4
$\begingroup$

Let $K=\mathbb Q(\xi_{39})$ be the 39-th cyclotomic field. Pari-GP told me that the prime ideals above $3$ and $13$ are not principal. Is there a way to prove that by hand (no computation made by computer)

$\endgroup$
6
  • $\begingroup$ There are no prime ideals above 39. You meant prime ideals over 3 and 13. Can you compute the class number by hand? $\endgroup$
    – KConrad
    Feb 7, 2015 at 19:12
  • 4
    $\begingroup$ If you are willing to accept a table entry from a book without checking it yourself, why not accept what a computer tells you too? :) Since you believe $h = 2$, if two ideals are not principal then their product must be principal, so check first that there is an element whose norm is the product of norms of primes over 3 and 39 (norms from a cyclotomic field down to $\mathbf Q$, other than $\mathbf Q$ itself, are never negative). That would show you do not have to check that primes over both 3 and 39 are not principal, but just over one of them. $\endgroup$
    – KConrad
    Feb 7, 2015 at 19:59
  • 1
    $\begingroup$ Here is a way: Assume the prime ideal to be principal say generated by a cyclotomic integer $\alpha$. Calculate the norm of the prime above. Then the $|N(\alpha)|$, is also norma of a suitable integer in a quadratic subfield. Choose an imaginary quadratic subfield in $Q[\zeta_{39}]$, where the norm is a positive definite quadratic form; for a binary form it should be easy to check if it assumes a specific value or not. $\endgroup$ Feb 8, 2015 at 0:53
  • 1
    $\begingroup$ @Vanchinathan:Why is |N(α)| norm of a suitable integer in a quadratic subfield? Moreover, there is only one quadratic subfield in a cyclotomic field. So how do you choose it? $\endgroup$
    – joaopa
    Feb 8, 2015 at 2:18
  • 1
    $\begingroup$ @joaopa, the norm is transitive, so if you have a tower of finite extensions $L/K/F$ then for $\alpha \in L$ we have ${\rm N}_{L/F}(\alpha) = {\rm N}_{K/F}({\rm N}_{L/K}(\alpha))$. Thus the norm from $L$ to $F$ of any element of $L$ is also the norm from $K$ to $F$ of a related element of $K$ for any field $K$ between $L$ and $F$. (Your previous comment says there is only one quadratic subfield in a cyclotomic field, but surely you meant "possibly more than one" in general, since the Galois group $({\mathbf Z}/n)^\times$ often has more than one subgroup of index two.) $\endgroup$
    – KConrad
    Feb 8, 2015 at 18:07

1 Answer 1

10
$\begingroup$

The quadratic number field $k = {\mathbb Q}(\sqrt{-39})$ has a cyclic class group of order $4$. Its genus field is $K = {\mathbb Q}(\sqrt{-3},\sqrt{13})$, which has class number $2$. The Hilbert class field of $K$ is dihedral over ${\mathbb Q}$, hence not contained in the cyclotomic field $L = {\mathbb Q}(\zeta_{39})$. This already shows that the class number of $L$ is even.

Now verify that one of the prime ideals above $13$ in $K$ is not principal (obvious since its norm down to $k$ is not principal) and that it ramifies completely in $L/K$.

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.