[Paper Review] Intervals Between Farey Fractions in the Limit of Infinite Level

Jan Fiala, Peter Kleban|arXiv (Cornell University)|May 19, 2005

Mathematical Dynamics and Fractals4 references3 citations

TL;DR

This paper investigates the asymptotic behavior of intervals between newly introduced Farey fractions in the modified Farey sequence as the level $k$ approaches infinity. Using recursive bounds on interval lengths derived from parent intervals at lower levels, the authors prove that the lim inf of the total length of even-indexed new intervals $I_k^{(e)}$ is zero, and conjecture that the limit itself is zero. This result, while geometrically simple, is surprisingly subtle and connects to statistical mechanics via partition functions and multifractal phase transitions.

ABSTRACT

The modified Farey sequence consists, at each level k, of rational fractions r_k^{(n)}, with n=1, 2, ...,2^k+1. We consider I_k^{(e)}, the total length of (one set of) alternate intervals between Farey fractions that are new (i.e., appear for the first time) at level k, I^{(e)}_k := \sum_{i=1}^{2^{k-2}} (r_k^{(4i)}- r_k^{(4i-2)}) . We prove that \liminf_{k o \infty} I_k^{(e)}=0, and conjecture that in fact \lim_{k o \infty}I_k^{(e)}=0. This simple geometrical property of the Farey fractions turns out to be surprisingly subtle, with no apparent simple interpretation. The conjecture is equivalent to $ lim_{k o \infty}S_{k}=0, where S_{k} is the sum over the inverse squares of the new denominators at level k, S_{k}:=\sum_{n=1}^{2^{k-1}} 1/ (d_k^{(2n)} )^2. Our result makes use of bounds for Farey fraction intervals in terms of their "parent" intervals at lower levels.

Motivation & Objective

• To analyze the limiting behavior of the total length of alternating intervals between newly introduced Farey fractions at level $k$.
• To understand why the sum of lengths of these 'even' intervals vanishes in the infinite-level limit, despite no obvious geometric reason.
• To establish a connection between the interval length sum $I_k^{(e)}$ and the sum of inverse squares of new denominators $S_k$, suggesting deeper number-theoretic or dynamical structure.
• To provide a foundation for understanding the thermodynamic behavior of Farey-based spin chains and their critical point properties.

Proposed method

• Define the modified Farey sequence recursively via matrix multiplication, generating new fractions at each level $k$ as mediants of neighboring fractions from level $k-1$.
• Identify 'even' intervals as those between fractions of even index at level $k$, which correspond to new fractions introduced at that level.
• Establish recursive bounds on the length of an odd interval at level $k$ in terms of its parent even interval at a lower level $m = k - j$, using the relation $I^{(o,j)}_k \leq I^{(e)}_m \cdot \frac{3}{2j+3}$.
• Use these bounds to derive a recursive inequality for $I_k^{(e)}$, showing that if $I_m^{(e)}$ is small, then $I_k^{(e)}$ is even smaller for sufficiently large $k$, leading to the lim inf result.
• Relate $I_k^{(e)}$ to $S_k = \sum_{n=1}^{2^{k-1}} \frac{1}{(d_k^{(2n)})^2}$, the sum of inverse squares of new denominators, and show that $\lim_{k\to\infty} I_k^{(e)} = 0$ is equivalent to $\lim_{k\to\infty} S_k = 0$.
• Use numerical evidence up to $k=34$ from A. Zhigljavsky to support the conjecture that $I_k^{(e)} \to 0$.

Experimental results

Research questions

• RQ1 Why does the total length of alternating intervals between newly introduced Farey fractions vanish in the limit of infinite level, despite no apparent geometric or number-theoretic reason?
• RQ2 Is the sequence $I_k^{(e)}$ monotonically decreasing, and can this be proven using interval evolution rules?
• RQ3 What is the precise asymptotic behavior of $I_k^{(e)}$ as $k \to \infty$, and does it follow $I_k^{(e)} \sim 1/\log_2 k$ as suggested by prior work?
• RQ4 How are the interval lengths $I_k^{(e)}$ and the sum of inverse squares of denominators $S_k$ related, and does the convergence of one imply the other?
• RQ5 What physical implications does the vanishing of $I_k^{(e)}$ have in the context of Farey-based statistical mechanical models and their phase transitions?

Key findings

• The paper proves that $\liminf_{k\to\infty} I_k^{(e)} = 0$, showing that the total length of even-indexed new intervals between Farey fractions tends to zero in the limit of infinite level.
• The authors conjecture that $\lim_{k\to\infty} I_k^{(e)} = 0$, which would imply that the sum of lengths of these intervals not only has a zero lim inf but actually converges to zero.
• The conjecture is equivalent to $\lim_{k\to\infty} S_k = 0$, where $S_k$ is the sum of the inverse squares of the denominators of the new fractions introduced at level $k$, indicating a deep number-theoretic decay in denominator size distribution.
• Numerical evidence up to $k=34$ supports the conjecture, showing that $I_k^{(e)}$ decreases with $k$, consistent with a decay rate of approximately $1/\log_2 k$.
• The result is physically significant: $I_k^{(e)}$ corresponds to the partition function $Z_k^F(1)$ of the Farey tree model at its critical point $\beta=1$, and its vanishing limit implies non-trivial thermodynamic behavior.
• The authors note that while their method establishes the lim inf result, it is insufficient to prove the full conjecture, and recent independent approaches using ergodic theory, continued fractions, and measure theory have since been proposed.