On stronger conjectures that imply the Erdős-Moser conjecture (Q531841): Difference between revisions

The Erdős-Moser conjecture states that the Diophantine equation \(S_k(m)=m^k\), where \(S_k(m)=1^k+2^k+\cdots+(m-1)^k\) has no solution for \(m>3\). The best result to date is that it has no solution with \(m>10^{10^9}\), due to \textit{Y. Gallot}, the reviewer and \textit{W. Zudilin} [Math. Comput. 80, No. 274, 1221--1237 (2011; Zbl 1231.11038)]. Here the author states the conjecture that \(S_k(m+1)/S_k(m)\) is never an integer for \(m>3\). This is equivalent with the statement that the Diophantine equation \(aS_k(m)=m^k\) has no solution with \(m>3\) and \(a\geq 1\) an integer. Meanwhile the reviewer [\textit{P. Moree}, ``Moser's mathemagical work on the equation \(1^k+2^k+\cdots+(m-1)^k=m^k\)'', \url{arXiv:1011.2940}, Rocky Mt. J. Math. 43, No. 5, 1707--1737 (2013; Zbl 1362.11045)] has shown that there are infinitely many \(a\) for which this equation has no solution. Let \(g_k(m)\) be the greatest common divisor of \(S_k(m)\) and \(S_k(m+1)\) divided by \(m\). Let \(N_k\) denote the numerator of the \(k\)th Bernoulli number. The author proves that if \[ \max_{m\geq 1}g_k(m)< |N_k|(\log |N_k|)^6 \] for \(k\geq 10\), then the Erdős-Moser conjecture holds true and that more generally the exponent 6 can be replaced by an arbitrary exponent \(e\), at cost of also requiring \(k\geq C_e\), where \(C_e\) is effectively computable. Furthermore, he conjectures that the latter upper bound for \(\max_{m\geq 1}g_k(m)\) holds true for some exponent \(e\) and established this in case \(N_k\) is square-free by showing that \(\max_{m\geq 1}g_k(m)=|N_k|\) then.