🔢 mathematics

Improved Power Laws for the Favard Length Problem in All Dimensions

This paper establishes improved quantitative upper bounds for the Favard length of purely 1-unrectifiable rational product Cantor sets in all dimensions $d \geq 2$ by combining coordinatewise estimates with an enhanced symbolic-cylinder combinatorial argument, yielding sharper decay rates such as $N^{-1/5}$ for the classical four-corner Cantor set and general exponents dependent on the properties of coordinate mask polynomials.

Original authors: Caleb Marshall

Published 2026-08-14

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Caleb Marshall

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 you are a detective trying to solve a mystery about shapes that are infinitely crinkly. In the world of mathematics, there is a special branch called "geometric measure theory" that studies these weird, jagged shapes, often called fractals. Some of these shapes are so twisted that if you tried to draw a smooth line through them, you'd never get anywhere; mathematicians call these "purely unrectifiable." Now, here is the puzzle: if you shine a light on such a shape from every possible angle and look at the shadow it casts on a wall, what happens? A famous theorem from the 1930s tells us that for these twisted shapes, the average length of all those shadows is exactly zero. It's as if the shape is a ghost that disappears when you look at it from the side.

But mathematicians are curious creatures. They don't just want to know that the shadow disappears; they want to know how fast it vanishes. If you look at a slightly blurry version of the shape (a "neighbourhood" that includes a tiny bit of space around it), the shadow has a tiny, non-zero length. The question is: as you make that blur smaller and smaller, does the shadow length shrink slowly like a snail, or does it vanish quickly like a rocket? This rate of shrinking is called the "Favard length," and figuring out the exact speed of this disappearance is a major challenge in modern math. It's like trying to predict exactly how fast a melting ice cube will disappear as the temperature drops, but for shapes that exist in multiple dimensions and have no smooth edges.

In this paper, Caleb Marshall tackles this mystery for a whole new class of these jagged shapes, moving beyond the flat, two-dimensional world into higher dimensions. He focuses on a specific type of fractal called a "rational product Cantor set," which can be thought of as a digital maze built by repeatedly chopping a block into smaller pieces and keeping only certain corners. The author proves that for these shapes, the shadow length doesn't just vanish; it disappears at a specific, predictable speed, following a "power law." Think of a power law as a strict rulebook for how fast the shadow shrinks: if you shrink the blur by a factor of 10, the shadow might shrink by a factor of 100, or 1,000, depending on the shape's hidden structure.

Marshall's main achievement is finding a faster, more efficient way to calculate this shrinking speed. He combines two powerful tools: a "Fourier" analysis (which is like breaking a complex sound into its individual musical notes to see which ones are quiet) and a clever combinatorial argument (a counting game played with the shape's building blocks). By improving the rules of this counting game, he shows that the shadows of these shapes disappear much faster than previously thought for many cases. For the most famous example, the "four-corner Cantor set" (a shape made of four corners in a square), he improves the known speed limit from a factor of $N^{-1/6}$ to $N^{-1/5}$ . While this might sound like a small change, in the world of fractals, it's a massive leap, like discovering a shortcut through a maze that everyone thought was a dead end.

Even more exciting, Marshall shows that for some specially designed shapes, the shadow vanishes even faster. He constructs examples where the shadow shrinks at a rate of $N^{-1/3}$ or even slightly faster (around $N^{-0.34}$ ), beating the old "one-third" barrier that many mathematicians thought was a hard limit. He does this by carefully choosing the "digits" used to build the fractal so that they avoid certain mathematical "traps" (zeros) that usually slow down the shrinking process. The paper doesn't just guess these numbers; it provides a rigorous, step-by-step proof that these faster rates are real.

The paper also clarifies what doesn't work. It shows that you can't just use a generic method to get these fast results; the specific "cyclotomic" structure of the shape's building blocks matters immensely. If the shape is built with the wrong kind of digits, the shadow won't vanish as quickly. The author proves that for the classic four-corner set, you need to use a specific, improved counting method to get the $1/5$ result, and that simply tweaking the old methods isn't enough to break the $1/3$ barrier for that specific shape. However, for other, custom-built shapes, the barrier can be broken. The confidence in these results is high; the author doesn't just suggest these rates are possible, he proves them using a combination of algebra, geometry, and probability, showing that the math holds up under strict scrutiny.

In short, this paper is a masterclass in how to measure the "ghostliness" of jagged shapes. It takes a difficult problem that has stumped mathematicians for decades, refines the tools used to solve it, and discovers that for many of these shapes, the shadows vanish much faster than anyone expected. It's a story of finding hidden order in chaos, proving that even the most twisted, unrectifiable shapes follow a strict, predictable rhythm when you know how to listen to their mathematical music.

Technical Summary: Improved Power Laws for the Favard Length Problem in All Dimensions

Problem Statement
The paper addresses the quantitative decay of the Favard length for compact, purely 1-unrectifiable sets in $\mathbb{R}^d$ ( $d \ge 2$ ). The Favard length, defined as the average one-dimensional Hausdorff measure of orthogonal projections onto lines, vanishes for such sets. The central problem is to establish quantitative upper bounds for the Favard length of $\delta$ -neighbourhoods of these sets, specifically seeking "power law" bounds of the form $\text{Fav}(N_\delta(E)) \lesssim (\log \delta^{-1})^{-\epsilon}$ or $\delta^\epsilon$ . While qualitative results (Besicovitch–Federer theorem) are known, quantitative rates for deterministic self-similar sets, particularly in dimensions $d \ge 3$ , have remained elusive or suboptimal.

