why is there no quintic formula?

explained by a sixth form student

2026-06-07

This is part 2 of a series on group theory. You'll probably want to read part 1 if you want to understand the conclusion on group solubility.

I want to start this post with an admission that there are better teachers than me. Here are a couple of resources I strongly recommend:

However, I'm going to try and close up what that article didn't do for me. I also want to keep the same level of approachability and understanding as my past writings on group theory.1 So if you're willing to stick it through, here's my take on elementary Galois theory!

introduction

You know the quadratic formula. You've heard horror stories of the cubic formula, and the quartic formula keeps you up at night. The quintic formula??

Oh. Well, it doesn't exist. In fact, it doesn't exist for any polynomial of order greater than 4. The reason why lies in an extremely surprising field of mathematics - group theory! (Yeah, who could've guessed I'm writing about group theory.) I think the question of why the quintic formula doesn't exist is one of those freaky maths things the nerd in your class cites randomly, and so there's some fun in trying to pick apart what's going on.

I mean - what would it mean for the quintic formula to exist? The fact that it doesn't seems to imply that if you give me a quintic, I can't always express it as a number made of rationals, pluses, minuses and roots. Sure, some may have a solution that you can express like that (called being solvable by radicals), but the fact that not all of them do means you can never have a single expression that solves every quintic.

But then, how can we tell whether a polynomial is solvable by radicals? And how did we prove that this isn't always true for quintics and above?

a few definitions to get out the way

Hopefully you remember what a field is - essentially a set that behaves like the rational numbers, with properly defined addition, subtraction, multiplication and non-zero division. Actually, throughout this post, we're only gonna be concerned with the field of rational numbers itself, and its extensions.

What's an extension? Well, an extension $K$ of a field $F$ is a bigger field that we made from $F$ by adding some new elements. This bigger field has the same operations as $F$.

For example, the field of real numbers is an extension of the field of rational numbers, because:

Essentially, the original field lies untouched within the extension, but there are some new members to play around with.

Now, we're talking about quadratics, cubics, quartics, quintics, which means we're concerned with polynomials (in a single variable, with rational coefficients).

Throughout this entire post, I refer to polynomials a lot. I want to make it clear that when I say polynomial, I always mean a polynomial with rational coefficients.

There's something curious about the solutions to these polynomial equations. In fact, you probably know it already; which of these complex numbers do you think would be a root of a polynomial?

Of these, $4 \pi + 2$ and $e^2$ can never be solutions to a polynomial.2 This is because $\pi$ and $e$ are transcendental numbers. Roots of polynomials can be complex, and they can be radicals, but they can't be any complex number. This isn't anything crazy, but that means we don't actually need all of $\mathbb C$ to express the roots of any polynomial. All roots of polynomials fall in a set known as the algebraic numbers $\mathbb A$. In fact, that's how we define an algebraic number! But then what do we need to solve a given polynomial?

Let's look back at field extensions. We can extend a field with what I'm going to call a root extension (not a real term) by "dropping in" a root of a number that is already in $\mathbb Q$. The proper term for this is adjoining a radical. For example, $\mathbb Q [\sqrt 2]$ is the field of rational numbers with $\sqrt 2$ added;3 however, since fields are by definition closed, we have to make sure that all possible results of operations involving $\sqrt 2$ are also in $\mathbb Q [\sqrt 2]$. This means that all these numbers are part of $\mathbb Q [\sqrt 2]$:

We can keep compounding these root extensions; now that $1 + \sqrt 2$ is in our field $\mathbb Q [\sqrt 2]$, we can extend it even further to $Q[\sqrt{1 + \sqrt 2}]$,4 with elements like:

(Makes sense, I hope 🐶)

This gives us a radical extension of $\mathbb Q$; i.e. a field extension we can get to by just doing a bunch of root extensions. We call the extensions that we used to get up to a radical extension the tower of fields.

For a given polynomial, we can actually try to create its own bespoke extension of $\mathbb Q$ - the smallest one that contains its roots! This is called the splitting field of the polynomial.5

For example:

And so, we can create these cool companion fields for every polynomial that contains its roots, and these companion fields are extensions of the rational numbers. What now?

"how can i make this about group theory" - Évariste Galois

The thing is, field theory is new and difficult. We've got so many neat little tools in group theory... Well, stick with me in the realm of field extensions for just one more moment.

