We know that for any $n$ relatively prime to $p$, $\mathbb{Q}_p(\zeta_n)$ is an unramified extension over $\mathbb{Q}$. On the other hand, we know that finite unramified extensions of a field complete with respect to a discrete valuation corresponds to finite extensions of its residue field. The residue field of $\mathbb{Q}_p$ is $\mathbb{F}_p$, and we know the finite extensions of $\mathbb{F}_p$ are $\mathbb{F}_{p^r} = \mathbb{F}_p(\zeta_{p^r-1})$, so the finite unramified extensions of $\mathbb{Q}_p$ should be $\mathbb{Q}_p(\zeta_{p^r-1})$.

Is it true that for any $n$ such that $(n,p) = 1$, there is an $r$ such that $\mathbb{Q}_p(\zeta_n) = \mathbb{Q}_p(\zeta_{p^r-1})$?


Yes, it’s true. Let the residue field of the unramified extension be $\Bbb F_{p^r}=\Bbb F_p(\zeta_{p^r-1})$. The primitive $(p^r-1)$-th root of unity in the residue field can be lifted to characteristic zero — there are any number of ways of seeing this, like Hensel, or the neat trick of lifting $\zeta_{p^r-1}\in\Bbb F_{p^r}$ to any $z$ in your unramified field, and then repeatedly taking $p^r$-th powers. The sequence you get is $p$-adically convergent.

  • $\begingroup$ But then is it not true that $\mathbb{Q}_p(\zeta_{p^r-1})$ are all the unramified extensions of $\mathbb{Q}_p$? $\endgroup$ – Aaron Johnson Feb 5 '18 at 5:44
  • 1
    $\begingroup$ All the finite unramified extensions, yes. There’s one of each degree. $\endgroup$ – Lubin Feb 5 '18 at 5:46
  • 1
    $\begingroup$ But if it's not true that for any $(n,p) = 1$, $\mathbb{Q}_p(\zeta_n) = \mathbb{Q}_p(\zeta_{p^r-1})$ for some $r$, it seems like this gives a contradiction that all finite unramified extensions are of the form $\mathbb{Q}_p(\zeta_{p^r-1})$? $\endgroup$ – Aaron Johnson Feb 5 '18 at 6:19
  • $\begingroup$ @AaronJohnson, you may be forgetting that the $n$th cyclotomic polynomial may easily fail to be irreducible over $\mathbb Q_p$, so that $\mathbb Q(\zeta_n)$ becomes ambiguous. The question is then about the degree(s) of the irreducible factors. $\endgroup$ – paul garrett Feb 5 '18 at 19:10
  • $\begingroup$ @paulgarrett: good point! Thanks! $\endgroup$ – Aaron Johnson Feb 6 '18 at 0:23

The proof you give is correct. Let me try to exhibit why we should expect such a thing.

The field of complex numbers $\mathbb C$ is the unique finite extension of $\mathbb R$ and that $$ \mathbb R(i) = \mathbb R(\zeta_4) = \mathbb C. $$

Roots of unity are not real except for $\pm 1$. Hence $$ \mathbb R(\zeta_n) = \mathbb R(\zeta_4) =\mathbb C \qquad \text{if $n>2$} $$ and $$ \mathbb R(\zeta_n) = \mathbb R \qquad \text{if $n\leq2$} $$ since there is no more room between $\mathbb R$ and its algebraic closure.

Similar things happen in $\mathbb Q_p$, take $p=5$. By Hensel's lemma roots of unity are not in $\mathbb Q_5$ except for $\pm 1, \pm i$. In particular $$ \mathbb Q_5(\zeta_n) = \mathbb Q_5 \quad \Longleftrightarrow\quad n\mid 4. $$ Now since $\bar{\mathbb Q}_5/\mathbb Q_5$ is an infinite extension there is more room for intermediate extensions but still diferent roots of unity might define the same extension. For instance $\mathbb Q_5(\zeta_{5^2- 1})$ contains $\mathbb Q_5(\zeta_8)$ and $\mathbb Q_5(\zeta_3)$. You can check that $$ \mathbb Q_5(\zeta_{5^2- 1})=\mathbb Q_5(\zeta_8)=\mathbb Q_5(\zeta_3). $$

  • $\begingroup$ Good example. I was thinking of using just that, but I wouldn’t have stated it as succinctly as you have here. $\endgroup$ – Lubin Feb 5 '18 at 22:56
  • $\begingroup$ I really appreciate your comment. $\endgroup$ – eduard Feb 6 '18 at 0:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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