I had an example in the book given as follows:

Find the minimal polynomial of $\sqrt{2}+\sqrt{3}$ over $\mathbb Q$ .

Solution: $~~~~~$$(\sqrt{2}+\sqrt{3})^2=5+2 \sqrt6$

$(\sqrt{2}+\sqrt{3})^4=49+20 \sqrt6$

Then $(\sqrt{2}+\sqrt{3})^4-10(\sqrt{2}+\sqrt{3})^2+1=0.$

Thus $a=\sqrt{2}+\sqrt{3}$ satisfies $f(x)=x^4-10x^2+1$ over $\mathbb Q$.

Let $p(x)$ be the minimal polynomial of $\sqrt{2}+\sqrt{3}$ over $\mathbb Q$.Then $(\sqrt{2}-\sqrt{3}),(-\sqrt{2}+\sqrt{3}),(-\sqrt{2}-\sqrt{3})$ are also roots of $p(x).$ So degree of $p(x)$ is atleast $4$.

But $f(a)=0 $ and $f(x) \in \mathbb Q[x] \implies p(x)$ divides $f(x)$ .

So $f(x)$ is minimal polynomial of $\sqrt{2}+\sqrt{3}.$

But I can't get the step how are $(\sqrt{2}-\sqrt{3}),(-\sqrt{2}+\sqrt{3}),(-\sqrt{2}-\sqrt{3})$ also roots of $p(x).$

Kindly help with this.


That fact comes from Galois Theory. Under the assumption that $K/F$ is Galois, the (monic) minimal polynomial $f$ of an element $\alpha \in K$ (over $F$) is of the form $$f(x) = \prod_{\sigma \in \text{Gal}(K/F)} (x - \sigma \alpha).$$ This implies that $\sigma \alpha$ are all roots of the minimal polynomial for $\alpha$.

In your case, let $K = \mathbb{Q}(\sqrt{2}, \sqrt{3})$ and $F = \mathbb{Q}$. Then $K/F$ is Galois and there are 4 automorphisms in $\text{Gal}(K/F)$: namely

  • identity on $K$
  • the one that sends $\sqrt{2} \mapsto -\sqrt{2}$ keeping $\sqrt{3}$ fixed: this automorphism sends $\sqrt{2} + \sqrt{3}$ to $-\sqrt{2} + \sqrt{3}$; hence $-\sqrt{2} + \sqrt{3}$ must be a root of $p(x)$.
  • the one that keeps $\sqrt{2}$ fixed and sends $\sqrt{3} \mapsto -\sqrt{3}$: like above, we deduce $\sqrt{2} - \sqrt{3}$ is root of minimal polynomial $p(x)$
  • the one that negate both roots: finally $- \sqrt{2} - \sqrt{3}$ must also be a root.
| cite | improve this answer | |
  • $\begingroup$ How do we know $K/F$ Is Galois? How do we know these maps are automorphisms? Somewhere we need to use some facts about $\sqrt{2}$ and $\sqrt{3}$. $\endgroup$ – 6005 Sep 3 '17 at 0:53

Good question: that is a big leap in the argument in my opinion.

If we let $K = \mathbb{Q}(\sqrt{2},\sqrt{3})$, then $\sqrt{2} + \sqrt{3}$, $\sqrt{2} - \sqrt{3}$, $-\sqrt{2} + \sqrt{3}$, and $-\sqrt{2} - \sqrt{3}$ all lie in $K$. In order to prove that they are all roots of $p(x)$, the minimal polynomial of $\sqrt{2} + \sqrt{3}$, we follow An Hoa's idea: find automorphisms $\sigma, \tau$ of $K / F$ such that $\sigma(\sqrt{2}) = -\sqrt{2}$, $\sigma(\sqrt{3}) = \sqrt{3}$, $\tau(\sqrt{2}) = \sqrt{2}$, and $\tau(\sqrt{3}) = -\sqrt{3}$. The argument is then completed by noting that $\sigma, \tau$ both preserve $p(x)$ (since its coefficients are in $K$), so they preserve the roots of $p(x)$. (If $\alpha$ is a root of $p(x)$, then $\sigma(\alpha)$ is a root of $\sigma(p(x)) = p(x)$ and $\tau(\alpha)$ is a root of $\tau(p(x)) = p(x)$.)

Therefore what we need to show is that these automorphisms $\sigma$ and $\tau$ exist. $K = \mathbb{Q}(\sqrt{2},\sqrt{3})/\mathbb{Q}$ is the splitting field of $(x^2 - 2)(x^2 - 3)$, so it is a Galois extension. To argue the extension has degree $4$, further notice that it contains both $\mathbb{Q}(\sqrt{2})$ and $\mathbb{Q}(\sqrt{3})$. These two fields are distinct because one contains a square root of two and one doesn't (all elements of $\mathbb{Q}(\sqrt{3})$ can be written $a + b\sqrt{3}$ and you can show that such a thing squared cannot be $2$). So $K$ has degree divisible by $2$ and strictly greater than $2$ over $\mathbb{Q}$, and it also has degree at most $4$ since it is the splitting field of a degree-4 polynomial, so it has degree exactly $4$.