An automorphism is a relabelling of elements of a field that preserves the structure of the field. Formally, an automorphism on a field $F$ is a bijective function $f: F \rightarrow F$ (meaning it maps each element in $F$ to a unique element in $F$) such that (for $x, y \in F$):

So basically, elements exhibit the same behaviours as they did before they were transformed, which is why I like to call it a relabelling.

NRICH gives a good example of an automorphism on $\mathbb Q [\sqrt 2]$: a function $f$ which maps all numbers that look like $a + b \sqrt 2$ to $a - b \sqrt 2$, which I'm going to write as

$$f: \sqrt 2 \mapsto - \sqrt 2$$

You can check it yourself; operations look exactly the same if you "step into" the world of $fF$ and then step back (like this)

$$ \frac{1}{1 + \sqrt 2} = \sqrt 2 - 1 \\\\ f\left(\frac{1}{1 + \sqrt 2}\right) = \frac{1}{1 - \sqrt 2} = -\sqrt 2 - 1 = f\left(\sqrt 2 -1\right) $$

And that's the important quality of an automorphism.

In addition, if you have a field extension $K / F$, $K$ can have what's called an F-automorphism. This is an automorphism which acts on $K$ but leaves the elements of $F$ untouched; i.e. $f(x) = x \ \forall \ x \in F$.
The automorphism we just looked at is actually a $\mathbb Q$-automorphism on $\mathbb Q [\sqrt 2]$, since $f(a)$ where $a \in \mathbb Q$ is just $a$.

Well, the splitting field of a polynomial also has its own $\mathbb Q$-automorphisms. Something really cool about these automorphisms, though, is that they always rearrange the roots of the polynomial. For example: $x^4 - 2$ has 4 roots: $\pm \alpha, \pm i \alpha$, where $\alpha = \sqrt[4] 2$ - so it has splitting field $\mathbb Q [\alpha, i]$.

That splitting field has 8 automorphisms. Since they act symmetrically on the 4 roots of our polynomial, I can just show you what they do to every single $\alpha$ and $i$ they see, similar to the automorphism on $\mathbb Q [\sqrt 2]$.

Listing them out, we have:

Label $\alpha$ goes to... $i$ goes to... Label $\alpha$ goes to... $i$ goes to...
$\text{id}$ $\alpha$ $i$ $\tau$ $\alpha$ $-i$
$\sigma$ $i \alpha$ $i$ $\sigma \tau$ $i \alpha$ $-i$
$\sigma^2$ $- \alpha$ $i$ $\sigma^2 \tau$ $- \alpha$ $-i$
$\sigma^3$ $-i \alpha$ $i$ $\sigma^3 \tau$ $-i \alpha$ $-i$

It's a lot to take in at once! Although, they do have a little bit of structure to them; for example, the automorphisms labelled with $\tau$ are like dual counterparts, and they seem to follow a sort of rotation pattern.

Hang on. Here's the set of $\mathbb Q$-automorphisms of $\mathbb Q [\alpha, i]$:

$$ \{ \text{id}, \sigma, \sigma^2, \sigma^3,\ \tau, \sigma \tau, \sigma^2 \tau, \sigma^3 \tau \} $$

\blink\

$$ \{ e, r, r^2, r^3, \ f, rf, r^2f, r^3f \} $$

It was $D_4$ this whole time - our old friend, the dihedral group of order 8! If we start to compose automorphisms by applying them one after the other, we can see that with $\sigma$ representing the rotation of roots and $\tau$ representing reflection in the real axis, the set of automorphisms becomes a group! This special little group is called the Galois group of our polynomial $z^4 - 2$, or equivalently the Galois group of our splitting field $\mathbb Q [\alpha, i]$ over $\mathbb Q$, written as

$$ \text{Gal}(\mathbb Q [\alpha, i] / \mathbb Q) \text{ or Gal}(z^4-2) $$

bonus section (extra credit)

Have you ever seen this image?

Infamous "Proof. Ask a toddler on the street." math meme.

A lot of people use it as an example for how strange and obscure mathematics can look as you start to use more and more technical notation.. But what if I said you actually understand the joke right now?

