4

I have the following problem with me:

Show that there are infinitely many prime numbers $p, q$ such that $p \mid 2^{q - 1} - 1, q \mid 2^{p - 1} - 1$.

Obviously $p$ and $q$ both are not equal to 2. When I have $p = q$, The result is again obvious by Fermat's Little theorem. Now comes the case where I found the difficulty, whenever $p$ is not equal to $q$ and both are not equal to 2. In such a case, I need two primes $p$ and $q$ in such a way that the product $pq$ divides $2^{p - 1} - 1$. Now how do I prove that there exist an infinite number of them.

A silly doubt: When I took the case $p = q$ and showed it is true by Fermat, and since there exist infinite number of primes, can I say that I have proved the given statement? Either ways I am interested in the case where $p$ is not equal to $q$.

Robert Soupe
  • 14,663
saisanjeev
  • 2,050

1 Answers1

4

Judging directly by the wording of the problem, yes, you have proven the claim - all pairs $(p,p)$ for any prime $p>2$ work. This is really unsatisfying.

I present a proof that there are infinitely many pairs $(p,q)$ of distinct primes that satisfy this. It is a strange proof and I believe it can be made better, but it is a proof nonetheless.

Lemma. If $r$ is a prime number for which $2^r-1$ is not prime, there exist two distinct primes $p,q>r$ so that the condition holds for $(p,q)$.

Proof. We first claim that $2^r-1$ is not a prime power. Indeed, if $2^r=p^k+1$ for some prime $p$ and exponent $k>1$, if $k$ is even we have that the RHS is $\equiv 2\bmod 8$, a contradiction, and it $k$ is odd we may write

$$2^r=(p+1)\left(p^{k-1}-p^{k-2}+\cdots-p+1\right)$$

where the second factor is odd and greater than $1$, a contradiction. So, there exist at least two distinct prime factors of $2^r-1$.

Claim. Let $r$ be a prime number. If a prime $p$ divides $2^r-1$, then $r|p-1$.

Proof. We have that the order $\mathrm{ord}_p(2)|r$, so it is either $1$ or $r$. It is clearly not $1$ as $p\nmid 1$, so $\mathrm{ord}_p(2)=r$. As $\mathrm{ord}_p(x)|p-1$ for any prime $p$ and integer $x$, we see that $r|p-1$. $\blacksquare$

From here, let $p,q$ be two distinct primes that divide $2^r-1$. We have $r|p-1,q-1$, so

$$p|2^r-1|2^{q-1}-1,$$

and similarly $q|2^{p-1}-1$. Finally, $r|p-1,q-1\implies r\leq p-1,q-1 \implies p,q>r$, so we are done. $\blacksquare$


If there are infinitely many primes for which $2^r-1$ is composite (I believe this is open), we then have that we can find pairs $(p,q)$ satisfying the problem condition with arbitrarily large values of $\min(p,q)$, so we have infinitely many.

Otherwise, let $r_0$ be the largest prime for which $2^{r_0}-1$ is composite (such a prime exists - take $2^{11}-1=23\cdot 89$ for example), and let $p_0,q_0$ be distinct primes dividing $2^{r_0}-1$.

Define $p_{n+1}=2^{p_n}-1,q_{n+1}=2^{q_n}-1$ for all $n\geq 0$. We claim that the pair $(p_{n+1},q_{n+1})$ satisfies the condition. First we see that $(p_n,q_n)$ is actually a pair of distinct primes, which is immediate by the fact that $p_0,q_0>r_0$ and $f(x)=2^x-1$ is monotone increasing (and in fact injective over $\mathbb{N}$). We show the remaining claim by induction.

By our lemma, the pair $(p_0,q_0)$ satisfies the problem condition, so our base case is true. For any pair $(p,q)$ satisfying the problem condition, we have that

$$p|2^{q-1}-1\implies p|2^q-2 \implies 2^p-1|2^{(2^q-1)-1}-1,$$

and similarly $2^q-1|2^{(2^p-1)-1}-1,$ so the condition holds for $(2^p-1,2^q-1)$. So, if $(p_n,q_n)$ satisfies the condition, then $(p_{n+1},q_{n+1})$ does too. So, we have constructed an infinitely increasing sequence of pairs of distinct primes satisfying the problem condition, and thus we have infinitely many such pairs. $\blacksquare$

  • Very nice! (+1) – quasi Sep 24 '18 at 07:48
  • @quasi Thank you! I just noticed that this condition is equivalent to $pq$ being a base-$2$ Fermat pseudoprime - it might be easier to solve it from there. – Carl Schildkraut Sep 24 '18 at 07:51
  • I think the lemma should be stated as: If $r$ is prime and $2^r-1$ is not prime, ... – quasi Sep 24 '18 at 13:59
  • @quasi Yes, it should. I have edited my answer to reflect this. – Carl Schildkraut Sep 24 '18 at 15:18
  • @quasi In your lemma what is the condition that you are referring to in the second part? – saisanjeev Sep 25 '18 at 14:09
  • @saisanjeev: The answer was posted by Carl Schildkraut, not me. But presumably, the condition you're referring to is just the requirement of the problem for the primes $p,q$. – quasi Sep 25 '18 at 14:18
  • oh i am sorry got confused @CarlSchildkraut is that the condition you are referring to? – saisanjeev Sep 25 '18 at 15:16
  • @saisanjeev Sorry, I probably could have made the wording clearer. Whenever I use the phrase "the condition", it's referencing the problem condition. For two primes $(p,q)$, "the condition" holds iff $p|2^{q-1}-1$ and $q|2^{p-1}-1$. Also, what is the source of this problem? – Carl Schildkraut Sep 25 '18 at 16:48
  • @CarlSchildkraut Although your answer is from over $5$ years ago, just in case you're still interested in the source of the problem, please see my comment to this question, indicating the basic source (although this was more "satisfying" as it had the requirement that $p$ & $q$ be distinct primes) is "Romania TST $7$ $2009$, Problem $3$". – John Omielan Dec 31 '23 at 06:00