kappa$ depending on the probability vector and the base.

Significance and Claims
The paper claims that its primary contribution lies in the refinement of the Fourier input used in the mechanism of [1] and [3]. By establishing the sharper estimates on the partial sums of Fourier coefficients (specifically the $L^1$ and $L^2$ bounds along the sequence $h2^n$ ), the authors are able to push the known thresholds for Velani's conjecture closer to the conjectured value of $\tau = 1$ .

The authors explicitly state that their approach generalizes to self-similar measures on missing-digit sets, demonstrating the robustness of their Fourier-analytic technique beyond the specific case of the middle-third Cantor set. They acknowledge that while these results represent significant progress, the conjecture remains open for the range $0.052 \le \tau \le 1.429$ . The work does not propose new applications or experimental frameworks but focuses strictly on advancing the theoretical understanding of metric Diophantine approximation on fractals.