4

I have a question regarding an algorithm for finding the ring of algebraic integers $\mathfrak{O}_K$ in an algebraic number field $K$. The algorithm that I have been using is as follows:

1) Take a $\Bbb{Q}$-basis $w = \{ w_1, ..., w_n\}$ with $w \subset \mathfrak{O}_K$. Let $M$ be the $\Bbb{Z}$ span of $w$, so $M \subset \mathfrak{O}_K$.
Calculate $\Delta[M]^2=\Delta[w_1, ..., w_n]^2$ - the usual discriminant.

2) If $\;[\mathfrak{O}_K:M]=m\quad$then $\quad|\Delta(M)^2| = m^2|\Delta(\mathfrak{O}_K)^2|$
If $|\Delta(M)^2|$ is squarefree then we know that $m=1$ and $w$ is a $\Bbb{Z}$ basis for $\mathfrak{O}_K$
Otherwise we check for each prime $p$ with $p^2$ dividing $|\Delta(M)^2|\;$ we look for $\alpha\in\mathfrak{O}_K$ of the form $\alpha = \frac1p\sum_jc_jw_j$ where $c_j\in\Bbb{Z}$ not all divisible by p. We can multiply through by an integer to make the $c_k$ which is not divisible by $p$ by 1. Clearly only need to check co-efficients $0\le c_j \lt p$.
If we find such an $\alpha$ with $c_k$ not equal to zero, we replace $w_k$ with our $\alpha$ to get another $\Bbb{Q}$-basis, $M'$ and our discriminant is divided by $p$.

3) We repeat with our new basis $M'$. If such an $\alpha$ doesn't exist then $p$ doesn't divide $\;[\mathfrak{O}_K:M]=m\quad$ -we use a theorem of Cauchy on finite groups here I believe.

My question is as follows, if we check a prime $p_1$ for basis $M$ and we find no such $\alpha$ of the form $\alpha = \frac{1}{p_1}\sum_jc_jw_j$ where $c_j\in\Bbb{Z}$ not all divisible by $p_1$, -so we move to the next prime $p_2$ with $p_2^2$ dividing $|\Delta(M)^2|\;$; if we find such an $\alpha$ and get a new basis $M'$, do we have to re-check all the possible $\alpha$'s for prime $p_1$ in our new basis $M'$ again??

Thanks

SEWillB
  • 1,701
  • I do start with algebraic integers as each $w_i$ is an algebraic integer and $M$ is the $\Bbb{Z}$-linear span of the $w_i$ -But I also don't believe that this is difficult as we can multiply each $w_i$ by a sufficient $n\in\Bbb{N}$ to put it in $\mathfrak{O}_K$ and the $\Bbb{Q}$-linear span stays the same. – SEWillB May 02 '17 at 23:59
  • I missed $w_i \in \mathfrak{O}_K$. It is not difficult once you know the $\sigma_n$ (embeddings $K \to \mathbb{C}$) I'd say. – reuns May 03 '17 at 00:03
  • I'm sorry I'm not sure how knowing the embeddings will help here; maybe in each specific case they might but I'm asking really about the procedure in its generality. I am asking whether once we discover that a prime $p$ doesn't divide $[\mathfrak{O}_K:M]$ whether or not we have to retry this prime when $p^2$ divides $|\Delta(M')^2|$ for our new generated $\Bbb{Q}$ basis $M'$ in the next iteration... – SEWillB May 03 '17 at 13:11
  • 1
    http://gaetan.chenevier.perso.math.cnrs.fr/MAT552/TAN_cours5.pdf look at page $82$ :) – Maman May 12 '17 at 13:47
  • @Maman The link is dead. – TheOutZ Aug 29 '24 at 22:48
  • 1
    @TheOutZ http://gaetan.chenevier.perso.math.cnrs.fr/TAN_chenevier.pdf – Maman Oct 02 '24 at 10:19

0 Answers0