L.S.,
Studying for my exam on algebraic number theory, I was thinking of writing down tricks for computing the class group of a number field fast. I thought of the following one, but I don't know if it is true. Could one maybe help me show whether it is or isn't?
Trick:
Let $K:\mathbb{Q}$ be a Galois extension of the degree $n$. Let $(p)$ split completely in $\mathcal{O}_K$ as $(p) = \prod_{i = 1}^n \mathfrak{p}_i$, with all $\mathfrak{p}_i$ of norm $p$. Now all $[\mathfrak{p}_i] \in CL(K)$ are generated by one $[\mathfrak{p}_j]$ for some $1 \leq j \leq n$.
The reason I suspect this might somehow be true, is because I already know that the Galois group acts transitively on a set of prime ideals lying over some $(p)$, and also because it happened in most class groups I computed so far.
Many thanks!
After some playing around with Sage, here is a counterexample.
Let $K=\mathbb Q(\sqrt[3]{11}, \zeta)$ where $\zeta$ is a cubed root of unity. Then $K$ is Galois with class group $C_2\times C_2$.
The prime $19$ splits completely in $K$. Two of the primes of $K$ lying above $19$ are $$(19, \zeta-\sqrt[3]{11}-2)\qquad\text{ and }\qquad(19, \zeta-\sqrt[3]{11}+9)$$
Neither of these primes are principal. Hence, if their ideal classes are to lie in a cyclic subgroup of $C_2\times C_2$ (which must be of order $2$), then both ideals must be in the same ideal class, and hence their product should be principal. However, their product is not principal.
Sage code:
and output: