Equal sums in random sets and the concentration of divisors (Q6040622): Difference between revisions

Let \(\mathcal A\) be a random set of positive integers in which an integer \(a\) is included with probability \(1/a\), let \(k\in\mathbb Z, k\geq 2\), and let \(\beta (k)\) denote the supremum of the exponents \(c<1\) for which the following is true: with probability tending to 1 as \(D\rightarrow\infty\), there are distinct sets \(A_{1},\dots,A_{k}\) such that \[ \bigcup_{i=1}^{k}A_{i}\subset (\mathcal{A}\cap [ D^{c}, D])\text{ and }\sum_{a\in A_{1}} a =\dots=\sum_{a\in A_{k}} a. \] Then \[ \liminf_{r\rightarrow\infty}\beta (2^{r})^{1/r}\geq\rho/2,\text{ where }\rho=0.281211... \] is a specific constant defined as a limit of a certain rapidly convergent sequence. The proof of this theorem is rather long and complicated and can hardly be described in a few words. The authors write: ``We characterise \(\beta (k)\) as a solution to a certain optimisation problem over measures on the discrete cube \(\{0, 1\}^{k}\), and obtain lower bounds for \(\beta (k)\) which we believe to be asymptotically sharp.'' This theorem is then applied to study the distribution of divisors of natural numbers, the concentration of divisors of a random permutation in a symmetric group and of a random polynomial over a finite field. In particular, the authors prove the following theorem. For \(n\in\mathbb N\) and \(t\in\mathbb R\), let \[ \delta (n, t):=\mathrm{card}\{d|d\in\mathbb N, \log d\in [t, t+1], d|n\} \] and \[ \Delta (n):=\max\{\delta (n, t)|t\in\mathbb R, t\geq 0\}. \] Then \(\Delta (n)\geq (\log\log n)^{\alpha}\) with \(\alpha > 0.3533...\) for almost all \(n\). The best previously known lower bound \(\alpha > 0.3382\ldots\) was obtained in the work of \textit{H. Maier} and \textit{G. Tenenbaum} [Math. Proc. Camb. Philos. Soc. 147, No. 3, 513--540 (2009; Zbl 1239.11105)], who conjectured that their lower bound was optimal. According to the authors, one of the main results of their work is a disproof of this conjecture.