Let $K = \mathbb{Q}(\sqrt{2},\sqrt{5},\sqrt{7})$: I know that $K/F$, with $F =\mathbb{Q}$, is Galois and has degree 8. By the primitive element theorem there is a a $\theta$ such that $K = F(\theta)$, in this case it could be a linear combination given by $\sqrt{2}+\sqrt{5}+\sqrt{7}$. My question is: What is the degree $d$ of the minimal polynomial for $F(\theta)$. Show that $d$ is independent of $\theta$.

The minimal polynomial should have degree 8 as $[F(\theta):\mathbb{Q}] = deg(m_{\theta}(x))$. How do I prove the independence?


  • 1
    $\begingroup$ Note that you didn't define $F$. $\endgroup$ – user21820 May 16 '14 at 6:34

Since you have $d = \deg(m_θ) = [\mathbb{Q}(θ):\mathbb{Q}] = [K:\mathbb{Q}] = 8$, you already have that $d$ is independent of the choice of $θ$ as long as it is a primitive element for $K/\mathbb{Q}$. But I'm not so sure you understand the second equality, and how did you know that $a = \sqrt{2}+\sqrt{5}+\sqrt{7}$ is a primitive element?

From your comment I see that you don't actually know how to prove that $a$ is a primitive element. Jyrki gave one hint for proving it, which is to observe that any element in $Gal(K/F)$ must preserve the roots of $(t^2-2)$ and $(t^2-7)$ and $(t^2-5)$ as polynomials in $t$, and so it can only flip the signs of the radicals. Since the only element of $Gal(K/F)$ that fixes $a$ is the identity automorphism, $Gal(K/F(a))$ has size $1$ and so $[K:F(a)] = 1$, and hence $K = F(a)$. However, this method is curiously specific to this particular primitive element, and cannot be used to prove that some other choices of signs also gives a primitive element.

Here is a generic method that works for many other fields as well. Let $(r,s,t)=(\sqrt{2},\sqrt{5},\sqrt{7})$. Then we know that the subextensions of $K/F$ correspond to the subgroups of $Gal(K/F)$, which is itself a subgroup of ${S_2}^3$ because any element in $Gal(K/F)$ must preserve each of $\{r,-r\}$ and $\{s,-s\}$ and $\{t,-t\}$, and hence must be ${S_2}^3$ because $|Gal(K/F)| = [K:F] = 8$. Now consider the eight subextensions $F,F(r),F(s),F(t),F(r,s),F(r,t),F(s,t),F(r,s,t)$. They are all distinct because if any two are the same then $F(r,s,t)$ collapses to a smaller field. For example, if $F(r) = F(s)$ then $F(r,s,t) = F(r,t)$ which has too low extension degree. Therefore these are all the subextensions of $K/F$. If $F(r+s+t)$ is equal to any of them except $K$, then again $F(r,s,t)$ collapses. For example, if $F(r+s+t) = F(r,s)$ then $F(r,s,t) = F(r,s,r+s+t) = F(r,s)$. Therefore $F(r+s+t) = F(r,s,t)$ and likewise this shows that in fact any non-zero coefficients for a linear combination of $r,s,t$ would give a primitive element. (Note that this is not true in general!)

Nevertheless in this particular problem the crucial point is that $[K:F] = 8$, which is most easily proved by proving the independence of $r,s,t$, and so the above method is redundant! Do you really have a correct proof of $[K:F] = 8$?

  • $\begingroup$ $\sqrt{2}+\sqrt{5}+\sqrt{7}$ is a primitive element by the proof of the primitive element theorem. As the proof of the primitive element indicates, a primitive element for an extension can be obtained as a simple linear combination of the generators for the extension. $\endgroup$ – Leonhard Leibniz May 16 '14 at 6:35
  • 1
    $\begingroup$ @LeonhardLeibniz: Yes, but the theorem does not tell you which linear combination works, so your claim does not follow. $\endgroup$ – user21820 May 16 '14 at 6:38
  • 1
    $\begingroup$ If you know that $K/F$ is Galois over of degree 8, and that the Galois group consists of mappings determined by all sign combinations of $\pm\sqrt2$, $\pm\sqrt5$ and $\pm\sqrt7$, then the primitivity of $\alpha=\sqrt2+\sqrt5+\sqrt7$ follows from the fact that all its conjugates are strictly less than $\alpha$. $\endgroup$ – Jyrki Lahtonen May 16 '14 at 13:23
  • 1
    $\begingroup$ @JyrkiLahtonen: My point was that Leonhard Leibniz did not know that the primitive element theorem does not give what he claims at all. $\endgroup$ – user21820 May 16 '14 at 13:53
  • 2
    $\begingroup$ @JyrkiLahtonen: It's okay I was just making clear that it's an unproven claim that is not trivial. I've added an explanation of your comment, in case it is useful to Leonhard Leibniz. =) $\endgroup$ – user21820 May 16 '14 at 14:50

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.