Skip to main content
18 events
when toggle format what by license comment
Oct 15, 2025 at 17:29 history edited Eduardo Rodríguez Golvano CC BY-SA 4.0
added 19 characters in body
Oct 15, 2025 at 17:16 comment added Eduardo Rodríguez Golvano @GHfromMO The proof is complete now. As Terry Tao pointed out, it does not make use of the prime number theorem.
Oct 15, 2025 at 17:05 history edited Eduardo Rodríguez Golvano CC BY-SA 4.0
Provided the details for big n. The proof is complete now.
Oct 10, 2025 at 14:25 comment added Eduardo Rodríguez Golvano @GHfromMO absolutely, the proof is not complete without the last part. I am trying to remember what I had thought about this. In the meantime I have edited my answer to add an alternative for big n (based on your proof)
Oct 10, 2025 at 14:23 history edited Eduardo Rodríguez Golvano CC BY-SA 4.0
I added an alternative for big n
Oct 10, 2025 at 0:38 comment added GH from MO Before seeing the proof above, based on Terry's ideas, I worked out my own proof that $n!$ is not highly abundant for $n\geq 168$. In this range, by Theorem 1 in Rosser-Schoenfeld (1962), we have primes $n/3<p_1<p_2<p_3\leq n/2$. Let us also take a prime $n<q<2n$, and define the contender $M:=\frac{qrn!}{p_1p_2p_3}$, where $qr<p_1p_2p_3<q(r+1)$. The contender wins: $$\frac{\sigma(M)}{n!}>\frac{\sigma(q)}{q}\left(\prod_j\frac{p_j\sigma(p_j)}{\sigma(p_j^2)}\right)\left(1-\frac{q}{p_1p_2p_3}\right)>\left(1+\frac{1}{2n}\right)\left(\frac{n^2+3n}{n^2+3n+9}\right)^3\left(1-\frac{54}{n^2}\right)>1.$$
Oct 10, 2025 at 0:25 comment added GH from MO By the way, if one wants a quick proof for $n\geq n_0$ based on the results of Erdős-Alaoglu (1944), I recommend to look at their Corollary below Theorem 16. By this Corollary, if $n!$ is highly abundant, then every prime $q<c_4\log(n!)$ divides $n!$. So for such $n$, there is no prime strictly between $n$ and $c_4\log(n!)\sim c_4 n\log n$, which means that $n$ is bounded. The proof of this Corollary contains an annoying typo. In the second display of the proof, $\sigma(qn)/nq^{k+1}$ should be $\sigma(n)/nq\sigma(q^k)$.
Oct 10, 2025 at 0:19 comment added GH from MO @EduardoRodríguezGolvano Your proof looks nice, but can you provide the details for $n>1320$? How to choose $k$ precisely, and why does it really work? Thanks in advance.
Oct 9, 2025 at 18:55 comment added Terry Tao Nice! In particular one is able to handle the asymptotic regime without needing any form of the prime number theorem (as opposed to the Alaoglu-Erdos analysis, which needs variants of PNT at many places in their paper).
Oct 9, 2025 at 15:54 comment added Eduardo Rodríguez Golvano @MaxAlekseyev yes, I mentioned Mersenne primes to mean the requirement could obviously be satisfied
Oct 9, 2025 at 15:29 comment added Matthew Bolan This together with the OEIS data means $n!$ is highly abundant if and only if $n \le 8$.
Oct 9, 2025 at 15:27 comment added Max Alekseyev "$2^k-1$ is not a Mersenne prime" can be easily enforced by taking composite $k$.
S Oct 9, 2025 at 12:45 history suggested G. Melfi CC BY-SA 4.0
improved formatting
Oct 9, 2025 at 12:19 review Suggested edits
S Oct 9, 2025 at 12:45
Oct 9, 2025 at 11:36 comment added Sam Hopkins As Terry Tao pointed out, this was also mentioned in the Alaoglu-Erdos paper.
Oct 9, 2025 at 11:35 comment added Saúl RM Eduardo decided to create an account (this is the same person that told me that $L_{2310}$ is not highly abundant, which then led to the proof that $L_{n}$ is not highly abundant for big $n$)
S Oct 9, 2025 at 11:31 review First answers
Oct 9, 2025 at 12:40
S Oct 9, 2025 at 11:31 history answered Eduardo Rodríguez Golvano CC BY-SA 4.0