Searcharxiv⌕ Search

arXiv subjects

Ralf Stephan

Publications and source records attributed to Ralf Stephan.

11 recordsLinked to original sources

The Evertse-Ferretti interval result for a set of weights, and the exponent $1/2$ in three complexity bounds

The interval result of Evertse and Ferretti (2013) concerns a tuple $L$ of linear forms and one tuple $c$ of weights. It says that the parameters $Q$ for which the twisted height $H_{L,c,Q}$ has a small value outside an exceptional subspace $T(L,c)$ lie in $O(δ^{-2}\logδ^{-1})$ intervals $[Q_i,Q_i^ω)$. We prove this for a whole set $\mathcal{C}$ of weights with the same forms. At most $J$ subspaces $T(L,c)$ occur, where $J$ depends only on the dimension and on the number of places, and one family of $O(δ^{-2}\logδ^{-1})$ intervals serves every $c\in\mathcal{C}$ at once. Neither count depends on $\mathcal{C}$. In applications of Bugeaud-Evertse type the solutions are sorted into about $\varepsilon^{-1}$ classes with the same forms and different weights, and one application now serves all classes. We deduce three complexity bounds. For a real algebraic irrational number $ξ$, the number $p(n,ξ,b)$ of blocks of length $n$ in its base-$b$ expansion satisfies $\limsup p(n,ξ,b)/(n(\log n)^η)=\infty$ for every $η<1/2$; Bugeaud and Evertse proved this for $η<1/11$. The same holds for the continued fraction expansion of an algebraic number of degree at least three, and for the steering word of $(3/2)^n$. We explain why $1/2$ is the limit of the method. All results are machine-verified in Lean~4.

math.NT↗

The exponent $1/4$ in three complexity bounds: integer bases, continued fractions, and the $(3/2)^n$ steering word

Let $ξ$ be a real algebraic irrational number and $p(n,ξ,b)$ the number of distinct blocks of length $n$ in its base-$b$ expansion. Bugeaud and Evertse proved in 2008 that $p(n,ξ,b)/(n(\log n)^η)$ is unbounded for every $η<1/11$. We prove this for every $η<1/4$. We also prove the analogous bound for the continued fraction expansion of an algebraic number of degree at least three, with every exponent $δ<1/4$. Bugeaud announced this statement in 2013 with an unspecified $δ>0$. Finally, the subword complexity of the steering word of $(3/2)^n$ exceeds $C\,k(\log k)^η$ infinitely often for every $η<1/4$ and every $C$. The three proofs share one count: $\varepsilon^{-1}$ classes times $\varepsilon^{-3}$ subspaces from the quantitative Subspace Theorem of Evertse and Ferretti. The subspaces are then disposed of at the cost of a logarithm, with no further power of $\varepsilon^{-1}$. All three results, and the quantitative Subspace Theorem they rest on, are machine-verified in Lean~4.

math.NT↗

Superlinear complexity of the $(3/2)^n$ steering word

Write $(3/2)^n = m_n + \varepsilon_n$ with $m_n$ the nearest integer and $\varepsilon_n\in[-\tfrac12,\tfrac12)$, and let $T=(t_n)$, $t_n=2m_{n+1}-3m_n$, be the resulting \emph{steering word}: the step-by-step record of the map $x\mapsto\tfrac32 x$ on the orbit of 1, coded by nearest-integer rounding. We prove that the subword complexity $p_T(k)$ of $T$ is superlinear, $p_T(k)/k\to\infty$. A pigeonhole argument reduces this to exponential pair repulsion of the orbit modulo one, which we establish by two applications of the $p$-adic Subspace Theorem: once in three variables, which handles all gaps between the two orbit points at once, and once in two variables (Ridout's theorem) for the single degenerate family of constant gap. The argument is completely formalized in Lean~4.

math.NT↗

Isaac's Condition (TU) is false

In 2006 Isaac reduced the simple normality to base $2$ of $\sqrt s$, for $s$ not a perfect square, to a property of dyadic expansions that he called Condition~(TU): the upper frequency of the digit~$1$ in $\sqrtν$, along any subsequence, depends only on the tail of the expansion of $ν$. We show that Condition~(TU) is false, already for the full sequence. We give two proofs. The first is elementary and turns Isaac's own reduction against the condition. The second uses the theorem of Kaufman and Mosquera--Shmerkin on the Fourier decay of smooth images of self-similar measures, and gives counterexamples for almost every point of a Cantor measure. All results are machine-verified in Lean~4; the second proof uses the two cited theorems as hypotheses.

math.NT↗

Transcendence criteria for the minimal word of the rational base $3/2$

Let $x_{0}$ be a positive integer, let $x_{n}=\lceil 3x_{n-1}/2\rceil$, and let $w_{n}=2x_{n+1}-3x_{n}\in\{0,1\}$ be the associated word studied by Dubickas; for $x_{0}=1$ the orbit is A061419 and $w$ is the minimal word $g_{3/2}$ of the rational base number system of Akiyama, Frougny and Sakarovitch. The orbit encodes a real constant $K=\lim_{n}x_{n}(2/3)^{n}$, equal for $x_{0}=1$ to $ω_{3/2}=K(3)=1.6222705028\ldots$, whose irrationality has been open since 1977. We prove that if $w$ is automatic then $K$ is transcendental; equivalently, an algebraic $K$ forces $w$ to be non-automatic. Further we prove that either the complexity of $w$ exceeds every linear bound or $K$ is irrational. All results are formally verified in the Lean~4 proof assistant, on three cited axioms.

math.NT↗

Aperiodicity and subword complexity in the binary expansion of powers of three

We prove two results on the fine structure of the binary digits of $3^{m}$. First, for every fixed period $p$, the number of positions at which the binary expansion of $3^{m}$ breaks $p$-periodicity grows in order like $\log m/\log\log m$; equivalently, no window of the expansion deeper than a fixed power of $\log m$ is $p$-periodic. Second, the finite binary word formed by the low-order digits of $3^{m}$ has full low-order subword complexity: its complexity function satisfies $p_{3^{m}}(n)\ge n+1$ for every length $n$, once $m$ is large enough.

math.CO↗

Interleaved snowballing: Reducing the workload of literature curators

We formally define the literature (reference) snowballing method and present a refined version of it. We show that the improved algorithm can substantially reduce curator work, even before application of text classification, by reducing the number of candidates to classify. We also present a desktop application named LitBall that implements this and other literature collection methods, through access to the Semantic Scholar academic graph (S2AG).

cs.DL↗

Lineare Rekurrenzen, Potenzreihen und ihre erzeugenden Funktionen

Diese kurze Einfuehrung in Theorie und Berechnung linearer Rekurrenzen versucht, eine Luecke in der Literatur zu fuellen. Zu diesem Zweck sind viele ausfuehrliche Beispiele angegeben. This short introduction to theory and usage of linear recurrences tries to fill a gap in the literature by giving many extensive examples.

math.HO↗

Divide-and-conquer generating functions. Part I. Elementary sequences

Divide-and-conquer functions satisfy equations in F(z),F(z^2),F(z^4)... Their generated sequences are mainly used in computer science, and they were analyzed pragmatically, that is, now and then a sequence was picked out for scrutiny. By giving several classes of ordinary generating functions together with recurrences, we hope to help with the analysis of many such sequences, and try to classify a part of the divide-and-conquer sequence zoo.

math.CO↗