Paperlane

no. 10 · Friday 6 November 2026

A Collatz-type conjecture on the set of rational numbers (revised and extended version)

Mohammad Javaheri

math.NT · v2

Abstract

Let $θ(x)=(x-1)/3$ if $x\ge1$ and $θ(x)=2x/(1-x)$ if $0\le x<1$. We conjecture that the $θ$-orbit of every nonnegative rational number ends in $0$. A weaker conjecture asserts that no map in the semigroup generated by $3x+1$ and $x/(x+2)$, other than the generators' powers, has a positive rational fixed point; equivalently, the only matrices in the semigroup generated by $\left(\begin{smallmatrix}3&1\\0&1\end{smallmatrix}\right)$ and $\left(\begin{smallmatrix}1&0\\1&2\end{smallmatrix}\right)$ with an integer eigenvalue are their pure powers. We prove sharp two-sided bounds for the smallest eigenvalue of every word in the more general semigroups attached to the maps $\varphi_{a,b}$, which explain why the pair $\{2,3\}$ is distinguished: it is the only pair with $(a-1)(b-1)=2$. Combining these bounds with congruences for the trace, we determine the possible integer eigenvalues of words, and we prove that $θ$ has no rational $1$-circuits, no rational $2$-circuits and no even rational circuits. We show that, for each $k$, a random $k$-syllable word with exponents at most $M$ has an integer eigenvalue with probability at most $2k(k+\log_2 2k)/M$. Finally, we verify by computer that $θ$ has no positive rational cycle of length at most $36$.

comments
This is a revised and extended version of a published article
journal
M. Javaheri, A Collatz-type conjecture on the set of rational numbers, J. Number Theory 132 (2012), no. 11, 2582--2594
fields
Number Theory
pdf ↗ page arXiv:1010.3692 ↗