Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm trying to express the splitting field of a cubic equation as a vector space over the rationals. Specifically I am looking for a set of six independent vectors that span the space. If the roots are a,b, and c then I know I need to be able to express all expressions in a,b and c up to second degree: namely,

a, a^2

b, b^2

c, c^2

ab, bc, ca

a^2b, b^2c, c^2a, ab^2, bc^2, ca^2

Obviously there are 15 of these and they are not all independent. Given a and b, I easily get c as a linear term in a, b, and (a+b+c). Likewise c^2 is not independent of a^2 and b^2. And I find that I can generate ab etc as linear combinations of c, c^2, and abc.

So what are my six vectors? I have allowed myself 1, a, b, a^2, and b^2 ....this gives me five and I think they are all linearly independent. I'm only allowed one more and I can't get it to work. I'm inclined to try (a^2b + b^2c + c^2a) because I'm pretty sure I need it, but having done so I can't see how I generate terms like a^2b on their own.

Any ideas? I'd be especially interesteds if there is a basis which is more symmetric in some sense than the arbitrary collection of terms which I'm cobbling together.

share|cite|improve this question
A cubic extension has degree $3$, not degree $6$. Or do you mean the splitting field of a cubic? In that case, it depends on the Galois group of the cubic. – Qiaochu Yuan Jul 3 '11 at 5:01
Have you tried a Gröbner basis algorithm (Buchberger?) on the obvious set of generators? Also, is it immediately clear that you don't need to worry about monomials like $a^2b^2$? – Jyrki Lahtonen Jul 3 '11 at 5:25
@qiaochu Although English is my first language, I am very weak on the terminology in this subject. Yes, I think I meant the splitting field of the cubic; and I know that some cubics have a different Galois group, but I meant the most general case of S3. – Marty Green Jul 3 '11 at 5:45
@Jyrki I don't know anything about the methods you refered to; but yes, I forgot about terms like a^2b^2 when I posted my question. In fact, I get them the same way I got terms like ab; since a^2b^2c^2 is an integer, I divide by c^2 which is the same as multiplying by a linear term in c. I think. – Marty Green Jul 3 '11 at 5:48
@Marty: Sorry about bringing up Gröbner here. Too heavy a tool for this job. – Jyrki Lahtonen Jul 3 '11 at 8:17
up vote 4 down vote accepted

The simplest basis I can think of consists of monomials $\mathcal{B}=\{1,a,a^2,b,ab,a^2b\}$. One way to see this goes as follows. Let's first write $$ P(x)=(x-a)(x-b)(x-c)=x^3-s_1x^2+s_2x-s_3, $$ with the elementary symmetric polynomials $s_1=a+b+c$, $s_2=ab+bc+ac$ and $s_3=abc$. If $K$ is your base field, I claim that all the monomials $a^ib^jc^k$ can be written as linear combinations of monomials from the set $\mathcal{B}$ with coefficients from the field $L=K(s_1,s_2,s_3)$ (in your case surely $s_1,s_2,s_3\in K$, so the combinations are really $K$-linear).

The first and most obvious thing is to replace everywhere $c$ with $s_1-a-b$. After that we only need to take care of monomials $a^ib^j$. The quantities $a$ and $b$ are zeros of $P(x)$, so we know how to replace $a^i, i\ge 3$ and $b^j, j\ge3$ with lower powers. So we are left with the 9 monomials $\mathcal{B}\cup\{b^2,ab^2,a^2b^2\}$. Substituting $c=s_1-a-b$ to the equation $ab+ac+bc-s_2=0$ gives us a relation $$ b^2=ab+a(s_1-a-b)+b(s_1-a)-s_2 $$ that allows us to write $b^2$ as an $L$-linear combination of $a^2,ab,a,b$ and $1$. When we do this substitution to a monomial like $ab^2$ or $a^2b^2$ we introduce higher powers of $a$. But we can reduce these to lower powers of $a$ as before. It may feel like $s_3$ wasn't really used, but it did make an appearance, when we reduced the higher powers of $a$ and $b$.

share|cite|improve this answer
Nicely done, and very well explained. Thanks. – Marty Green Jul 3 '11 at 12:23
In retrospect the basis is obvious given that this is the usual way of getting a basis for a tower of field extensions. $L(a)/L$ has a basis $\{1,a,a^2\}$ and $L(a,b)/L(a)$ a basis $\{1,b\}$. A basis for $L(a,b)/L$ is gotten by including all the products of elements of the two bases. – Jyrki Lahtonen Jul 3 '11 at 18:41

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.