Assume that $K/\Bbb Q$ is a cyclic Galois extension, and $\mathfrak{p}$ a prime ideal of $K$ and $\sigma$ an element of the Galois group. What can we say about the classes $[\mathfrak{p}]$ and $[\sigma(\mathfrak{p})]$ in the ideal class group? Let $\left \langle [\mathfrak{p}] \right \rangle$ be the subgroup of the ideal class group generated by $[\mathfrak{p}]$. Do we have that necessarily $[\sigma(\mathfrak{p})] \in \left \langle [\mathfrak{p}] \right \rangle$? I cannot prove this. Also, I do not have any idea how to find a counter example.

  • $\begingroup$ you could have a look at [[math/9812171] Perfect forms and the Vandiver conjecture](arxiv.org/abs/math/9812171) $\endgroup$
    – Niels
    Jan 26, 2021 at 20:48
  • 1
    $\begingroup$ Look up genus theory. $\endgroup$ Jan 26, 2021 at 21:02
  • $\begingroup$ @Niels I can not see how the results in that paper are related to my question. Could you please give an exposition on it? $\endgroup$ Jan 26, 2021 at 21:20
  • 3
    $\begingroup$ @FranzLemmermeyer Could you please introduce a specific book or note or ... to me about this subject? If I am not mistaken, probably genus theory would deal with the cases when the Galois group acts trivially on the ideal class group. $\endgroup$ Jan 26, 2021 at 21:23
  • $\begingroup$ looks like you got interested: mathoverflow.net/questions/382930/… $\endgroup$
    – Niels
    Feb 3, 2021 at 15:13

1 Answer 1


This is true if $K$ is quadratic, since then $[\sigma(\mathfrak p)] = [\mathfrak p^{-1}]$.

It should be false for every other degree.

The only relation that the Galois action should satisfy is (for cyclic fields of degree $n$) that $ \prod_{i=0}^{n-1} [\sigma^{i}(\mathfrak p)] $ is trivial in the class group, since it's the class of the norm of $\mathfrak p$, which is an ideal of $\mathbb Q$ and thus is principal.

For example, for cubic fields, if this were true then you could never have $p$ congruent to $2$ mod $3$ dividing the order of the class group, as the Galois automorphism of order $3$ would have to act on elements of order $p$ by scalar multiplication by an element of $\mathbb F_p^\times$, and thus would have to be trivial, which is impossible since it would violate the above identity. So it suffices to find a cubic field whose class number is a multiple of $2,$ or $5,$ or ...

Searching quickly on the LMFDB, I found this counterexample, whose class number is $4$.

  • $\begingroup$ Your statements on quadratic extension and cyclic extensions are very clear. Also, I can check why your counterexample contradicts my claim (I can see it in another way), and I get the answer to my question. But I have another request, could you please tell more on the "Galois automorphism should have to act on elements of order $p$ by scalar multiplication by an element of $\mathbb F_p^\times$"? Can you write down the action of the Galois group? $\endgroup$ Jan 27, 2021 at 9:54
  • 1
    $\begingroup$ @NeoTheComputer No, I just mean that if $\sigma([\mathfrak p]) \in \langle [\mathfrak p \rangle$ then $\sigma([\mathfrak p]) = [\mathfrak p]^a$ for some $a$ mod $p$, because $ [\mathfrak p]$ has order $p$. $\endgroup$
    – Will Sawin
    Jan 27, 2021 at 14:50
  • $\begingroup$ Thank you for your response. $\endgroup$ Jan 27, 2021 at 14:53

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.