The following problem was raised in a Mathlinks thread:

If $a,b,c\in\mathbb Z$ such that $a\ne0$ and $b^2-4ac\ne 0$, for how many consecutive integers $x$ can $ax^2+bx+c$ ba a perfect square ?

The polynomial $-15x^2+64$ is obviously a square for the five numbers $x=-2,...,2$, but the method used for finding this in the above thread cannot be extended further. Should $5$ really be the best possible answer?

Has this problem been treated somewhere else?

  • $\begingroup$ For starters $15w^2-14$ is a square for $w=\pm1,\pm3,\pm5$ so you get six by taking $w=2x+1$. But eventually you get to surfaces of general type. See my answer to mathoverflow.net/questions/73346 . $\endgroup$ – Noam D. Elkies Feb 11 '12 at 23:01
  • 4
    $\begingroup$ ...and if I did it right, you can get to $w = \pm 7$ using points on an elliptic curve of (conductor 360 and) rank 1. That might be the maximum. FWIW $0,\pm1,\pm2,\pm3$ seems to give a curve of conductor 30 with 12 torsion points but rank zero. $\endgroup$ – Noam D. Elkies Feb 11 '12 at 23:12

To resonate with Noam Elkies' comments, it is conjectured that $8$ squares is the maximum for arbitrary $a$, and $4$ squares is the maximum for $a=1$. For $5$ symmetric squares the smallest known leading coefficients are $a=15$ and $a=-20$, while for $5$ increasing squares they are $a=60$ and $a=-56$. It is known that there are infinitely many examples with $5$ or $8$ symmetric squares, or with $6$ increasing squares. It is also known that there is no symmetric sequence of $7$ squares, and only finitely many of $10$ squares up to obvious equivalences (this one follows from Falting's theorem applied to a specific hyperelliptic curve).

Good starters are:

Browkin-Brzeziński: On sequences of squares with constant second differences, Canad. Math. Bull. 49 (2006), 481–491.

Bremner: On square values of quadratics, Acta Arith. 108 (2003), 95–111.

  • 4
    $\begingroup$ A recent paper of Gonzales-Jimenez and Xarles [Acta Arith. 149 (2011)] (apologies for the lack of accents) gets a (sharp) upper bound of $8$ for those quadratics with an axis of symmetry half-way between two integers. $\endgroup$ – Mike Bennett Feb 12 '12 at 4:42

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.