This is ripped from a textbook in the middle of a section where it was talking about the roots of unity; so $K$ is actually the splitting field of the polynomial $z^m - 1$, which means $\text{Gal}(K/\mathbb Q) = \text{Gal}(z^m-1)$.

So we're thinking about the different ways of "rearranging the roots" of this polynomial: how ever could we permute $m$ roots of unity...?

Although it's not as simple as a geometric rotation, it's not really surprising that these transformations form a cycle, is it? Which actually means that the group of automorphisms form the cyclical group $C_m$. You know what else forms the cyclical group $C_m$? The integers modulo $m$ under multiplication!

So what that image really says is that the Galois group of the $m$th roots of unity is isomorphic to ($\cong$) the group of integers modulo $m$ under multiplication. The proof isn't that easy, but it's not exactly surprising that the Galois group we get is cyclical. $\text{Gal}(K/\mathbb Q) \cong (\mathbb Z/m \mathbb Z)^\times$!!

One thing to notice is that the Galois groups of root extensions are actually quite special - no matter what the extension is, if and only if it adjoins a radical, its Galois group is cyclical. That's a surprise tool that will help us later ;46

Although this is a pretty cool connection already, there's something very exciting about these groups that we're about to see.

solubility (just add water!)

Remember the context that we're in: we have a splitting field of a polynomial, which is the minimal extension of $\mathbb Q$ that contains its roots. For some polynomials, like the examples I gave, it's easy to see how we get to its splitting field using only root extensions. For others that are less obvious, it's a little bit more of an effort.

Our goal is to figure out whether we can even get to a polynomial's splitting field using only root extensions - because remember, if the splitting field is a radical extension, that means the polynomial itself is solvable by radicals!

As a reminder, what we're looking for is a tower of fields

$$ \mathbb Q = F_0 \subseteq F_1 \subseteq F_2 \subseteq \cdots \subseteq F_k $$

where $F_k$ is our splitting field.

What if I told you that the tower of fields we use to get to a splitting field directly corresponds to the Galois group of the splitting field? What's important to us specifically is how we can write that group $G$ as a subnormal series.

$$ 1 = H_0 \triangleleft H_1 \triangleleft H_2 \triangleleft \cdots \triangleleft H_n = G $$

I talk about subnormal series in part 1 :]

When we take this subnormal series of $G$, a specific quotient group in the chain, say

$$H_n / H_{n-1}$$

..Is actually the exact same group as the Galois group on the other end of the field tower:

$$\text{Gal}(F_1/F_0)$$

So what we want to find is a subnormal series for $G$ where every single quotient group is cyclical; we call a group like this soluble.7 Why? Because that means every single field extension is a root extension! And so the polynomial is solvable by radicals!! (\explosion sound effect\, source: your mind)

Don't understand what I mean? Let's get more hands-on with our favourite polynomial $z^4 - 2$.

Its Galois group is $D_4$. I'm going to go through one of its subnormal series, and show you at each step how this relates to the field extensions that produce the splitting field of the polynomial.

The first normal subgroup we'll take is the one generated by the rotation element $r$.

$$ C_4 \triangleleft D_4 \\\\ \{e, r, r^2, r^3\} \triangleleft \{e, r, r^2, r^3, r^4, f, rf, r^2f, r^3f\} $$

The quotient group at this stage is $D_4 / C_4 = C_2$. This corresponds to extending the field $\mathbb Q$ to $\mathbb Q [i]$:

$$\text{Gal}(\mathbb Q [i]/\mathbb Q) = C_2$$

The group of $\mathbb Q$-automorphisms on $\mathbb Q [i]$ is, as we expect, $C_2$. Of course, it's cyclical, because we got it by adjoining $\sqrt{-1}$.

The second normal subgroup we'll take is generated by $r^2$:

$$ C_2 \triangleleft C_4 \\\\ \{e, r^2\} \triangleleft \{e, r, r^2, r^3\} \\\\ C_4 / C_2 = C_2 $$

And the corresponding field extension is adjoining $\sqrt 2$:

$$\text{Gal}(\mathbb Q [\sqrt 2, i] / \mathbb Q [i]) = C_2$$

Again, its $C_2$ - cyclic, so achieveable by radicals.

