Let's say we have the irreducible polynomial $f$ with roots $\alpha_1,\ldots,\alpha_n$. Now let $K$ be its splitting field, in other words $$K=\mathbb Q(\alpha_1,\ldots,\alpha_n).$$

When is it the case that $K=\mathbb Q(\alpha_i)$ for any $i$, i.e. when is the dimension of $K$ over $\mathbb Q$ equal to $n$, the degree of $f$? If this were true,then we would be able to express all the roots by rational combinations of each other. Is there a criterion that would decide whether an extension $\mathbb Q(\alpha)$ of $\mathbb Q$ has this property? Please explain your answers.

  • 1
    $\begingroup$ An equivalent condition is that $K$ must have cyclic Galois group of order $\deg(f)$. I'd be surprised if there was a nice, fully general, categorisation of what $\alpha$ or $f$ should look like. For example, it is not at all obvious at first glance that $\mathbb Q(\sqrt{2+\sqrt2})$ should be an example (and note that this answers your second question: no). There are things we can say: $f$ must be solvable by radicals, $\mathbb Q(\alpha)\subset\mathbb Q(e^{2i\pi/N})$ for some $N$ (Kronecker-Weber). It might help if you specified what kind of criterion you are after. $\endgroup$ – Mathmo123 May 4 '16 at 22:18
  • $\begingroup$ (I may have misunderstood what you meant by rational combinations: if you mean that every root is of the form $\frac{g(\alpha)}{h(\alpha)}$ for some rational polynomials $g,h$ then this is true by definition.) $\endgroup$ – Mathmo123 May 4 '16 at 22:34
  • $\begingroup$ @Mathmo123 i meant that when the splitting field can be generated by one element, then all its conjugates lie in Q(alpha).Also, I remember seeing something about a square discriminant having to do with this, but can't really recall. $\endgroup$ – Bogdan Simeonov May 5 '16 at 5:43
  • $\begingroup$ @Mathmo123 also, how does the proof that such an extension must have a cyclic galois group go? $\endgroup$ – Bogdan Simeonov May 5 '16 at 5:48
  • $\begingroup$ The splitting field can always be generated by one element (primitive element theorem, $\mathbb Q$ is perfect), but that element won't necessarily be a root of the same polynomial. Having square discriminant means the Galois group is a subgroup of $A_{\deg(f)}$. This doesn't help much when the degree is large. For your second question, the Galois group must be a transitive subgroup of $S_{\deg(f)}$ of order $\deg(f)$. $\endgroup$ – Mathmo123 May 5 '16 at 8:13

A sufficient condition is the Galois group $G$'s subgroups are all normal (e.g $G$ is abelian or $G=Q_8$).

Assume a monic polynomial $f(x) \in \mathbb Q[x]$ has Galois group $G$, and its splitting field is L. Choose a root $\alpha$ of $f(x)$ and consider $\mathbb Q(\alpha)$, which must be Galois over $\mathbb Q$ as all subgroups of $G$ are normal. Hence it contains every conjugation of $\alpha$ i.e the roots of $f$. Therefore it is just the splitting field L, so we have $[L:\mathbb Q]=[\mathbb Q(\alpha):\mathbb Q]=\operatorname{deg} f$.


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.