Now that we know $K$ has degree $4$ over $\mathbb{Q}$ and is Galois, it must have four automorphisms (this is the definition of Galois, that the number of automorphisms equals the degree). These automorphisms permute the roots of $(x^2 - 2)(x^2 - 3)$ and are defined by where they send these roots. They must send roots of $(x^2 - 2)$ to roots of $(x^2 - 2)$ and roots of $(x^2 - 3)$ to roots of $(x^2 - 3)$. But there are only $4$ different ways to do this. Therefore, every possible way of sending $\sqrt{2} \mapsto \pm \sqrt{2}$ and $\sqrt{3} \mapsto \pm \sqrt{3}$ is an automorphism. In particular, $\sigma$ and $\tau$ are automorphisms. $\square$

| cite | improve this answer | |
  • $\begingroup$ But couldn’t OP have done it by noticing that the original computation, but replacing $\sqrt2$ with $-\sqrt2$, would give the same conclusion? similarly for replacing $\sqrt3$ with $-\sqrt3$; and for making both replacements at once? $\endgroup$ – Lubin Sep 4 '17 at 22:11
  • $\begingroup$ @Lubin No. It is true that all four of those values are roots of $(x^2 - 5)^2 - 24 = 0$. But what you have to show is that this is actually the minimal polynomial of $\sqrt{2} + \sqrt{3}$. The minimal polynomial of $\alpha$ is the smallest degree monic polynomial $p$ such that $p(\alpha) = 0$. So just knowing that $\alpha$ satisfies that particular polynomial doesn't show it's minimal. $\endgroup$ – 6005 Sep 4 '17 at 22:46
  • $\begingroup$ You are quite right: I read too carelessly, and thought OP was referring to his $f$, not $p$. My apologies. $\endgroup$ – Lubin Sep 5 '17 at 0:30

Assume that $m$ and $n$ are square-free coprime integers. Then, if we prove that $Q(\sqrt m +\sqrt n) = Q(\sqrt m, \sqrt n)$, we can conclude that that $f(x) = x^4 -2(m+n)x +(m-n)^2$ must be the minimal polynomial of $\sqrt m +\sqrt n$ since $Q(\sqrt m +\sqrt n)$ is a four-dimensional $Q$-vector space. But if $\xi = (\sqrt m + \sqrt n)$, then ${\frac {{\xi}^2-(m+n)} {2}}={\sqrt m}{\sqrt n}$. Thus, $({\frac {{\xi}^2-(m+n)} {2}})\xi = m{\sqrt n}+n{\sqrt m}$. Hence, $$({\frac {{\xi}^2-(m+n)} {2}})\xi -n{\xi} = (m-n){\sqrt n},$$ and $$({\frac {{\xi}^2-(m+n)} {2}})\xi -m{\xi} = (n-m){\sqrt n}.$$ Thus, $Q(\sqrt m +\sqrt n)$ contains $\{\sqrt m, \sqrt n \}$, and so $Q(\sqrt m +\sqrt n)\supset Q(\sqrt m ,\sqrt n)$. The reverse set inclusion is immediate. The case $m =2$, and $n = 3$ is of course what inspired this train of thought which is hopefully correct.

| cite | improve this answer | |
  • 1
    $\begingroup$ This is a proof that $x^4 - 10x^2 + 1$ has those roots, not that the minimal polynomial has those roots. You have to prove that the minimal polynomial has those other roots. $\endgroup$ – 6005 Sep 3 '17 at 0:52
  • 1
    $\begingroup$ $x^2-3$ is the minimal polynomial of $\sqrt{3}$ over $\mathbb{Q}(\sqrt{2})$ thus $(x-\sqrt{2})^2-3$ is the minimal polynomial of $\sqrt{3}+\sqrt{2}$ over $\mathbb{Q}(\sqrt{2})$. Can you finish ? $\endgroup$ – reuns Sep 3 '17 at 1:53
  • $\begingroup$ This is definitely taking the "low road" but $f(x) = x^4 -10x^2 +1$ must be a multiple of the minimal polynomial of $\sqrt 2 +\sqrt 3$, and as $f(x)$ factors into four distinct linear factors $(x-(\sqrt 2 +\sqrt 3))$, $(x+(\sqrt 2 +\sqrt 3))$, $(x-(\sqrt 2 -\sqrt 3))$, and $(x+(\sqrt 2 -\sqrt 3))$, and since all 6 combinations into products of 2 factors do not yield a factorization of $f(x)$ as a product of two quadratics in $Q[x]$, we conclude that $f(x)$ must be irreducible in $Q[x]$, and must therefore be the minimal polynomial of $\sqrt 2 +\sqrt 3$. $\endgroup$ – student Sep 4 '17 at 2:26
  • $\begingroup$ "reuns" Does the above edit make sense? $\endgroup$ – student Sep 4 '17 at 18:25

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.