Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Consider the following property of subfields ${\mathbb K}$ of ${\mathbb C}$ :

$$ \text{Any polynomial of degree 3 with coefficients in} \ {\mathbb K} \ \text{has a root in } {\mathbb K} \ \ \ \ (*) $$

Thus, $\mathbb R$ itself satisfies (*), but it is easy to see that there are many other subfields with this property.

If ${\mathbb K} \subseteq {\mathbb R}$ satisfies $(*)$, does ${\mathbb K}(i)$ satisfy $(*)$ also ?

This question was inspired to me by that other one.

share|cite|improve this question
Many other subfields? The only other subfield that comes to mind is the real algebraic numbers. What did you have in mind? – JSchlather Jan 21 '13 at 6:55
@JacobSchlather Given any coutable field $\mathbb T$, we can construct a tower of degree 3 extensions of $\mathbb T$, whose union $F({\mathbb T})$ is again countable and satisfies (*). So we see that there are at least countably many such subfields. – Ewan Delanoy Jan 21 '13 at 7:40
Since my answer doesn't work and this question hasn't gotten any more attention. You might consider posting this problem on mathoverflow. – JSchlather Jan 23 '13 at 4:58
@JacobSchlather I believe the problem in your answer can be fixed, but I'm working on other problems right now. I'll post more about this when I have the time. – Ewan Delanoy Jan 23 '13 at 5:17
up vote 2 down vote accepted

I have a counter-example.

Consider the integer polynomial

$$ f(x) = (x^3 + x + 1)^2 + 1 $$

$f$ is irreduible over $\mathbb{Q}$ and factors into two cubic polynomials over $\mathbb{Q}(i)$. Let $g(x)$ be one of those polynomials$.

Using sage, I compute that the splitting field $E$ of $f$ is degree 72: exactly what you would expect due to the factorization $f = g \bar{g}$. Of particular note is that the Galois group has no subgroups of order 24, and thus the splitting field contains no cubic subfields.

Now suppose $K$ is any field constructed by starting with $\mathbb{Q}$ and iteratively making cubic extensions (transfinite iteration, unless you carefully order the extensions).

Since $K$ has no degree 3 extensions, every cubic polynomial over $K$ has a root in $K$. Also, every finitely generated subfield $L \subseteq K$ satisfies $[L:\mathbb{Q}] = 3^n$, and contains a subfield of degree 3 over $\mathbb{Q}$

In particular, let $L = E \cap K$. Because $E$ has no subfields of degree $3$, we must have $E \cap K = \mathbb{Q}$. However, this would imply $E \cap K(i) = \mathbb{Q}(i)$, and thus $g$ is irreducible over $K(i)$.

share|cite|improve this answer
Thanks for your answer. Can you say more about the Galois group ? Its structure is S3 x S3 x Z/2Z, right ? It would be nice also if you explained in a little bit more detail why it has no subgroup of index 3 (a sage command is fine). – Ewan Delanoy Jan 26 '13 at 9:19
$Z/2$ comes in as a semidirect product. It's the subgroup of $S_6$ generated by the symmetric group on 123, the symmetric group on 456, and by the permutation $(14)(25)(36)$. After constructing $f$, the command to see there were no index 3 subgroups was G = f.galois_group() and then listing for A in G.subgroups(): print len(A). – Hurkyl Jan 26 '13 at 13:16
By the way, you might be interested in a related question I asked here recently :… – Ewan Delanoy Jan 26 '13 at 16:37
@Ewan: I am; the very same question had went through my mind. – Hurkyl Jan 26 '13 at 21:34
So the exact place where I got stuck is why it's not true. Interesting. I couldn't decide whether or not my approach was too abstract and it actually did work in the specific case of $K(i)$. – JSchlather Jan 27 '13 at 16:11

Let $K$ be a perfect field such that every cubic polynomial over $K$ has a root in $K$. Let $i \in \overline{K}$ be a solution to $x^2+1 \in K$, if $i \in K$ there is nothing to prove so we assume not. We claim every cubic polynomial over $K(i)$ also takes a root. Suppose not then we can find cubic polynomial $p(x) \in K(i)[x]$ such that $p(x)$ is irreducible over $K(i)[x]$. Let $L$ be the splitting field of p(x), then $L$ is also an extension field of $K$. In particular $[L:K]=6,12$. In both cases the Sylow 2-subgroup of $\mathrm{Gal}(L/K)$ has index $3$, so its fixed field is a degree $3$ extension of $K$. This is impossible, so every cubic polynomial over $K(i)$ has a root. Notice that this proof works for any degree $2$ extension of $K$.

It seems likely to me that such a field does not have any extensions of degree divisible by $3$. But I don't think it's true that every group whose order is divisible by $3$ has a subgroup of index $3$, so this technique will likely not generalize.

Edit: It was pointed out in the comments that this approach is incorrect because $L/K$ need not be Galois. So you have to look at normal closure $E$ of $L$ over $K$. In order to salvage this approach in this case one would need to show that every transitive subgroup of $S_6$ whose order is divisible $3$ and $2$ has a subgroup of index $3$. I'll leave this up in case its useful to anyone else attempting to solve the problem.

share|cite|improve this answer
Right. You can kill two birds with one stone : I believe this answer will also fit in the question linked in my OP (a bountied question by the way). – Ewan Delanoy Jan 21 '13 at 7:23
@EwanDelanoy Hmm. The question linked is at a part of Stewart's book that hasn't talked about field extensions. So the OP there probably wouldn't appreciate a proof using Galois theory too much. – JSchlather Jan 21 '13 at 7:33
My mistake, the book is about Galois theory though. – Ewan Delanoy Jan 21 '13 at 7:37
@JacobSchlather : the extension $L/K$ may not be Galois. You have to take the splitting field of $x$ over $K$, but this field is of degree dividing $6!=720$. – user10676 Jan 21 '13 at 10:35
@user10676, Good point. This method probably isn't salvageable then. – JSchlather Jan 21 '13 at 17:01

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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