Methodology
The author develops a unified framework combining Fourier analysis, combinatorial geometry, and number theory to derive these bounds for a broad class of "rational product Cantor sets." The methodology proceeds through several distinct stages:

Geometric Reduction and Fourier Setup:
The problem is reduced to estimating the $L^2$ norm of a counting function associated with projected cylinder intervals. Using a projective chart reduction, the sphere $S^{d-1}$ is covered by finite charts, allowing the analysis to focus on a "safe" frequency interval. The counting function is expressed as a Riesz product, and the problem is transformed into bounding the $L^2$ mass of this product.
Algebraic Decomposition (Cyclotomic Factorization):
The mask polynomials associated with the digit sets of the Cantor sets are decomposed into four factors based on their zeros on the unit circle:
- $A^{(1)}$ : Cyclotomic factors with roots not accumulating under dilation.
- $A^{(2)}$ : Accumulating cyclotomic factors (roots of unity).
- $A^{(3)}$ : Factors with zeros on the unit circle that are not roots of unity.
- $A^{(4)}$ : Factors with no zeros on the unit circle.
  This decomposition dictates the analytic strategy: $A^{(1)}$ and $A^{(3)}$ are handled via "Set of Small Values" (SSV) estimates, while $A^{(2)}$ requires "Set of Large Values" (SLV) constructions to avoid accumulating zeros.
Analytic Estimates (SSV and SLV):
- SSV: The paper establishes quantitative covering properties for the sub-level sets of the Riesz product factors corresponding to non-accumulating zeros.
- SLV: For accumulating zeros, the author constructs a "Set of Large Values"—a measurable set $\Gamma$ whose difference set avoids the zeros of the accumulating factors. This construction relies on "fibered" digit sets and cluster methods to ensure the difference set has sufficient measure while avoiding the zero set.
Combinatorial Propagation:
A critical innovation is an improved combinatorial argument operating on the symbolic cylinder tree. The author introduces a "low-multiplicity" set of directions where the projection map is not highly non-injective.
- Low Multiplicity: For directions in this set, an $L^2$ estimate is derived using a refined "stopping-time" packing argument on the symbolic tree, improving upon previous bounds.
- Propagation: For directions outside this set (high multiplicity), the author employs a "block propagation" argument. By identifying a fixed proportion of "green" (good) terminal words and propagating this structure across scales, the paper proves that the projection length decays rapidly. This step introduces a propagation cost parameter $J_{K,\rho} \approx K^{\rho-1} \log K$ .

Key Contributions and Results

General Power Laws in All Dimensions: The paper establishes the first genuinely non-trivial power law bounds for deterministic self-similar sets in dimensions $d \ge 3$ . Theorem 2.20 provides a general exponent $\delta$ depending on the digit sets and the dimension, applicable to rational product Cantor sets where coordinate digit sets satisfy specific cardinality or cyclotomic conditions (e.g., small cardinality, two-prime conditions, or admitting fibered subsets).
Improved Exponent for the Four-Corner Cantor Set: Applying the general machinery to the classical four-corner Cantor set ( $d=2$ ), the author improves the known exponent. While Nazarov, Peres, and Volberg established a bound of $N^{-1/6+u}$ , this paper proves a bound of $N^{-1/5+u}$ for every $u > 0$ . This improvement stems directly from the refined combinatorial propagation argument, which reduces the propagation cost from $K^{2+o(1)}$ to $K^{1+o(1)}$ .
Zero-Free Digit Sets and Exponent Optimization: The paper demonstrates that if the coordinate mask polynomials are zero-free on the unit circle, the exponent is governed solely by the minimum modulus of these polynomials.
- For a specific base-25 planar example with digit set $\{0, 1, 2, 6, 9\}$ , the exponent is improved to $N^{-1/3+u}$ .
- For a base-36 example with digit set $\{0, 7, 14, 15, 18, 24\}$ , the exponent exceeds the $1/3$ threshold, attaining $\delta_{36} \approx 0.3407 - u$ .
Fibered Digit Sets: The work extends the scope of applicable sets to include those admitting "fibered subsets," a structural property related to vanishing sums of roots of unity, allowing for power laws in cases where previous cardinality bounds failed.

Significance and Claims
The paper claims to provide the first quantitative power law estimates for the Favard length problem in dimensions $d \ge 3$ for genuinely higher-dimensional self-similar sets (not merely planar sets embedded in higher dimensions). The significance lies in:

Unification: Extending planar techniques (originally developed by Nazarov, Peres, Volberg, and others) to arbitrary dimensions $d \ge 2$ .
Quantitative Improvement: Concretely improving the exponent for the four-corner Cantor set from $1/6$ to $1/5$ through a more efficient combinatorial propagation mechanism.
Structural Insight: Demonstrating that the decay rate is intimately tied to the cyclotomic structure of the digit sets. The paper shows that sets with "zero-free" mask polynomials can achieve significantly better exponents (crossing $1/3$ ), suggesting that the $1/4$ barrier (theoretical limit for the four-corner set if weighted estimates were available) is not a fundamental obstruction of the method but a limitation of the current Fourier input.

The author notes that while the current methods yield $1/5$ for the four-corner set, achieving the $1/4$ or $1/3$ barrier for this specific set would require new Fourier analytic inputs (specifically, weighted estimates) that are not currently available. However, for other constructed sets with favorable arithmetic properties, the paper achieves exponents strictly better than $1/3$ .

Technical Summary: Improved Power Laws for the Favard Length Problem in All Dimensions

More like this