Let 𝑝 be an odd prime and assume $𝑥^2+ax+1$ is irreducible over the field $\mathbb{F}_p$. The polynomial function

$$\Psi:\mathbb{F}_p^2⟶\mathbb{F}_p,\quad (x,y)\mapsto 𝑥^2+𝑦^2−x+y-axy$$

is surjective, as proved here: Image of a polynomial function $x^2+y^2-x+y-axy$ over $\mathbb{F}_p$.

I would like to compute a set of representatives of the classes of the kernel of $\Psi$ (i.e., the relation $\ker(\Psi)=\{(x,y,t,w)\in \mathbb{F}_p^4, \, \Psi(𝑥,𝑦)=\Psi(t,w)\})$. So basically I would like to have an explicit set of $p-1$ elements of $\mathbb{F}_p^2$ that take on all the nonzero values in $\mathbb{F}_p$ when you apply $\Psi$. (Finding a solution to $\Psi(x,y)=0$ is obvious.)

Is there a way to do this in general, regardless of the value of $p$?


Here is a solution for $\frac{p-1}{2}$ of the level sets. Let $\varphi=x^2+y^2-x+y-axy$, and assume that $a\not=\pm 2$.

Following the suggestion in https://mathoverflow.net/q/356936 we first apply the change of variables given by $x= z+(a/2)y$ to obtain $$\varphi=z^2-z-b(y^2+(1+a/2)^{-1}y),$$ where $b=a^2/4-1$. Then letting $u=z-1/2$, $v=y+(2+a)^{-1}$, and $c=\frac{-1}{4}\left(\frac{1-a/2}{1+a/2}\right)$ we have $$\varphi=u^2-bv^2+c.$$

To compute the level sets you need to compute the fibres $\varphi^{-1}(d)$ for an image point $d$. If $d-c$ is a quadratic residue, then $d-c=e^2$ and so letting $s=u/e$ and $t=v/e$ we obtain: $$s^2-bt^2=1.$$

Having all solutions $(s,t)$ determines the solutions $(u,v)$, in turn the solutions $(y,z)$, and in turn the solutions $(x,y)$.

So it suffices to find the solutions to $s^2-bt^2=1.$ But such solutions are explicitly determined in:

Tekcan, Ahmet, The number of rational points on conics $C_{p,k}:x^2−ky^2=1$ over finite fields $\mathbb{F}_p$. Int. J. Math. Sci. 1 (2007), no. 2, 150–153.

MathSciNet Summary:

"Let $p$ be a prime number, $\mathbb{F}_p$ be a finite field, and let $k\in \mathbb{F}_p^*$. In this paper, we consider the number of rational points on conics $C_{p,k}: x^2−ky^2=1$ over $\mathbb{F}_p$. We prove that the order of $C_{p,k}$ over $\mathbb{F}_p$ is $p−1$ if $k$ is a quadratic residue mod $p$ and is $p+1$ if $k$ is not a quadratic residue mod $p$....''

I am not sure what happens when $d-c$ is a non-residue (I haven't tried), but since $\varphi$ is onto and $(p-1)/2$ of the non-zero values are quadratic residues, this gets you roughly half way.

  • 2
    $\begingroup$ I am surprised that Tekcan could publish this result, which is a standard exercise in number theory (for any diagonal conic over $\mathbb{F}_p$, in any number of variables). This was also discussed multiple times on this site, see e.g. my response for mathoverflow.net/questions/65183/… $\endgroup$ – GH from MO Sep 2 '20 at 21:58

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.