Fix a number field $k$ and some $\alpha \in \mathcal{O}_k \setminus \mathbb{Z}$. Let $p$ be a rational prime which splits completely in $k$, so that $p \mathcal{O}_k = P_1 \cdots P_m$ for $m = [k : \mathbb{Q}]$ and some prime ideal $P_1, \ldots, P_m$ of $\mathcal{O}_k$. Now it is easy to see that for any $i=1,\ldots,m$ there exists an unique integer $0 \leq \alpha_i < p$ such that $\alpha \equiv \alpha_i$ (mod $P_i$). For no reason we can say that $\alpha_1, \ldots, \alpha_m$ are all distinct integers, indeed this is not certainly the case if $p < m$.

My question is: Is it true that for "most" of the primes $p$ which split completely in $k$ we have that $\alpha_1, \ldots, \alpha_m$ are all distinct?

Here by "most", I mean, for example, a set of relatively positive density, or even density $1$... I let to you give a precise meaning.

My guess is that the answer should be YES. Indeed, assuming that each $\alpha_i$ is a random non-negative integer less than $p$, the probability that $\alpha_1, \ldots, \alpha_m$ are all distinct is $$\left(1 - \frac1{p - 1}\right)\left(1 - \frac1{p - 2}\right)\cdots\left(1 - \frac1{p - m + 1}\right) \to 1 ,$$ as $p \to +\infty$. Moreover, in the special case $m = 2$, we have that $\alpha_1 = \alpha_2$ means that $\alpha \in \alpha_1 + p \mathcal{O}_k \subseteq \mathbb{Z} + p \mathcal{O}_k$ and it seems that the set $\mathbb{Z} + p \mathcal{O}_k$ is "small" for large $p$, so $\alpha$ should be not in it.

Thanks for any idea/reference.


Let $f\in\mathbb{Z}[X]$ be the characteristic polynomial of $\alpha$. If the reduction of $f$ mod $p$ has any repeated factors in $\mathbb{F}_p[X]$, then $p$ divides the discriminant of $f$. This is true regardless whether all the factors are linear or not.

Therefore the $\alpha_i$ will be distinct for all but finitely many of the completely split $p$ when $\alpha$ generates $k$ over the rationals (in which case its characteristic polynomial is also its minimal polynomial).

But no $p$ at all will work when $\alpha$ lies in a proper subfield between $k$ and $\mathbb{Q}$, since $f$ is then a proper power of the minimal polynomial.

  • $\begingroup$ Please can you expand a bit? In particular the last sentence: Do you mean that in such a case $\alpha_1 = \cdots = \alpha_m$? $\endgroup$
    – sercej
    Oct 26 '16 at 13:52
  • $\begingroup$ They won't necessarily all be equal, but each value will necessarily occur repeatedly, with multiplicity at least the degree of $k$ over $\mathbb{Q}(\alpha)$. $\endgroup$
    – GNiklasch
    Oct 26 '16 at 14:35
  • $\begingroup$ I still don't get why this should happened. What is the link between the $\alpha_i$'s and the characteristic polynomial? $\endgroup$
    – sercej
    Oct 26 '16 at 14:56
  • $\begingroup$ The $\alpha_i$ are the roots of the reduction mod $p$ of the characteristic polynomial. $\endgroup$
    – GNiklasch
    Oct 26 '16 at 15:10
  • $\begingroup$ OK, thanks. Do you have a reference for this fact? $\endgroup$
    – sercej
    Oct 26 '16 at 22:30

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.