A simple proof for the upper bound of a theorem of T. Łuczak (Q6614510)
From MaRDI portal
| This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use this page instead for the normal view: A simple proof for the upper bound of a theorem of T. Łuczak |
scientific article; zbMATH DE number 7922300
| Language | Label | Description | Also known as |
|---|---|---|---|
| English | A simple proof for the upper bound of a theorem of T. Łuczak |
scientific article; zbMATH DE number 7922300 |
Statements
A simple proof for the upper bound of a theorem of T. Łuczak (English)
0 references
7 October 2024
0 references
Given real numbers \(b,c >1\) and denoting by \(x = [0:a_1(x),a_2(x),\ldots]\) the classical continued fractions expansion of irrationals \(x \in [0,1]\), \textit{T. Łuczak} [Mathematika 44, No. 1, 50--53 (1997; Zbl 0884.11028)] proved that the Hausdorff dimension of\N\[\NE(b,c) = \{x \in [0,1): a_n(x) \geq c^{b^n} \text{ for infinitely many } n\}\N\]\Nis bounded above from \(1/(b+1)\). Since this matches the straight-forwardly obtained lower bound by Feng, Wu, Liang and Tseng [\textit{D.-J. Feng} et al., Mathematika 44, No. 1, 54--55 (1997; Zbl 0884.11029)] for\N\[\N\tilde{E}(b,c) = \{x \in [0,1): a_n(x) \geq c^{b^n} \text{ for all } n\},\N\]\Nand clearly \(\tilde{E}(b,c) \subseteq E(b,c)\), this proved the exact Hausdorff dimension of both sets.\N\NAlthough Łuczak's proof is quite short, it contained some non-trivial covering arguments. In this article, the authors provide an even shorter and simpler proof of Łuczak's result that comes without any covering argument at all. In short, the idea is to observe (which was essentially already a step in the proof of Łuczak) that for any \(1/b < s < 1\) and \(1/s < d < b\) and any \(x \in E(b,c)\) we obtain infinitely many \(n\) \[\tag{1}a_{n+1}(x) > \left(a_1(x) \cdots a_n(x)^{sd-1}c^{(1-s)d^{n+1}}\right).\]\N\NThe authors now show that the set of all \(x\) satisfying $(1)$ (after \(s \to 1\) and \(d \to b\)) has Hausdorff dimension at most \(1/(b+1)\), which clearly implies the claim. After writing the set that satisfies $(1)$ as a limsup set, the authors estimate the Lebesgue measure of cylinder sets \(I(b_1,\ldots,b_n) :=\{x \in [0,1]: a_1(x) = b_1,\ldots, a_n(x) = b_n\}\) and the ratios of \(\frac{\vert I(b_1,\ldots,b_n)\vert}{\vert I(b_1,\ldots,b_n,j)\vert}\) (which are both well-studied, classical estimates in the metric theory of continued fractions) to obtain the desired result.
0 references
continued fractions
0 references
partial quotients with doubly exponential growth
0 references
Hausdorff dimension
0 references
0 references