Every real number is a sum of two real numbers with diverging partial quotients
Dmitry Gayfulin, Erez Nesharim · 2025
Read Every real number is a sum of two real numbers with diverging partial quotients by Dmitry Gayfulin, Erez Nesharim (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-04521.txt.
Introduction
Dmitry Gayfulin and Erez Nesharim prove a number splitting result about continued fraction growth in 2025, from the short abstract in the text head. Every irrational number, they show, is a sum of two real numbers whose continued fraction entries grow without bound. The proof builds the two numbers step by step rather than just claiming they exist. It leans on a recent algorithm by Nikita Shulga, which the authors also study in its own right.
Catalog record
| CREATOR | Dmitry Gayfulin, Erez Nesharim |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-04521
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Every real number is a sum of two real numbers with diverging partial quotients Dmitry Gayfulin, Erez Nesharim Abstract We show that every irrational number is a sum of two real numbers with diverging partial quotients. The proof is constructive. The key towards these results is an algorithm which was recently developed by Nikita Shulga [ 8 ] , and our study of this algorithm is of independent interest. 1 Introduction Define the set G:=\{\alpha\in\mathbb{R}\,:\,\alpha\in\mathbb{Q}\textrm{ or }a_{n}(\alpha)\xrightarrow[n\to\infty]{}\infty\}\,, where a_{n}(\alpha) is the n th partial quotient in the continued fraction expansion of a real number \alpha , whenever its continued fraction expansion is longer than n . The set G was first studied by Good [ 3 ] , where it is proved that its Hausdorff dimension is 1/2 . In this paper we resolve a question suggested by Shulga [ 8 , Problem 6.2] and show that every real number is a sum of two real numbers in G . Theorem 1 . For every \alpha\in\mathbb{R} there exist \beta,\gamma\in G such that \alpha=\beta+\gamma\,. (1) Decompositions of real numbers as a sum of two numbers satisfying a Diophantine property are abundant in the literature. Arguably, the most notable examples of this sort are the theorem of Hall [ 4 ] , where G is replaced by F(a):=\{\alpha\in\mathbb{R}\,:\,a_{n}(\alpha)\leq a\textrm{ for every $n\geq 1$}\}\,, for a=4 , and the theorem of Erdős [ 2 ] , where G is replaced by the set of Liouville numbers. Another Diophatine property was studied by Cusick [ 1 ] . Let S(a):=\{\alpha\in\mathbb{R}\,:\,a_{n}(\alpha)\geq a\textrm{ for every $n\geq 1$}\}\,, where a\geq 2 is a fixed integer. Following Cusick’s result for a=2 , Shulga [ 8 ] proved that for any a\geq 2 and any 0\leq\alpha\leq\frac{1}{a-1} , there exists a decomposition ( 1 ) with \beta,\gamma\in S(a) . Our proof of Theorem 1 is constructive. In fact, our decomposition ( 1 ) is given by an algorithm that was developed and analyzed by Shulga [ 8 ] , and the proof relies on complementing its analysis. Shulga’s algorithm is described in the next subsection, and the analysis we carry out is of independent interest. The reader is referred to Section 2 for more background on continued fractions. 1.1 Shulga’s decomposition In [ 8 ] Shulga established a nice and simple algorithm which decomposes real numbers into a sum of two real numbers with remarkable properties. Definition 2 (Shulga) . For any \alpha\in[0,1] define sequences \{b_{n}\}_{n\geq 1} and \{c_{n}\}_{n\geq 1} , and real numbers \beta and \gamma as follows: 1. For every n\geq 0 , if the integers b_{k} and c_{k} are defined for every 1\leq k\leq n , then if \alpha=[0;b_{1},\ldots,b_{n}]+[0;c_{1},\ldots,c_{n}]\,, we put \beta:=[0;b_{1},\ldots,b_{n}]\,,\quad\textrm{and}\quad\gamma:=[0;c_{1},\ldots,c_{n}]\,, and finish the algorithm. Otherwise, define \displaystyle b_{n+1} \displaystyle:=a_{n+1}(\alpha-[0;c_{1},\ldots,c_{n}])+1\,,\quad\textrm{and} \displaystyle c_{n+1} \displaystyle:=a_{n+1}(\alpha-[0;b_{1},\ldots,b_{n},b_{n+1}])\,. 