[Paper Review] An Improvement To Zaremba's Conjecture

ShinnYih Huang|arXiv (Cornell University)|Oct 14, 2013

Mathematical Dynamics and Fractals13 references4 citations

TL;DR

This paper improves Zaremba's conjecture by proving that for alphabets with Hausdorff dimension greater than 5/6, the set of denominators of continued fractions with bounded partial quotients contains a density one subset of the admissible integers. Using a refined circle method and local-global analysis, the authors establish an effective density one result with improved error bounds, showing that almost all admissible integers are represented as denominators with high multiplicity when the alphabet's dimension exceeds 5/6.

ABSTRACT

We prove there exists a density one subset $\dd \subset \N$ such that each $n \in \dd$ is the denominator of a finite continued fraction with partial quotients bounded by 5.

Motivation & Objective

• To strengthen the local-global principle for Zaremba’s conjecture by improving the threshold for the Hausdorff dimension of the alphabet.
• To establish an effective density one result for denominators of continued fractions with partial quotients bounded by 5, improving upon prior bounds.
• To refine the error term analysis in the circle method to achieve sharper estimates in the context of bounded continued fraction expansions.
• To investigate the role of local obstructions in the denominator set and show that alphabets with dimension >5/6 are likely free of such obstructions.

Proposed method

• Applies the circle method to decompose the counting function into a main term and an error term, leveraging spectral gap estimates and exponential sum bounds.
• Uses a recursive construction of the set Ω_N to model the distribution of continued fraction denominators with bounded partial quotients.
• Employs a matrix group action on the modular group to analyze local solubility and lift solutions modulo prime powers.
• Implements a lifting argument via the matrix (1 k; 0 1) to propagate solutions from mod k to mod k^n, ensuring full residue coverage.
• Combines techniques from Bourgain-Kontorovich and Frolenkov-Kan to refine the error term decay to O(e^{-c√log N}) for the density one statement.
• Analyzes local obstructions via group-theoretic conditions on the alphabet, showing that only specific arithmetic progressions can cause obstructions.

Experimental results

Research questions

• RQ1 Can the threshold for the Hausdorff dimension in the local-global conjecture be improved from 307/312 to 5/6?
• RQ2 Is it possible to achieve a density one result for the denominator set with an effective error term under a lower dimension threshold?
• RQ3 What is the maximal possible Hausdorff dimension of an alphabet that still exhibits local obstructions?
• RQ4 How does the multiplicity of denominators grow as a function of the dimension δ_A when δ_A > 5/6?
• RQ5 Under what conditions on the alphabet does the set of denominators contain almost all admissible integers?

Key findings

• For any alphabet 𝒜 with Hausdorff dimension δ_𝒜 > 5/6, there exists a subset 𝔻̃_𝒜 ⊂ 𝔻_𝒜 that contains almost every admissible integer, with relative density 1 + O(e^{-c√log N}) as N → ∞.
• The error term in the density estimate decays faster than any power of log N, specifically as O(e^{-c√log N}), with c = c(𝒜) > 0 effectively computable.
• Each d ∈ 𝔻̃_𝒜 appears with multiplicity ≫ N^{2δ_𝒜 - 1 - r} for any r < (1/9)(δ_𝒜 - 5/6), with implied constants depending only on 𝒜 and r.
• The alphabet 𝒜 = {1,2,3,4,5} has δ_𝒜 ≈ 0.8368 > 5/6, so Theorem 1.6 applies, confirming that almost all admissible integers are denominators of continued fractions with partial quotients bounded by 5.
• The maximal Hausdorff dimension of an alphabet with local obstructions is bounded above by max(δ_oct, δ_even), where δ_oct ≈ 0.70 and δ_even ≈ 0.70 from finite truncations, suggesting δ_𝒜 > 5/6 implies no local obstructions.
• The paper proves that if an alphabet 𝒜 satisfies the local condition (i.e., full coverage modulo k^n for all n), then 𝔻_𝒜 contains a density one subset of the admissible integers when δ_𝒜 > 5/6.