Sarcia3's math and computer science ramblings

A place for my thoughts (the list).


2026-01-25-a-galois-theory-problem

Introduction

As the exam season approaches there is an intense test season preceding it1. One of those tests that I had to take was the one from the Algebra course, which had the following as the problem number three:

Prove that the splitting field L of $X^3+tX-1$ over $\mathbb Q(t)$2 is a Galois extension and that $\mathrm{Gal}(L/Q(t)) \simeq S_3$.

The setting

This is a fairly standard problem: proving that this indeed is an Galois extension is just noting that $0=\mathrm{char}\ \mathbb Q=\mathrm{char}\ \mathbb Q(t)$. Showing that the Galois group is isomorphic to $S_3$ is a bit more complicated. As the polynomial is of degree $3$, we can notice that it suffices to show that

\[|\mathrm{Gal}(L/\mathbb Q(t))|=3!.\]

Most of the solutions I saw comprised of two steps:

  1. Proving that $3\mid [L:\mathbb Q(t)]$ (useful because $[L\colon \mathbb Q(t)]=|\mathrm{Gal}(L/\mathbb Q(t))|$). For example by noticing that $X^3+tX-1$ is irreducible.
  2. Proving that $2\mid |\mathrm{Gal}(L/\mathbb Q(t))|$. For example by stating the group explicitly or noticing that the discriminant of the polynomial is not a square3.

In this blog post I will show an interesting way to prove the latter.

The fun part

Let us remind ourselves of the construction of $\mathbb Q(t)$:

\[\mathbb Q(t)=\left\{\frac{P(t)}{Q(t)} \mid P, Q \in \mathbb Q[t]\right\}.\]

We can notice from the construction that this object would be algebraically the same if in the place of the formal variable $t$ we put any transcendental number $\alpha$.

With that isomorphism we can assume $t$ to be some transcendental number $\alpha$ and thus $\mathbb Q(t) \subseteq \mathbb R$. For some $\alpha$s our original polynomial has only one real solution — this happens for example when the derivative is never zero. Let’s try to find such $\alpha$.

\[\begin{align*} (X^3+\alpha X-1)' &= 0 \iff\\ 3X^2+\alpha &= 0 \iff\\ X &= \pm i\sqrt{\alpha/3}. \end{align*}\]

So for any transcendental $\alpha > 0$ (like $e$) our polynomial’s derivative is never $0$ over the reals, so it cannot have more than one real solution.

This in turn means that $\mathbb R \subset L \subset \mathbb C$. And we know that $\overline{(\cdot)}= (z \mapsto \overline z)$ is an $\mathbb R$-automorphism of $\mathbb C$ with the order $2$, and in particular it is an $\mathbb R$-automorphism of $L$ of the order two.

This means that for $\alpha > 0$, we get $2\mid |\mathrm{Gal}(L/\mathbb Q(\alpha))|$ as desired.

Footnotes:

  1. math courses at polish universities are usually divided into two parts — the lecture and the practicals. The lecture is theory and the practicals are problem-solving. Before being allowed to attempt the oral theory exam one must usually pass the practicals first, and to pass them, you have to write a test. 

  2. by $\mathbb Q(t)$ we mean the field of rational functions over $\mathbb Q$. 

  3. a well known lemma states that if $L$ is the splitting field of the polynomial $P$ of degree $n$ over $F$, then $\mathrm{Gal}(L/F) \subseteq A_n \iff $ the discriminant of $P$ is a square in $F$. 


2025-01-18-Hello-World!

I have just set-up this website, using GitHub Pages.

As one can notice I don't really have much HTML/CSS experience, so this website's code can be a little rusty. I hope to grow it in the future and add some interesting blogposts ^^.