Along the lines of the general question "How much does the discriminant of a number field reveal about the field?", I was wondering how often it happens that the discriminant of some number field uniquely identifies the Galois group of (the Galois closure over $\mathbb{Q}$ of) that field. Of course not always: there are $S_n$-extensions whose discriminant is the same as that of some quadratic field; on the other hand, certain small discriminants (such as -3, which must belong to the group $C_2$) can only occur once by standard discriminant bounds. Intuitively, I would think that such unique discriminants are pretty common, so:

Q1: Is there a reasonable heuristic for the asymptotic of, e.g., number of degree-$n$ number fields whose discriminant suffices to identify the Galois group? Is it even clear that such fields exist for all $n$?

I was also surprised by the following easy example: If $F/\mathbb{Q}$ is a tamely ramified quartic $A_4$-extension, then its discriminant can NEVER identify the Galois group! Namely, this discriminant would have to be a square; if it involves more than one prime factor, then some $C_2\times C_2$-extension has the same discriminant; but if it has only one prime factor, then this prime must be $\equiv 1$ mod $3$, and so there is a $C_3$-extension with the same discriminant!

Q2: Are there other examples as this, maybe infinite families, of permutation groups $G$ that can never be recognized from the discriminant of a (possibly, tame) $G$-extension?

  • $\begingroup$ Take $F_1/\mathbb{Q},F_2/\mathbb{Q}$ two (tame) extensions with same discriminant $d$ but different Galois groups $G_1,G_2$. Then for any (tame) Galois extension $L/\mathbb{Q}$ of group $G$ with discriminant $e$ coprime to $d$, the (tame) extensions $LF_1//\mathbb{Q}$ and $LF_2/\mathbb{Q}$ have same discriminant $ed$ but non isomorphic Galois groups $G\times G_1$ and $G\times G_2$ (it is a classical result that these groups are isomorphic iff $G_1\simeq G_2$). Hence you have infinitely many examples of discriminants which cannot determine Galois groups. Did I misunderstand the question ? $\endgroup$
    – GreginGre
    Commented Aug 21, 2018 at 9:51
  • $\begingroup$ @GreginGre: This gives examples of particular discriminant values which are not unique. This happens a lot (for suitable values). What I want in Q2 is examples of groups where there is not even a single discriminant value which is unique. The $A_4$-example is essentially due to a group-theoretical condition (roughly speaking, the group is somehow ``dominated" by the two abelian groups $C_3$ and $C_2\times C_2$), which could a priori work for infinitely many other groups, but I haven't found any other example so far. $\endgroup$ Commented Aug 21, 2018 at 19:02


Your Answer

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

Browse other questions tagged or ask your own question.