0
$\begingroup$

I'm having a bit of trouble understanding an example from the introduction to galois theory section from Gallian's text.

We have that $\omega = -1/2 + i\sqrt 3/2$ satisfies the equations $\omega^3 = 1$ and $\omega^2 + \omega + 1 = 0$. We now consider the extension $Q(\omega,2^{1/3})$ of $Q$.

I understand that any automorphism of a field containing Q must fix Q, and since the basis of the extension field will involve powers and products of $\omega$ and $2^{1/3}$, we just need to find out where the automorphisms send those elements.

What I don't understand is how we determine that there are six possible automorphisms and where they send these elements. The automorphisms he lists in the text send

$\omega \to \omega$ or $\omega^2$ and $2^{1/3} \to 2^{1/3}$ or $\omega(2^{1/3})$ or $\omega^2(2^{1/3})$.

I'm just having a bit of trouble seeing how he determined where the automorphism sends the elements, I suspect it has something to do with the basis of the extension? But I'm not sure.

In a previous example when we had the extension of $Q(\sqrt 2)$ over $Q$, the automorphisms could send $\sqrt 2$ to $\pm \sqrt 2$ since

$2 = \phi(2) = \phi(\sqrt 2 \sqrt 2) = (\phi(\sqrt 2))^2$

In this case the basis of the extension is $\{1,\sqrt 2\}$, so we're not just sending the basis element $\sqrt 2$ to another basis element, rather its conjugate.

I'm guessing there's a fundamental misunderstanding I'm having about determining the automorphisms.

Thanks

$\endgroup$
  • 1
    $\begingroup$ Any automorphism sends an element to its conjugates. That's how author is determining those automorphism. Also if you are familiar with Galois extensions and splitting fields , in this case you can determine the order of group of automorphisms (Galois group) without explicitly calculating the maps. $\endgroup$ – iron feliks Apr 21 '16 at 18:09
  • $\begingroup$ As @skmehta says, $\omega$ must go to a root of $X^2+X+1$, and your chosen cube root of $2$ (you’ve called it $2^{1/3}$) must go a root of $X^3-2$. There’s your six members of the Galois group. $\endgroup$ – Lubin Apr 21 '16 at 19:45
1
$\begingroup$

The extension $\Bbb Q(\sqrt[3]{2},\omega)$ is the splitting field of the irreducible polynomial $x^3 - 2 \in \Bbb Q[x]$. It is not hard to see that:

$x^3 - 2 = (x - \sqrt[3]{2})(x - \sqrt[3]{2}\omega)(x - \sqrt[3]{2}\omega^2)$

that is, each of: $\sqrt[3]{2},\sqrt[3]{2}\omega,\sqrt[3]{2}\omega^2$ is a cube root of $2$.

Part of the behavior of the Galois group of $x^3 - 2$ has to do with the "threeness of three", an odd degree rational polynomial has either one, or three real roots. This is because complex-conjugation (the algebraic closure of $\Bbb Q$ is a subfield of $\Bbb C$) is a field automorphism of $\Bbb C$ that fixes the reals (and thus the rationals).

In this case, we have the partial factorization over $\Bbb Q(\sqrt[3]{2})$:

$x^3 - 2 = (x - \sqrt[3]{2})(x^2 + \sqrt[3]{2}x + \sqrt[3]{4})$

and we have "two types" of automorphisms of $\Bbb Q(\sqrt[3]{2},\omega)/\Bbb Q$: ones that send all three roots of $x^3 - 2$ "in a circle" (cycle), and ones that swap just two of them (with a quadratic only the second type is possible).

Since we have 3 roots, we have "some" subgroup of $S_3$, and since this group has a 3-cycle AND a two-cycle (we know it has a two-cycle, because the restriction of complex conjugation to $\Bbb Q(\sqrt[3]{2},\omega)$ is obviously an automorphism of $\Bbb Q(\sqrt[3]{2},\omega)/\Bbb Q$ of order 2), it must be $S_3$. From here it's just a matter of "back-figuring" what each permutation of the roots must do to the generators. That is:

$\text{Gal}(\Bbb Q(\sqrt[3]{2}/\Bbb Q) = \langle \sigma, \tau\rangle$ where:

$\sigma: \sqrt[3]{2}\mapsto \sqrt[3]{2}\omega \mapsto \sqrt[3]{2}\omega^2 \mapsto \sqrt[3]{2}$

$\tau: \sqrt[2]{2}\omega \mapsto \sqrt[3]{2}\omega^2 \mapsto \sqrt[3]{2}\omega$.

The roots of $x^3 - 2$, when plotted on the complex plane, form the vertices of an equilateral triangle, making it appear less mysterious that their symmetry group is the symmetry group of said triangle.

$\endgroup$

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.