On the distribution of balanced subgroups (Q2637191)

The author proves a conjecture of [\textit{C. Pomerance} and \textit{D. Ulmer}, in: Number theory and related fields, Borwein, Jonathan M. (ed.) et al., Springer Proceedings in Mathematics \& Statistics 43, 253--270 (2013; Zbl 1332.11089)] regarding so-called balanced cyclic subgroups of the group of units \({\mathbb U}_n\) of the ring \({\mathbb Z}/n{\mathbb Z}\). The group \({\mathbb U}_n\), for \(n>2\), may be split into two ``halves'': elements represented by integers in \((0, n/2)\), and the elements represented by integers in \((n/2, n)\). A subgroup \(H\) of \({\mathbb U}_n\) is called balanced if every coset \(aH\) has the same number of elements in each half. What motivates the study of this concept, beside the interplay between the additive and multiplicative structure of \({\mathbb U}_n\), is the relation with elliptic curves. If \(H\) contains \(-1\), then it is clearly balanced. Moreover, if \(n=4k\) and \(2k+1\in H\), then \(H\) is also balanced. A balanced subgroup of \({\mathbb U}_n\) is called sporadic if it does not fall into the above two categories. Let \(g\) be an odd prime. Let \({\mathcal B}_{g,0}(x)\) denote the number of all \(n>2\), such that \(n=4k\leq x\), \(g\in{\mathbb U}_n\), and \(2k+1\in \langle g\rangle\). Likewise let \({\mathcal B}_{g,1}(x)\) denote the number of all \(n>2\), such that \(4\neq n\leq x\), \(g\in{\mathbb U}_n\), and \(-1\in \langle g\rangle\). Finally, let \({\mathcal B}_{g,*}(x)\) denote the number of all \(n>2\), such that \(n\leq x\), \(g\in{\mathbb U}_n\), and \(\langle g\rangle\) is sporadic balanced. The main result of the paper is that there exists some \(c>0.00108\dots\) such that for every \(g\) and sufficiently large \(x\), i.e. \(x\geq x_0(g)\), we have \[ {\mathcal B}_{g,*}(x) \leq \frac{x}{(\log x)^{2/3+c}}. \] The bound for \({\mathcal B}_{g,*}(x)\) is thus smaller (by a power of \(\log x\)) than the sizes of \({\mathcal B}_{g,0}(x)\) and \({\mathcal B}_{g,1}(x)\). This shows that, when we fix the generator and vary the modulus, sporadic cyclic subgroups are indeed rare. In order to achieve the stated upper bound the author establishes a complex system of necessary conditions that an integer \(n\) counted by \({\mathcal B}_{g,*}(x)\) must satisfy. These conditions involve, among others, the sets \({\mathcal P}_j\) of odd primes such that \(2^j\) divides exactly the order of the subgroup \(\langle g\rangle\). Another type of condition is related to distinct odd primes \(p_1, p_2 \mid n\) such that \(p_1-1\) and \(p_2-1\) have a common odd prime factor. There are also other conditions involving, e.g., various Legendre symbols. The rest of the paper is concerned with translating this system of conditions to an upper bound for the counting function. The author does this using, among others, a general theorem of \textit{H. L Montgomery} and \textit{R. C. Vaughan} [Multiplicative number theory. I. Classical theory. Cambridge Studies in Advanced Mathematics 97, (2007; Zbl 1142.11001)], a result of Wiertelak concerning the sets \({\mathcal P}_j\) [\textit{K. Wiertelak}, Acta Arith. 43, 177--190 (1984; Zbl 0531.10049)], a modification of a theorem of \textit{J. Grantham} [J. Number Theory 130, No. 5, 1117--1128 (2010; Zbl 1216.11110)], and sieve methods. The reviewer believes that the second inequality in Proposition 5.1 follows from the inequality \(-\log(1-\theta)\leq \theta+\theta^2\) for \(\theta\in (0,1/2]\) and the fact that \(\sum_p \frac{1}{p^2} < +\infty\).