The Divisor Function along a Deterministic Orbit and the Emergence of Ladders

Marco Mantovanelli · 2026

Read The Divisor Function along a Deterministic Orbit and the Emergence of Ladders by Marco Mantovanelli (2026) 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-2604-25446.txt.

Introduction

Mantovanelli studies the subtractive divisor orbit, subtracting the divisor count each step, with a(x) the steps to zero. The mean divisor count suggests a(x) grows like x over log x, but sampling left no estimate. He proves orbit divisor counts sum to x, with bounds between roughly x over log-cubed and three-eighths of x. Parity flips exactly at squares. On dyadic segments, he sets out a local-to-global criterion and a truncation. Under two hypotheses, he obtains a(x) like x over log x.

Catalog record

CREATORMarco Mantovanelli
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE44323
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.25446

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

The Divisor Function along a Deterministic Orbit and the Emergence of Ladders Marco Mantovanelli Abstract We study the deterministic recursion n j + 1 = n j − τ ⁡ ( n j ) n_{j+1}=n_{j}-\tau(n_{j}) , where τ ⁡ ( n ) \tau(n) is the divisor function, and the associated orbit length a ⁡ ( x ) a(x) . Heuristics based on the average order of τ ⁡ ( n ) \tau(n) suggest that a ⁡ ( x ) ≍ x / log ⁡ x a(x)\asymp x/\log x , but the strong dependence along the orbit makes this difficult to analyze. We develop a deterministic framework based on energy identities and dyadic decomposition, reducing the problem to the study of the distribution of τ ⁡ ( n ) \tau(n) along the orbit. Our main conceptual contribution is to reduce this endogenous multiplicative problem to a single structural obstruction, captured by a phase-rigidity mechanism for the orbit. Within this framework we prove several unconditional structural results: large values of τ ⁡ ( n ) \tau(n) contribute negligibly on dyadic scales, any energy-concentrating obstruction must occur at a single divisor scale, and small variance of τ ⁡ ( n j ) \tau(n_{j}) along the orbit forces the emergence of a long near-arithmetic progression on which τ ⁡ ( n ) \tau(n) is essentially constant. The only potential obstruction is the formation of energy-saturating divisor ladders. This is addressed via two explicit hypotheses: a phase-rigidity principle for the endogenous orbit, and an anti-concentration statement ruling out such configurations. Under these hypotheses we obtain a ⁡ ( x ) ≍ x log ⁡ x . a(x)\asymp\frac{x}{\log x}. Keywords: Divisor function; deterministic recursion; multiplicative functions; short intervals; orbit dynamics; energy concentration; inverse theorem; near-arithmetic progressions. MSC2020: Primary 11N37; Secondary 11A25, 11N60, 37C10. 1 Introduction The study of arithmetic functions along deterministic sequences lies at the interface of analytic number theory, probability, and dynamical systems. Classical results provide detailed information on the average behavior of multiplicative functions such as the divisor function \tau(n), but these results rely on exogenous sampling , for instance averaging over intervals or arithmetic progressions. In contrast, far less is understood when the sampling is generated endogenously by the function itself. In this paper we study such a setting for the divisor function. Starting from an initial value x , consider the deterministic orbit n_{0}=x,\qquad n_{j+1}=n_{j}-\tau(n_{j}), and define the orbit length a(x):=\min\{k\geq 0:n_{k}\leq 0\}. The sequence a(x) appears in the OEIS as A155043 [ 6 ] . A heuristic based on the average order \sum_{n\leq x}\tau(n)=x\log x+O(x) suggests that typical step sizes are of order \log n , leading to the prediction a(x)\asymp\frac{x}{\log x}. However, this heuristic relies on independence assumptions that fail in the present setting, where the sequence (n_{j}) is highly dependent. Endogenous sampling and the main difficulty The orbit (n_{j}) constitutes an endogenous sample of the divisor function. In particular: • the values \tau(n_{j}) determine future sampling locations, • standard results for multiplicative functions in short intervals or arithmetic progressions [ 4 , 5 ] do not apply directly, • independence heuristics break down due to the recursive structure. This places the problem outside the scope of existing techniques. Structure versus randomness Our approach is based on a structural dichotomy: Either the orbit exhibits mixing, or it develops strong additive structure. We formalize this principle via a phase-rigidity framework , which analyzes the orbit through residue distributions and Fourier bias. This leads to a reduction of the problem to a single, explicit structural obstruction governing the orbit. Using the recursion n_{j+1}=n_{j}-\tau(n_{j}), we isolate the mechanism by which failure of mixing is expected to force strong constraints on the step sizes \tau(n_{j}) . From non-mixing to ladders The key mechanism is the passage from non-mixing to rigidity. Persistent Fourier bias in the orbit constrains the phase increments e\!