9
$\begingroup$

The semiprime $87 = 3*29$ has a curious property: it's the fact that both

$87^2 + 29^2 + 3^2 = 8419$

and

$87^2 - 29^2 - 3^2 = 6719$

are prime numbers.

This intrigued me and led me to wonder if there are other semiprimes with the same property, and I found that

$21 = 3*7$ is another example, since both

$21^2 + 7^2 + 3^2= 499$

and

$21^2 - 7^2 - 3^2 = 383$ are prime numbers

So the following question arises: Are there infinitely many prime numbers $p$ and $q$, with $p \neq q$, such that both

$(pq)^2 + p^2 + q^2$

$(pq)^2 - p^2 - q^2$

are also primes?

Does this follows from some known theorem or conjecture?

Improve this question
$\endgroup$
4
  • $\begingroup$ Does it help anyhow to consider a second degree equation $ X^2-SX+P=0 $ with $ P=(a+b)(a-b) $ , $ a=(pq)^{2}=S/2 $, $b=p^2+q^2 $ as well as the equation $Y^2-bY+a=0 $? $\endgroup$
    – Sylvain JULIEN
    Aug 13, 2018 at 21:41
  • $\begingroup$ I don't know if it helps, but by some basic algebra massaging you can show your two polynomials to be equal to $(pq-1)^2 + (p+q)^2 - 1$ and $(pq+1)^2 - (p+q)^2 -1$ respectively. The latter implies that the composite number $(p-1)(q-1)(p+1)(q+1)$ must be one more than a prime number, while the former (interestingly?) implies that the Gaussian composite $(p-i)(q-i)(p+i)(q+i)$ must be one more than a prime number (they are the same statement but with $p, q \mapsto ip, ip$). $\endgroup$
    – Michael Cromer
    Aug 14, 2018 at 2:18
  • $\begingroup$ As usual with this kind of question about primes, I dare say that the answer is "probably, yes". $\endgroup$
    – Eric Duminil
    Aug 14, 2018 at 12:10
  • $\begingroup$ See Bateman-Horn conjecture $\endgroup$
    – Kevin Casto
    Aug 14, 2018 at 19:22

1 Answer 1

Reset to default
18
$\begingroup$

If both $p$ and $q$ are $\pm 1\mod 6$, then $(pq)^2 + p^2 + q^2$ is divisible by 3, so that there is only a possibility if one of the primes is 2 or 3. For $q = 2$ (and $p \ne 3$), $(pq)^2 + p^2 + q^2 = 5p^2 + 4$ also is divisible by 3. The question then becomes, are there infinitely many primes $p$ so that both $10p^2 + 9$ and $8p^2 - 9$ are prime?

Since it is not known if there is any polynomial of degree greater than 1 that produces infinitely many primes, let alone a pair of polynomials that in an infinite number of primes produce prime numbers, I guess it will be extremely hard to prove that there are infinitely many such pairs $(p, q = 3)$. It may be possible to prove that they don't, though I expect that would be hard as well.

Improve this answer
$\endgroup$
5
  • 1
    $\begingroup$ There are 26 such primes $p$ up to 10,000: 7, 29, 83, 181, 197, 337, 601, 631, 1303, 1847, 2029, 3023, 3109, 3359, 4591, 4649, 4831, 6397, 6791, 7489, 7559, 7573, 7951, 8609, 8933, and 9857 (provided you trust Maple and my programming skills). $\endgroup$
    – Gerry Myerson
    Aug 13, 2018 at 23:11
  • 5
    $\begingroup$ I should have checked OEIS first. They're tabulated at oeis.org/A079796 and called "nonomatic primes". $\endgroup$
    – Gerry Myerson
    Aug 13, 2018 at 23:13
  • $\begingroup$ The first $10,000$ such primes are given at the OEIS and a comment that the number is likely infinite. It should be possible to calculate how many such primes one would expect up to $N$ and use that data to see how well that seems to work. $\endgroup$
    – Aaron Meyerowitz
    Aug 14, 2018 at 6:18
  • $\begingroup$ @GerryMyerson great find! Do you have any idea why they are called that way? $\endgroup$
    – doetoe
    Aug 14, 2018 at 7:17
  • $\begingroup$ No idea. Was wondering myself. $\endgroup$
    – Gerry Myerson
    Aug 14, 2018 at 7:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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