I'm in the middle of a proof where I'd like to show that $\sqrt{2 - \sqrt{2}} \notin \mathbb{Q}(\sqrt{2 + \sqrt{2}})$

The only way I can think of involves finding an explicit set representation for $\mathbb{Q}(\sqrt{2 + \sqrt{2}})$.

At first I tried showing $\mathbb{Q}(\sqrt{2 + \sqrt{2}}) = \{a + b\sqrt{2 + \sqrt{2}}: a,b\in\mathbb{Q}\}$

and then realised this is probably false, as it doesn't look like it contains $\sqrt{2}$.

I figured I could try $\mathbb{Q}(\sqrt{2 + \sqrt{2}}) = \{a + b\sqrt{2 + \sqrt{2}} + c\sqrt{2}: a,b,c\in\mathbb{Q}\}$

but this method seems really long-winded. I'm pretty sure there'll be plenty of shorter methods, but I don't know any method to show this.

Any pointers appreciated!

  • $\begingroup$ Wouldn't it be enough to find minimal polynomials for each number and then show that they have no common factors? $\endgroup$ – MJD Nov 15 '12 at 16:21
  • $\begingroup$ @MJD Don't they have the same minimal polynomial, $(x^2-2)^2-2$? $\endgroup$ – Thomas Andrews Nov 15 '12 at 16:24
  • $\begingroup$ @MJD: What would it imply if that were the case? $\endgroup$ – maliky0_o Nov 15 '12 at 16:26
  • 3
    $\begingroup$ I'm not even sure that this is true - isn't $\frac{2}{2+\sqrt{2}}=2-\sqrt{2}$, so $\frac{\sqrt{2}}{\sqrt{2+\sqrt{2}}} = \sqrt{2-\sqrt{2}}$? Since $\sqrt{2}\in\mathbb Q[\sqrt{2+\sqrt{2}}]$, your statement seems false. $\endgroup$ – Thomas Andrews Nov 15 '12 at 16:26
  • 1
    $\begingroup$ From my personal point of view, I would be happy to see this one not deleted, as I have learned something from it - but of course you are entitled to do as you prefer. $\endgroup$ – Old John Nov 15 '12 at 16:54

Actually, this statement is not true.

Using $(2+\sqrt{2})(2-\sqrt{2})=2$, taking square roots and dividing, we see that $$\sqrt{2-\sqrt{2}} = \frac{\sqrt{2}}{\sqrt{2+\sqrt{2}}}$$

So, since $\sqrt{2}\in\mathbb Q(\sqrt{2+\sqrt{2}})$, $\sqrt{2-\sqrt{2}}\in\mathbb Q(\sqrt{2+\sqrt{2}})$.

  • $\begingroup$ Ah, thanks. That makes it easier to see how you derived $\sqrt{2 - \sqrt{2}} \in \mathbb{Q}(\sqrt{2 + \sqrt{2}})$. But this trick wouldn't work for, say, telling if $\sqrt{3 - \sqrt{7}} \in \mathbb{Q}(\sqrt{3 + \sqrt{7}})$, since $(3 + \sqrt{7})(3 - \sqrt{7}) = 2$ doesn't give you anything. $\endgroup$ – maliky0_o Nov 16 '12 at 12:13
  • $\begingroup$ True, but (1) that wasn't the problem given, and (2) this indicates that you can't just make a blanket statement blind about expression of this sort. You have to use something specific about the numbers. $\endgroup$ – Thomas Andrews Nov 16 '12 at 13:47
  • $\begingroup$ For example, it becomes clear that, if $a\neq b$, then $\sqrt{\sqrt{a}-\sqrt{b}}\in \mathbb Q\left(\sqrt{\sqrt{a}+\sqrt{b}}\right)$ if and only if $\sqrt{a-b}\in \mathbb Q\left(\sqrt{\sqrt{a}+\sqrt{b}}\right)$. $\endgroup$ – Thomas Andrews Nov 16 '12 at 14:13

$\begin{eqnarray}{\rm\bf Hint}\ \ w-1 && =\ \ (w+1)\,(w-1)^2\ \ \ {\rm by}\ \ \ w^2-1 = 1\ \ \ {\rm for}\ \ \ w := \sqrt{2} \\ \Rightarrow\ \ \ \ 2-w && =\ \ (2+w)\, (w-1)^2\ \ \ {\rm by}\ \ \ w * {\rm prior,\ \ using}\,\ \ w^2 = 2 \\ \Rightarrow\ \sqrt{2-w} && =\, \sqrt{2+w}\,\ (w-1) \ \ \ \ \ {\rm by\ taking\ \sqrt{prior}} \\ \Rightarrow\ \sqrt{2-w} && \in\Bbb Q(\sqrt{2+w}) = {\rm R}\ \ \ \ \ \ {\rm by}\ \ \sqrt{2+w}^{\,2}\! =\, 2+w\in {\rm R}\:\Rightarrow\: w \in\rm R \end{eqnarray} $


These expressions show up when computing the sine and cosine of $\pi/8 = 22.5^\circ$. Since the corresponding extensions are cyclotomic, hence normal (and even abelian), any extension of the automorphism sending $\sqrt{2}$ to $-\sqrt{2}$ must fix the extension.


Let $\alpha = \sqrt{2 + \sqrt{2}}$ and $\beta = \sqrt{2 - \sqrt{2}}$. I want to find out if $\mathbb Q(\alpha)$ is equal to $\mathbb Q(\beta)$ or not. They are equal but I was wondering about proving it with Galois theory instead of direct computation.

Since $\alpha,-\alpha,\beta,-\beta$ are the roots of the polynomial $P(x) = x^4 - 4x^2 + 2$ we could try building the splitting field of $P$ (which has degree 4 since the polynomial is irreducible with degree 4) by a chain of extensions like this:

enter image description here

the bottom ones all have degree 2 and the full tower has degree 4, so I think that implies the greyed out ones at the very top have degree 1 so it's a trivial extension, proving $\mathbb Q(\alpha) = \mathbb Q(\beta)$.

  • $\begingroup$ Consider the extension $S/\mathbb{Q}(\sqrt{2})$ instead. It is a degree $2$ extension by Galois theory/tower of fields. Now $\mathbb{Q}(\alpha)\neq \mathbb{Q}(\sqrt{2})$, similarly for $\mathbb{Q}(\beta)$ (since they are degree $4$ extensions of $\mathbb{Q}$ whereas $\mathbb{Q}(\sqrt{2})$ is of degree $2$. Hence both equal $S$. $\endgroup$ – fretty Nov 16 '12 at 8:02
  • $\begingroup$ @sperners-lemma: Is there a result that tells you that the degree of a splitting field of a polynomial p is the degree of p if p is irreducible? $\endgroup$ – maliky0_o Nov 16 '12 at 12:07
  • $\begingroup$ @maliky0_o, i guess there isn't there I don't know how to fix this... $\endgroup$ – sperners lemma Nov 16 '12 at 12:40
  • $\begingroup$ all we know is degree of splitting field of degree n polynomial divides n! $\endgroup$ – sperners lemma Nov 16 '12 at 13:06

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.