Higher-order colossally abundant numbers
Abstract
We construct nested subclasses of colossally abundant numbers by changing the coordinates in their supporting-line description.
The least member $m_k$ of the class of order $k$ is called its first contact.
Our main questions concern the growth of $m_k$, the factorization of the numbers in each class, and the intersection of all classes.
For coordinates $(e^{a u^\beta},-(\sigma(n)/n)^k)$, where $u=\log\log n$, every class is infinite, the intersection is empty, and $u(m_k)\sim(k/(a\beta))^{1/\beta}$, unconditionally.
Every fixed integer divides all members of sufficiently high classes; in a specified range of coordinates, $m_k\mid m_{k+1}$.
Robin's inequality restricted to any one of these classes is equivalent to the Riemann hypothesis (RH).
We also construct families whose intersection is exactly the set of global maximizers of $\sigma(n)/(n\log\log n)$ on the CA tail; for these families, $m_k\to\infty$ is equivalent to RH.
Finally, a bound $\Re\varrho\le b<1$ for the nontrivial zeta zeros gives, for the natural choice of abscissa $\log n$, $\log\log m_k=k+O(k^3e^{-(1-b)k})$ and estimates for the prime factors of $m_{k+1}/m_k$.
The value $b=7/8$ uses the result stated in the cited OpenAI preprint and requires no assumption of RH.