Let $p\neq 3$ be a Mersenne prime. Is it true that $(p^2-1)/2$ has a prime divisor $r$ such that $r^{2}$ does not divide $n$?

  • 1
    $\begingroup$ The condition "for every prime $q$ dividing $n$, $q^2$ also divides $n$" is called being powerful in the literature. So after dividing out the factors of $2$, you'd like to know if $2^{p-1} - 1$ is ever powerful when $2^p - 1$ is prime. Ribenboim has some conjectures related to similar questions in "The New Book Of Prime Number Records", in the section on Wieferich primes. $\endgroup$ – Zack Wolske Apr 6 '12 at 18:25

Well, maybe I'm missing the point, but the first Mersenne prime is $3=2^2-1$, and $\frac{3^2-1}{2} = 4 = 2^2$, so... no, I guess.


Considering there are only 47 known Mersenne primes, and finding the factors of $2^p-1$ is a difficult task, I'm not sure this question is fully tractable. But we can certainly show that all of the known Mersenne primes satisfy your question. First, take $m=2^p-1$ to be a Mersenne number, and rewrite $n= \frac{m^2 - 1}{2} = 2^{2p-1} - 2^p$. This is powerful when $p=2$, which gives Philip van Reeuwijk's counterexample. For larger $p$, we know $4$ will always divide $n$, so we can ignore powers of $2$ and just consider the odd part $n'= 2^{p-1} - 1$.

We know $p$ is prime, and odd since it is not $2$, so we are looking at $n' = 4^k - 1$, with $k=\frac{p-1}{2}$. Clearly $3$ divides $n'$, and $9$ divides iff $3$ divides $k$. So we require that $p \equiv 1 \mod6$. This is not enough; plenty of known Mersenne primes have this property.

Since $3$ divides $k$, we can write $n'=64^{k'}-1$. Then $7$ divides $n'$, and $7^2$ divides $n'$ iff $7$ divides $k'$. So now we require that $p \equiv 1 \mod42$. Sadly, again this is not enough.

One step further, we see that when $7$ divides $k'$, $43$ divides $64^{k'}-1$, and $43^2$ divides iff $43$ divides $k'$. Now we're happy (for the time being), because no known Mersenne prime has $p\equiv 1 \mod 1806 = 43*7*6$. But it seems there's no reason they can't have this property, so you may have to continue your search once such a Mersenne prime is found. Expect one by the 504th instance: $504 = \phi(1806)$.

  • $\begingroup$ I have checked a few cases, and if $m = 2^p-1$, it actually seems the case that $p$ is a factor of $n = \frac{m^2-1}{2}$ and $p^2$ isn't. This even seems to work when $m$ isn't a prime but just the $p$-th Mersenne number with $p$ prime. I can't prove it, though... $\endgroup$ – Philip van Reeuwijk Apr 6 '12 at 19:32
  • $\begingroup$ If you could that would be something interesting. Showing that $p$ divides is an application of little Fermat. Conjecture $W'_2$ on $p. 343$ of Ribenboim's "New Book of Prime Numbers" is "There exist only finitely many primes with $2^{p-1} \equiv 1 \mod p^2$", and he goes on to show that this would imply solutions to other well-known open problems. $\endgroup$ – Zack Wolske Apr 6 '12 at 20:35
  • $\begingroup$ Yes, the little Fermat part I had figured out already; the part about $p^2$ not being a factor is beyond me, I fear. Ribenboim's book looks interesting, I'll check it out. Thanks for the tip! $\endgroup$ – Philip van Reeuwijk Apr 6 '12 at 20:52
  • $\begingroup$ Just to clarify, such primes are called Wieferich, and only two are known: 1093, and 3511. $\endgroup$ – Zack Wolske Apr 6 '12 at 21:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.