On sets of integers whose shifted products are powers (Q2427559): Difference between revisions

Suppose that \(A \subset \{1,\dots,N\}\) has the property that \(aa'+1=x^k\) for some \(k >1\) whenever \(a\) and \(a'\) are distinct elements of \(A\). Gyarmati, Sárközy and the author introduced the problem of bounding \(| A| \) from above and conjectured that one has \(| A| =O(1)\). This has since been shown contingent on the \textsl{abc} conjecture, however unconditionally somewhat less is known. In the paper under review the author shows that \(| A| =O_\varepsilon(\log^{2/3+\varepsilon}N)\) for all \(\varepsilon>0\) improving on the previous best of \(O(\log N)\). The proof is very appealing, combining some basic extremal graph theory in the form of bounds for the Zarankiewicz numbers, some estimates for simultaneous linear forms in the logarithms of algebraic numbers due to Loxton (this is the new ingredient over previous work), and some estimates for linear forms in the logarithms of rational numbers. All in all the argument is short but compelling, making progress on a very attractive problem.