Searcharxiv⌕ Search

arXiv · 2609.30921

Counterexamples to the inhomogeneous Duffin-Schaeffer Conjecture

Abstract

The Duffin-Schaeffer Conjecture, proven in 2020, gives a zero-full dichotomy for the approximation of irrational numbers with reduced fractions for arbitrary approximation functions. We construct approximation functions such that the inhomogeneous analogue of the Duffin-Schaeffer Conjecture fails. The counterexamples are obtained for all non-zero rationals and certain Liouville numbers.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Manuel Hauke-Treuer, James Maynard, Andrew Pollington. 2026-09-25. Counterexamples to the inhomogeneous Duffin-Schaeffer Conjecture. https://arxiv.org/abs/2609.30921

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

Small improvements to the Ball-Rivoal theorem and its $p$-adic variant

We prove that the dimension of the $\mathbb{Q}$-linear span of the Riemann zeta values $ζ(3),ζ(5),\ldots,ζ(s-1)$ is at least $(1.119 \cdot \log s)/(1+\log 2)$ for any sufficiently large even integer $s \geqslant s_0$. This result slightly refines the theorem of Rivoal (2000) or Ball--Rivoal (2001). The proof incorporates the arithmetic factor $Φ_n$ introduced by Zudilin (2001) into a slight generalization of the Ball--Rivoal hypergeometric construction. The asymptotic impact of the arithmetic factor $Φ_n$ was not anticipated prior to this work. Although this result is subsumed by a much stronger recent development by Fischler (2026), our approach yields a new result in the $p$-adic setting, slightly refining a theorem of Sprang (2020). We prove that the dimension of the $\mathbb{Q}$-linear span of the $p$-adic zeta values $ζ_p(3),ζ_p(5),\ldots,ζ_p(s-1)$ is at least $(1.119 \cdot \log s)/(1+\log 2)$ for any prime $p$ and any sufficiently large even integer $s \geqslant s_0(p)$.

math.NT↗

A Fibonacci theorem for Collatz trajectories via modular graph structure

Let $T(n)=n/2$ if $n$ is even and $T(n)=(3n+1)/2$ if $n$ is odd. We prove that for each $m\ge1$, exactly $F(m+1)$ odd integers $n$ in $\{1,\ldots,2^m\}$ have the property that none of the iterates $T(n),T^2(n),\ldots,T^{m-1}(n)$ lies in the residue class $4\pmod6$, where $F(m+1)$ is the $(m+1)$-th Fibonacci number; the proportion decays at rate $(φ/2)^m$, $φ=(1+\sqrt{5})/2$. Equivalently, these are the odd $n\le2^m$ for which no two consecutive terms of $n,T(n),\ldots,T^{m-1}(n)$ are even. The proof uses the directed graph $G$ of Collatz transitions modulo $6$ and its unique absorbing strongly connected component $G'=G[\{1,2,4,5\}]$. Removing vertex $4$ from $G'$ yields a subgraph of spectral radius $φ$, against $ρ(G')=2$; the Fibonacci count follows from this spectral gap. We construct an explicit bijection $Ψ_m:\{1,\ldots,6\cdot2^m\}\to\mathcal{P}_m(G)$ onto the directed paths of length $m$ in $G$. We further show that no vertex of $G'$ is dispensable: removing any single vertex reduces the spectral radius strictly below $2$, with hierarchy $1<\sqrt{2}<φ<2$. In particular, every positive cycle of $T$ must visit residue class $2\pmod6$, and a flow conservation identity forces this class to account for more than $18\%$ of the steps in any such cycle.

math.NT↗