\left(-\frac{h\tau(n_{j})}{q}\right), leading to residue concentration of \tau(n_{j}) . Combining this with a bounded-divisor reduction yields small variance of \tau(n_{j}) , which in turn implies approximate constancy of step sizes. As a consequence, the orbit follows a near-arithmetic progression, giving rise to a structured configuration that we call a divisor ladder : a sequence m_{1}>m_{2}>\cdots>m_{r} with m_{i+1}=m_{i}-T+o(T),\qquad\tau(m_{i})=T+o(T), where T\asymp\log N and r\asymp N/T . Main result (informal) We reduce the asymptotic behavior of the orbit to a static problem. Theorem (Informal) . If the divisor function does not admit energy-saturating concentration along long near-arithmetic progressions of step T\asymp\log N , then a(x)\asymp\frac{x}{\log x}. The required condition is an anti-concentration statement for \tau(n) , which we formulate as an anti-ladder hypothesis . Relation to existing work Our approach connects several areas: • multiplicative functions in short intervals and progressions [ 4 , 5 , 3 ] , • entropy methods in number theory [ 7 , 8 ] , • structure theorems in additive combinatorics, • classical results on the distribution of \tau(n) [ 1 , 9 , 2 ] . Contribution of this paper We introduce a deterministic framework for analyzing multiplicative functions along endogenous orbits and obtain the following: • a structure-versus-randomness reduction based on phase rigidity, • deterministic reductions showing that large divisor values are negligible and that any obstruction must occur at a single scale, • unconditional baseline bounds for a(x) , including a(x)\gg_{\varepsilon}x^{1-\varepsilon}, • a conditional proof of a(x)\asymp\frac{x}{\log x} under an anti-concentration hypothesis. Our main conceptual contribution is to reduce the problem to a single structural obstruction, captured by a phase-rigidity principle for the orbit. Outlook The central open problem is to establish the required anti-concentration estimate for \tau(n) along long arithmetic progressions. Progress in this direction would likely require new ideas beyond current methods for multiplicative functions. 2 Heuristics and numerical observations In this section we provide heuristic considerations and numerical observations concerning the behavior of the orbit length a(x)=\min\{k\geq 0:n_{k}(x)\leq 0\},\qquad n_{j+1}=n_{j}-\tau(n_{j}). These considerations are intended only as motivation and are not used in the proofs of the main results. 2.1 First-order heuristic A natural first approximation replaces the divisor function \tau(n) by its average order \log n , leading to the heuristic recursion n_{j+1}\approx n_{j}-\log n_{j}. Interpreting this as a continuous process yields \frac{dn}{dj}\approx-\log n, and hence a(x)\approx\int_{2}^{x}\frac{dt}{\log t}\sim\frac{x}{\log x}. This suggests the first-order approximation a(x)\asymp\frac{x}{\log x}, which is consistent with the expected asymptotic behavior. 2.2 Refined heuristic Numerical experiments indicate that the effective step size along the orbit may be slightly larger than \log n on average. A simple phenomenological refinement is to consider \tau_{\mathrm{eff}}(n)\approx\log n+\log\log n, which leads to a(x)\approx\frac{x}{\log x+\log\log x}. This refinement is purely heuristic and not derived from rigorous arguments; it should be viewed only as a possible second-order correction. 2.3 Numerical observations We computed a(n) for n\leq 10^{4} and compared it with \frac{n}{\log n}\quad\text{and}\quad\frac{n}{\log n+\log\log n}. Figure 1: Orbit length a(n) for n\leq 10^{4} compared with n/\log n and n/(\log n+\log\log n) . As shown in Figure 1 , both heuristic curves capture the overall growth trend, although noticeable fluctuations persist. To quantify deviations from the heuristics, we consider normalized ratios; see Table 1 . x a(x) \dfrac{a(x)}{x/\log x} \dfrac{a(x)}{x/(\log x+\log\log x)} \dfrac{a(x)}{\mathrm{Li}(x)} 10 3 0.6908 0.9410 0.0091 10^{2} 19 0.8750 1.1651 0.4670 10^{3} 116 0.8013 1.0255 0.6440 10^{4} 962 0.8860 1.0996 0.7719 10^{5} 7534 0.8674 1.0515 0.7825 10^{6} 65059

Read the complete text