The distribution functions of $σ(n)/n$ and $n/ϕ(n)$, II

Andreas Weingartner
DOI: https://doi.org/10.48550/arXiv.1011.4262
2010-11-19
Abstract:Let $\sigma(n)$ be the sum of the positive divisors of $n$, and let $A(t)$ be the natural density of the set of positive integers $n$ satisfying $\sigma(n)/n \ge t$. We give an improved asymptotic result for $\log A(t)$ as $t$ grows unbounded. The same result holds if $\sigma(n)/n$ is replaced by $n/\phi(n)$, where $\phi(n)$ is Euler's totient function.
Number Theory
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the asymptotic behavior of the distribution functions \(A(t)\) and \(B(t)\) of two number - theoretic functions \(\frac{\sigma(n)}{n}\) and \(\frac{n}{\phi(n)}\) as \(t\) approaches infinity. Specifically, the author hopes to improve the error - term estimates of these distribution functions when \(t\) is large. ### Problem Background Let: \[ A(t) := \lim_{N \to \infty} \frac{1}{N} \left| \{ n \leq N : \frac{\sigma(n)}{n} \geq t \} \right|, \] where \(\sigma(n)\) is the sum of all positive divisors of \(n\); \[ B(t) := \lim_{N \to \infty} \frac{1}{N} \left| \{ n \leq N : \frac{n}{\phi(n)} \geq t \} \right|, \] where \(\phi(n)\) is Euler's Totient function (i.e., the number of positive integers less than or equal to \(n\) and coprime to \(n\)). It is known that these two limits exist and are continuous functions of \(t\). The work of Erdős et al. shows that: \[ B(t) = \exp \left( -e^t e^{-\gamma} (1 + o(1)) \right), \quad (t \to \infty), \] where \(\gamma\approx0.5772\) is the Euler - Mascheroni constant. ### Main Contributions of the Paper On this basis, the author further improves the error terms and obtains the following results: #### Theorem 1 \[ A(t), B(t) = \exp \left\{ -e^t e^{-\gamma} \left( 1 + \sum_{j = 2}^{m} a_j t^{-j}+O(t^{-(m + 1)}) \right) \right\}, \] where the specific coefficients \(a_j\) are as follows: \[ a_2=-\frac{\pi^2}{6e^{2\gamma}}, \quad a_3=\frac{\pi^2}{6e^{3\gamma}}, \quad a_4 =-\left( \frac{\pi^2}{6}+\frac{37\pi^4}{360} \right) e^{4\gamma}. \] #### Theorem 2 By introducing a more complex main - term expression, the error term can be further reduced: \[ A(t), B(t) = \exp \left\{ -y+\min_{s\in J} I(y, s)+R(y) \right\}, \] where \(J = [y\log y - y, y\log y + y]\), and the order of \(R(y)\) is \(O\left(\frac{y}{L(y)^c}\right)\), \(L(y)=\exp \left\{ (\log y)^{3/5}(\log \log y)^{1/5} \right\}\). Assuming the Riemann hypothesis holds, we have: \[ R(y)=O(\sqrt{y}(\log y)^2). \] ### Conclusion By improving the error terms, the paper provides more accurate asymptotic formulas for \(A(t)\) and \(B(t)\), thereby better describing the distribution characteristics of \(\frac{\sigma(n)}{n}\) and \(\frac{n}{\phi(n)}\). This not only deepens the understanding of these number - theoretic functions but also provides more accurate tools for subsequent research.