2. If b_{n} and c_{n} are defined for every n\geq 1 , i.e., the algorithm takes infinitely many steps, then we put \beta:=[0;b_{1},b_{2},\ldots]\,,\quad\textrm{and}\quad\gamma:=[0;c_{1},c_{2},\ldots]\,. The following properties of the algorithm were established in [ 8 ] : Theorem 3 (Shulga) . For any \alpha\in[0,1] the numbers \beta and \gamma in Definition 2 are well defined and satisfy \alpha=\beta+\gamma\,. Moreover, the sequences \{b_{n}\}_{n\geq 1} and \{c_{n}\}_{n\geq 1} in Definition 2 satisfy \displaystyle b_{k} \displaystyle=a_{k}(\alpha-[0;c_{1},\ldots,c_{n}])\,\quad\textrm{and} \displaystyle c_{k} \displaystyle=a_{k}(\alpha-[0;b_{1},\ldots,b_{n}])\,, for any 1\leq k\leq n . As will be seen below, a key step in Shulga’s proof of Theorem 3 is to show that sequences \{b_{n}\}_{n\geq 1} and \{c_{n}\}_{n\geq 1} of Definition 2 satisfy the inequality (see Section 3 ) c_{n}\geq b_{n}\,, (2) for every n\geq 1 . Likewise, a key step in the proof of Theorem 1 is a strengthening of ( 2 ): Theorem 4 . For any \alpha\in[0,1] and n>1 the Shulga decomposition satisfies c_{n}>b_{n}\,. (3) In Section 4 we will show that Theorem 4 implies the following corollary, which implies Theorem 1 : Corollary 5 . For any \alpha\in[0,1] the Shulga decomposition ( 1 ) satisfies \beta,\gamma\in G . Moreover, if \alpha\in\mathbb{Q} then \beta,\gamma\in\mathbb{Q} . Remark 1 . If \alpha=p/q\in\mathbb{Q} , one can ask what is the length of the continued fraction expansions of the \beta and \gamma in the Shulga decomposition. It follows from the proof of Corollary 5 that this length is O\left(q^{2}\right) . Numerical evidence seem to suggest that the maximal length is O(\log{q}) , but the study of this phenomenon lies beyond the scope of the current paper. The key towards Theorem 4 is a detailed analysis of the sets of reals with a given first n steps of Shulga’s algorithm. This analysis allows us to prove further properties regarding the growth rate of the partial quotients of \beta and \gamma . Theorem 6 . For any \alpha\in[0,1] and n\geq 1 the Shulga decomposition satisfies \displaystyle b_{n+1} \displaystyle\geq b_{n}\,, (4) \displaystyle b_{n+2} \displaystyle\geq b_{n}+1\,, (5) \displaystyle b_{n} \displaystyle\geq n\,. (6) The estimates in Theorems 4 and 6 are non-improvable, as one can see from the Shulga decomposition 531/629=[0;2,2,3,5]+[0;2,3,4,8]\,, and the inequalities ( 4 ) and ( 5 ) have no analogues for the sequence \{c_{n}\}_{n\geq 1} . For example, the Shulga decomposition 28244/141973=[0;6,27]+[0;30,29]\,, shows that \{c_{n}\}_{n\geq 1} is not necessarily monotonic. This gives a negative answer to [ 8 , Problem 6.1] . In fact, one can show that for every l>1 , there exists \alpha such that c_{l}<c_{1} . Here is an example with l=4 : 9974074083712426/149649898029019789=[0;16,227,231,235]+[0;240,229,233,237]\,. Numerical evidence suggest that ( 6 ) can be improved for large n , and more specifically, that for every \alpha the Shulga decomposition should satisfy b_{n}>2n for all n large enough. However, the following theorem shows that a linear growth rate is best possible. Theorem 7 . There exists an irrational number such that the partial quotients of its Shulga decomposition ( 1 ) satisfy the inequalities 4n-2\leq b_{n}<c_{n}<5n. (7) Theorem 7 is proved by an explicit construction. Remark 2 . It is possible to deduce that the number constructed in the proof of Theorem 7 satisfies c_{n}<(4+\varepsilon)n for every \varepsilon>0 and all n large enough. However, we decided to settle with the estimate ( 7 ) in order to simplify the proof. 2 Basic properties of continued fractions In this section, we recall some basic facts about continued fractions that we use in the article. For a more detailed background, we refer to the classical literature on the subject (e.g., [ 5 , 6 , 7 ] ). Any real number \alpha can be represented as a continued fraction as follows \alpha=[a_{0};a_{1},a_{2},\ldots]:=a_{0}+\cfrac{1}{a_{1}+\cfrac{1}{a_{2}+\cdots}}\,, where a_{0}\in\mathbb{Z} and a_{i}\in\mathbb{Z}_{+} for i\geq 1 . This representation is finite for rational \alpha , and infinite and unique for irrational \alpha . Each rational number has exactly two different representations [a_{0};a_{1},a_{2},\ldots,a_{n}]=[a_{0};a_{1},a_{2},\ldots,a_{n}-1,1]\,,\quad a_{n}\geq 2\,. Thus, without loss of generality we will assume that the last partial quotient of rational numbers is greater than 1 . For any n\geq 1 we consider a_{n}(\cdot) as a function \mathbb{R}\to\mathbb{N} that returns the n th partial quotient of the continued fraction expansion. Of course, if \alpha is rational, then n should not exceed the length of the continued fraction expansion of \alpha ; otherwise, a_{n}(\alpha) is undefined. The convergents of a continued fraction are the rational numbers \frac{p_{n}(\alpha)}{q_{n}(\alpha)}:=[a_{0};a_{1},a_{2},\ldots,a_{n}] whose numerators and denominators are defined by the recurrence relations \displaystyle p_{0}(\alpha) \displaystyle=a_{0}\,,\; \displaystyle p_{1}(\alpha) \displaystyle=a_{1}a_{0}+1\,,\; \displaystyle p_{n}(\alpha) \displaystyle=a_{n}p_{n-1}(\alpha)+p_{n-2}(\alpha)\,,\quad\textrm{and} (8) \displaystyle q_{0}(\alpha) \displaystyle=1\,,\;