4

This is part of a past qualifying exam problem:

Let $\alpha = \sqrt 2 + \sqrt 7$ and let $F=\mathbb Q(\alpha)$. Find the minimal polynomial of $\alpha$ over $\mathbb Q$ and determine $[F: \mathbb Q]$.

Work so far. I was able to calculate the following polynomial over $\mathbb Q$ that has $\alpha$ as a root: $f(x):= x^4 - 18x^2 + 25$. I'm pretty sure that this is the minimal polynomial, but I also need to verify it is irreducible over $\mathbb Q$. Eisenstein's criterion doesn't apply as far as I can tell, and since it's a quartic, then it's not enough to check that $f$ has no roots in $\mathbb Q$ (using, e.g., rational roots test).

That being said, using rational roots test, I can deduce that if $f$ does factor, it must factor as a pair of quadratics. My question is: besides comparing coefficients and setting up a system of equations to rule out a factorization into quadratics, is there any easier way to show that $f$ is irreducible over $\mathbb Q$?

RobPratt
  • 50,938

3 Answers3

4

To solve this problem, we can first decompose the polynomial $f(x)$ into a product of 4 first-degree polynomials (Gauss's theorem implies that such a decomposition is unique).

$$ x^4 -18x^2+25=x^4+10x^2-28x^2+25=(x^2+5)^2 - (2x\sqrt{7})^2 = \\\\ = (x^2+5-2x\sqrt{7})(x^2+5+2x\sqrt{7})=(x^2-2x\sqrt7+7-2)(x^2+2x\sqrt{7}+7-2)=\\\\= [(x-\sqrt7)^2 - (\sqrt2)^2]\cdot [(x+\sqrt7)^2 - (\sqrt2)^2] = \\\\=\boxed{\bf (x-\sqrt7-\sqrt2)(x-\sqrt7+\sqrt2)(x+\sqrt7-\sqrt2)(x+\sqrt7+\sqrt2)}.$$

Now, let's assume that there exist two polyminals $g(x)$ and $h(x)$, such that $g(x)\cdot h(x)=f(x)$. These polyminals must, like all 2-degre polynominals, consist of 2 polyminals of degre 1 each - and these polyminals must be divisors of $f(x)$ because divisibility is a transitive relation. There are only 4 divisors of $f(x)$ like that - they are shown above. So, we have to consider only $3 $ possibilites, and none of them matches our search: for all 3 cases, both $g(x)$ and $h(x)$ are not in $\mathbb{Q}$.

So, we can conclude, that the polynominal $f(x) := x^4 - 18x^2 + 25$ is irreductible over $\mathbb{Q}$.

Ivan Borisyuk
  • 389
  • 1
  • 8
  • 2
    A more general result proved in the same way (see the last paragraph on page 2): https://kconrad.math.uconn.edu/blurbs/ringtheory/reducibleallp.pdf – KCd Dec 20 '24 at 20:27
2

In order to prove that $f(x)=x^4-18x^2+25$ is irreducible over $\mathbb Q$ we can prove that $f(n)$ is prime for some values of $n$. The reason is that if $f(x)=g(x)h(x)$ then $f(n)=g(n)h(n)$ can be prime only when one of the factors $g(n),f(n)$ is equal to $1$ and the other is prime.

Since $f(x)=f(-x)$ each time that $f(n)$ is prime so is $f(-n)$.

We have $$f(2)=-31\\f(4)=-7\\f(6)=673\\f(8)=2969$$ so we have show eight primes which is a proof that $f(x)$, of four degree, is irreducible over $\mathbb Q$.

Note.-According to the Bunyakovsky's conjecture, $f(x)$ is prime for infinite values of $f(n)$

Ataulfo
  • 32,657
  • Can you elaborate on why finding eight prime values is sufficient? Is there a theorem related to this? Thanks. – R. H. Vellstra Dec 21 '24 at 00:06
  • @pyridoxal_trigeminus Suppose it factors as a linear times a cubic. A linear polynomial can only take values $\pm1$ twice, and a cubic six times. If you had nine prime outputs from $f$, then either the linear took value $\pm1$ three times or the cubic seven times. Now do a similar analysis if $f$ factors as two quadratics. – 2'5 9'2 Dec 21 '24 at 00:17
  • Duplicate of this and many others. The question is also a dupe, so will likely soon be closed if anyone can find the time. – Bill Dubuque Dec 21 '24 at 01:32
  • @pyridoxal_trigeminus: This method, depending on luck, can be easy and fast. But sometimes it is very difficult or practically impossible to apply. The reason is that there are irreducible polynomials that do not present primes for not very large values ​​of $n$ and that in general $f(n)$ can be so large that a special method is needed to know if it is prime or not. On the other hand, there are irreducible polynomials that are always multiples of some constant number for which $f(n)$ can never be prime ($x^3-x+2$ is irreducible but never prime because it is always even) (suite) – Ataulfo Dec 21 '24 at 15:18
  • (suite) Respect to the sufficient primes, I don't remember well how I managed to bound their number but I do remember that it was equal to the degree plus a small number (my answer with double the degree, i.e. $8$, is then correct).

    The problem to bound is to examine $f(n)=g(n)h(n)$ in order to know the maximal times that one of the two factors can be equal to $\pm1$ which is not difficult as a function of the degree of the polynomial considered.

    – Ataulfo Dec 21 '24 at 15:19
0

Using elementary methods only...

You can quickly check $f(\pm1)$, $f(\pm5)$, $f(\pm25)$ are all nonzero so by the rational root theorem, $f$ has no linear factor.

So if $f$ factors, it factors as two quadratics: $$x^4-18x^2+25=\left(x^2+ax+b\right)\left(x^2+cx+d\right)$$ With no cubic term, $c=-a$: $$x^4-18x^2+25=\left(x^2+ax+b\right)\left(x^2-ax+d\right)$$ With no linear term, $ad=ab$. Either:

  • $a=0$ $$x^4-18x^2+25=\left(x^2+b\right)\left(x^2+d\right)$$ $$x^4-18x^2+25=x^4+(b+d)x^2+bd$$ But there are no integers that multiply to $25$ and sum to $-18$.

Or

  • $a\neq0$, so $d=b$: $$x^4-18x^2+25=\left(x^2+ax+b\right)\left(x^2-ax+b\right)$$ Now $b$ must be $\pm5$ considering the linear term: $$x^4-18x^2+25=\left(x^2+ax\pm5\right)\left(x^2-ax\pm5\right)$$ Finally, balancing the quadratic terms, $$-18=\pm10-a^2$$ But neither $8$ nor $28$ is a perfect square.

So this factorization is not possible.

2'5 9'2
  • 56,991