To no surprise, the third and last normal subgroup is the trivial group!

$$ 1 \triangleleft C_2 \\\\ \{e\} \triangleleft \{e, r^2\} \\\\ C_2 / 1 = C_2 \\\\ \text{Gal}(\mathbb Q [\sqrt[4] 2, i] / \mathbb Q [\sqrt 2, i]) $$

Which means we're done! We've shown that there's a subnormal series for $\text{Gal}(\mathbb Q [\alpha, i]/\mathbb Q)$ where each quotient group is cyclical, and you can see how this corresponds to a field tower where each extension is a root extension!

There's a really sick YouTube video that goes through the exact same decomposition I just did, and it explains how we can "extract" the field extensions from the quotient group/Galois group that I implicitly did in my explanation.

The foundational statement is that a polynomial is solvable by radicals if and only if its Galois group is soluble.

Also, if a group is soluble, all of its subgroups are soluble.8

the burden of floof: conclusion

That's cool and all, but what does it say about quintics?

A polynomial of degree $n$ has $n$ solutions (deepest statement in this blog post). That means that, no matter what its roots look like, its Galois group (actions that rearrange its roots) will always $\subseteq S_n$, which is the group of all permutations of $n$ elements.

$1, C_2, S_3, S_4$ are all soluble. This means that all polynomials of degree 4 or lower are solvable by radicals - which is why we have our handy quartic formula. All it takes is to show that $S_5$ is not soluble, because as we now know, if not all quintic polynomials are solvable by radicals, there's no way we can make a quintic formula.

The only normal subgroup of $S_5$ is $A_5$, and the only normal subgroup of $A_5$ is the trivial group, which means the only subnormal series of $S_5$ contains the quotient group $A_5/1 = A_5$. $A_5$ is simple, and not cyclical - so $S_5$ is not soluble.

$\text{Gal}(x^5 - 2x + 1)$ is $S_5$.9 Soooooo not all quintics are solvable by radicals! Which is why there's no quintic formula :)

closing

This is by far a formal, rigorous proof of the fact, but I hope I went into enough detail to show how and why this little branch of Galois theory works. The NRICH article has some more information on other awesome results you can get by using groups, and there's even more out there - trust me when I say group theory is everywhere in mathematics.

This is definitely my longest post so far, and possibly my most involved, considering the foreknowledge you need of normal subgroups and subnormal series. I hope you enjoyed it, and see you next time!


  1. If you know what I'm talking about and anything seems incorrect, misunderstood or not well researched, please don't hesitate to reach out to me on Tumblr! The last thing I want in a supposedly easy-to-digest primer on any topic is to be blatantly wrong, especially considering this post is a reflection of my own understanding. <

  2. In case you're wondering, here's a polynomial with rational coefficients which has all of those as roots: $$ x^8 - 7x^7 + 15x^6 - 59x^5 + 124x^4 - 24x^3 + 490x^2 - 540x - 1260 $$ This made me wonder if there's a way to get the lowest degree polynomial over $\mathbb Q$ for any given set of roots. <

  3. The field extension $\mathbb Q [\sqrt 2]$ means the field with the least extensions required to have it include $\sqrt 2$. You'll find me writing $\mathbb Q [\alpha]$ for fields that should contain $\alpha$ more generally. <

  4. This field extension can be written as $\mathbb Q [\sqrt 2 , \sqrt{1 + \sqrt 2}]$ to show that it needs to include both $\sqrt 2$ and $\sqrt{1 + \sqrt 2}$; but since $\mathbb Q [\sqrt{1 + \sqrt 2}]$ as the minimal field required to contain $\sqrt{1 + \sqrt 2}$ already contains $\sqrt 2$, we don't have to include it. <

  5. It's called a splitting field because that field $F$ lets us split the polynomial into factors $(x - \alpha_1)(x - \alpha_2) \cdots$ where $\alpha_n \in F$. Therefore, you can denote the splitting field of a polynomial as $\mathbb Q [\alpha_1, \alpha_2, \cdots]$ to show that $F$ needs to contain all of the polynomial's roots. <

  6. This is Kummer Theory, although there are a few conditions on the theory that I'm pretty sure are fulfilled in our context. Pretty sure. <

  7. Or solvable, but that's boring. I think soluble sounds cooler and it's the same root word anyways. <

  8. I should probably have proved this. Stack Exchange, anyone? <

  9. If you want to see how to get to this, here's a cool read. <