Given a single-variable polynomial, we all know that the number of its roots is bounded in terms of its degree. A polynomial here is a polynomial with integer coeffecients, and a root of a polynomial is a real root.

Given a multivariate polynomial, this phenomenon, in its general form, stops working. However, by a well-known theorem by Bezout, we still can say that either it has an infinite number of roots, or it has only a finite number which is bounded above in terms of its degree.

My questions is about the number of integer roots: Does there exist a functions $B=B_n:\mathbb N\rightarrow \mathbb N$ such that given a polynomial $P(x_1,...,x_n)$ of degree $d$, we can say that the set of INTEGER roots of $P$ contains at most $B(d)$ elements, or otherwise it is infinite?

Thanks in advance.

  • $\begingroup$ Sorry, what's the specific reference for that theorem of Bezout? I'm not familiar with that version. $\endgroup$
    – KReiser
    Jun 25 at 18:11
  • $\begingroup$ It considers a system of polynomial equatinons, and claims that the number of solutions, if finite, then bounded in terms of the degrees. Bézout's theorem $\endgroup$ Jun 25 at 18:46
  • 1
    $\begingroup$ I'm familiar with that form of Bezout, but I don't see how you're getting the claim that a multivariable polynomial has either infinitely many integer roots or finitely many integer roots and the number of such roots is bounded above by a function of the degree. $\endgroup$
    – KReiser
    Jun 25 at 18:50
  • $\begingroup$ Given a multivariate polynomial $P(x_1,...,x_n)$ of degree $d$ and with a finite number $m$ of zeros, I applied Bezout's theorem on the system consisting of $n$ copies of the equation $P(x_1,...,x_n)=0$. Since the solutions of this system are those of $P$, I concluded that $m\leq d^n$. But I realized now that this is actually a wrong use of the theorem, because the solutions of the system has to be counted over the algebraic closure, which is the complex numbers in this case. But then, $P$ must have an infinite number of zeros (assuming $n>1$). $\endgroup$ Jul 1 at 5:29
  • $\begingroup$ There's more wrong with your attempt to use Bezout there than just needing to work over the algebraic closure - taking $n$ copies of $P=0$ doesn't do anything new past considering $P=0$! Bezout tells you information about nontrivial intersections, but the scenario you describe is far from a nontrivial intersection. And yes, it is well-known that $P$ must have infinitely many zeroes (over an algebraically closed field) when $n>1$ - the interesting, tricky part is how many of those are integral (or belong to some smaller field, etc). Glad you got a good answer below! $\endgroup$
    – KReiser
    Jul 1 at 5:42

1 Answer 1


There is no such bound.

Given an equation $X$ over $\mathbf{Q}$ with infinitely many rational points, you can always scale the coefficients so that any finite subset of the rational points all become integral points. (Put them all under a common denominator $N$, then replace each variable $x$ by $x/N$.) Take an elliptic curve $E$ with positive rank, that is, $E(\mathbf{Q})$ is infinite. (They exist: for example, $E: y^2 = x^3 - 2$.) After scaling, you can find an elliptic curve with as many integral points as you like. (You can take $y^2 = x^3 - 2 N^6$ for the greatest common denominator $N$ of the finite set of rational points.) If such a bound in your question existed, then, since these curves all have the same degree, at some point these scaled elliptic curve would have to have infinitely many integral points. But $E(\mathbf{Z})$ is always finite for any elliptic curve by a theorem of Siegel.


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.