All the variables I deal with are integers. Given $q \geq 2, |k| \geq 2$, impose a constraints that we have an integer $p$ such that $q^2=1+pk=1\pmod k.$ For example, $3^2=1+1\cdot8, 5^2=1+3\cdot8.$ Now

I suspect that given the triple $(q,k,p)$, then at least one of the following is true.

(1) there is a shift integer $u$ such that $(q+ku)^2=1+p_1k$ where $p_1$ is a square

(2) there is a shift integer $u$ such that $(-q+ku)^2=1+p_1k$ where $p_1$ is a square.

For example, given $q=5, k=8, p=3$. I can find $u=1$ so that the second case $-q+ku=-5+8(1)=3$ and then $3^2=1+8\cdot1$ now $1$ is a square. Notice that if original $p$ is a square then we are done or $k=\pm 1$.

I try to prove this is true. But I somehow fail. Here is my attempt: My process follows like this, suppose now $p$ is not square and I want to solve $(q+ku)^2=1+p_1k.$ Now I can $q^2+2kuq+k^2u^2=1+pk+2kuq+k^2u^2=1+p_1 k.$ I get the following Diophantine equation:

$p+2uq+ku^2=p_1=\theta^2.$ Now I want to solve $u$ in order to let $u$ be an integer, I must have the discriminant of a quadratic equation $u$ to be a perfect square. So I must have

$4q^2-4k(p-\theta^2)=4\phi^2$ for some integer $\phi.$ So know I can plug $q^2=1+pk$ then I can get the following Pell's equation.

$1+k\theta^2=\phi^2.$ Due to the theory of Pell's equation, if $k$ is not a perfect square, then there will be infinite solutions. But in order for $u$ to be an integer. I have $u=\frac{-q \pm \phi}{k}.$ I will need to further require that $\phi-q$ or $\phi+q$ must be divisible by $k$ which I am not sure how to proceed. I guess since there are infinite solutions, so it would be ok or due to some quadratic residue issue. I can not find the answer. Bu t I believe if (1) can not satisfied and then we can do for $-q$ then one can always make another true.

Due to some reason, I only need either (1) or (2) is true if one of them can not be satisfied. Any comments are very welcome. Thanks a lot.

  • 2
    $\begingroup$ $19^2=1+10\times36$, so we can take $q=19$, $k=36$. But $(ku\pm q)^2=1+r^2k=1+36r^2=1+(6r)^2$ requires two squares to differ by $1$, which doesn't happen (except for zero and one, which no value of $u$ achieves). $\endgroup$ Oct 30, 2021 at 5:18
  • $\begingroup$ Any thoughts on my comment? $\endgroup$ Oct 31, 2021 at 12:20
  • $\begingroup$ Ok, thanks a lot. You beat me. $\endgroup$
    – En-Jui Kuo
    Nov 1, 2021 at 13:37
  • $\begingroup$ Oh, so due to the theory of Pell's equation I mentioned above, I hope when k is not a perfect square, the conjecture will be true. I will keep thinking about it. $\endgroup$
    – En-Jui Kuo
    Nov 1, 2021 at 13:38


Your Answer

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

Browse other questions tagged or ask your own question.