Algebraic Number Theory

2.2.15 Formulae for calculating discriminants

Let us now look at some alternative ways for calculating discriminants.

Proposition 2.2.16

Let \(K\) be a number field with basis \(B=\{ b_1,\dots ,b_n\} \) and let \(\sigma _1,\dots ,\sigma _n\) be the embeddings of \(K\) into \(\mathbb {C}\). Now let \(M\) be the matrix

\[ \left(\begin{matrix} \sigma _1(b_1) & \cdots & \sigma _1(b_n) \\ \vdots & & \vdots \\ \sigma _n(b_1) & \cdots & \sigma _n(b_n) \end{matrix} \right). \]

Then

\[ \Delta (B)=\det (M)^2. \]

Proof

By Proposition 1.7.6 we know that \(\operatorname {Tr}_{K/\mathbb {Q}}(b_i b_j)= \sum _k \sigma _k(b_i)\sigma _k(b_j)\) which is the same as the \((i,j)\) entry of \(M^t M\). Therefore

\[ \det (T_B)=\det (M^t M)=\det (M)^2. \]

Now, by the Primitive element theorem 1.5.11, for any number field \(K\) we can find some \(\alpha \) such that \(K=\mathbb {Q}(\alpha )\) and thus \(\{ 1,\alpha ,\alpha ^2,\dots ,\alpha ^{n-1}\} \) is a basis for \(K/\mathbb {Q}\) where \(n=[K:\mathbb {Q}]\). In this case the discriminant is given by:

Proposition 2.2.17

Let \(K\) be a number field and \(B=\{ 1,\alpha ,\alpha ^2,\dots ,\alpha ^{n-1}\} \) for some \(\alpha \in K\). Then

\[ \Delta (B)=\prod _{i {\lt} j} (\sigma _i(\alpha )-\sigma _j(\alpha ))^2 \]

where \(\sigma _i\) are the embeddings of \(K \) into \(\mathbb {C}\).

Proof

First we recall a classical linear algebra result relating to the Vandermonde matrix, which states that

\[ \det \left(\begin{matrix} 1 & x_1 & x_1^2 & \cdots & x_1^{n-1} \\ \vdots & & & \vdots \\ 1 & x_n & x_n^2 & \cdots & x_n^{n-1} \end{matrix} \right) =\prod _{i{\lt}j} (x_i-x_j). \]

Combining this with Proposition 2.2.16 gives the result.

Lastly, we have probably the more useful formula for computing the discriminant in this case, but before we state it we need the following lemma.

Lemma 2.2.18

Let \(f\) be a monic irreducible polynomial over a number field \(K\) and let \(\alpha \) be one of its roots in \(\mathbb {C}\). Then

\[ f'(\alpha )=\prod _{\beta \neq \alpha } (\alpha -\beta ), \]

where the product is over the roots of \(f\) different from \(\alpha \).

Proof

We first write \(f(x)=(x-\alpha )g(x)\) which we can do (over \(\mathbb {C}\)) as \(\alpha \) is a root of \(f\), where now \(g(x)=\prod _{\beta \neq \alpha } (x-\beta )\). Differentiating we get

\[ f'(x)=g(x)+(x-\alpha )g'(x). \]

If we now evaluate at \(\alpha \) we get the result.

Proposition 2.2.19

Let \(K=\mathbb {Q}(\alpha )\) be a number field with \(n=[K:\mathbb {Q}]\) and let \(B=\{ 1,\alpha ,\alpha ^2,\dots ,\alpha ^{n-1}\} \). Then

\[ \Delta (B)=(-1)^{\frac{n(n-1)}{2}}N_{K/\mathbb {Q}}(m_\alpha '(\alpha )) \]

where \(m_\alpha '\) is the derivative of \(m_\alpha (x)\) (which we recall denotes the minimal polynomial of \(\alpha \)).

Proof

By Proposition 2.2.17 we have \(\Delta (B)=\prod _{i {\lt} j}(\alpha _i-\alpha _j)^2\) where \(\alpha _k:=\sigma _k(\alpha )\). Next, we note that the number of terms in this product is \(1+2+\cdots +(n-1)=\frac{n(n-1)}{2}\). So if we write each term as \((\alpha _i-\alpha _j)^2=-(\alpha _i-\alpha _j)(\alpha _j-\alpha _i)\) we get

\[ \Delta (B)=(-1)^{\frac{n(n-1)}{2}}\prod _{i=1}^n \prod _{i \neq j} (\alpha _i-\alpha _j). \]

Now, by Lemma 2.2.18 and Proposition 1.7.6 we see that

\[ N_{K/\mathbb {Q}}(m_\alpha '(\alpha ))=\prod _{i=1}^n m_\alpha '(\alpha _i)=\prod _{i=1}^n \prod _{i \neq j} (\alpha _i-\alpha _j), \]

which gives the result.

Example 2.2.20
#

Let \(\alpha \) be a root of \(m_\alpha (x)=x^8-2\) (which is irreducible by Eisensteins criterion 1.2.15). Therefore \(\mathbb {Q}(\alpha )\) has degree \(8\) over \(\mathbb {Q}\) and a basis is \(B:=\{ 1,\alpha ,\dots ,\alpha ^7\} \). From the above we then have

\begin{align*} \Delta (B)= (-1)^{8\cdot 7/2} N_{\mathbb {Q}(\alpha )/\mathbb {Q}} (m_\alpha ’(\alpha ))& = N_{\mathbb {Q}(\alpha )/\mathbb {Q}} (8\alpha ^7)\\ & =N_{\mathbb {Q}(\alpha )/\mathbb {Q}}(8) N_{\mathbb {Q}(\alpha )/\mathbb {Q}}(\alpha ^7)=8^8 \cdot (-2)^7=-2^{31}\end{align*}

Here we use that \(N_{K/\mathbb {Q}}(\alpha ^7)=N_{K/\mathbb {Q}}(\alpha )^7\) and that \(N_{K/\mathbb {Q}}(\alpha )=-2\).

Now, by Lemma 2.2.7, if we want to find an integral basis, we need to first pick a basis \(B\) of algebraic integers, and then check for which primes \(p|\Delta (B)\) we have \(p^2| \Delta (B)\). Its at these primes where we might need to modify our basis candidate \(B\).

In order to make this simpler, here is a simple trick.

Lemma 2.2.21

Let \(K=\mathbb {Q}(\alpha )\)and \(\alpha \) be an algebraic integer such that \(m_\alpha \) satisfies Eisensteins Criterion 1.2.15 for a prime \(p\). Then none of the elements

\[ \phi =\frac{1}{p}(x_0+x_1\alpha +\cdots +x_{n-1}\alpha ^{n-1}) \]

is an algebraic integer, where \(n=\deg (m_\alpha )\) and \(x_i \in \{ 0,\dots ,p-1\} .\)

Proof

We will only prove the case when \(m_\alpha \) is Eisenstein, since the proof of the more general case is identical.

Suppose for contradiction that \(\phi \in \mathcal{O}_K\) and let \(x_d\) be the first non-zero coefficient, so

\[ \phi =\frac{1}{p}(x_d\alpha ^d+x_{d+1}\alpha ^{d+1}+\cdots +x_{n-1}\alpha ^{n-1}) \in \mathcal{O}_K. \]

Now, rewrite this as \(\phi =\frac{1}{p}(x_d\alpha ^d +\alpha ^{d+1}\beta )\) for some \(\beta \in \mathcal{O}_K\). Next, multiply through by \(\alpha ^{n-1-d}\), then we have

\[ \frac{x_d\alpha ^{n-1}}{p}+\frac{\alpha ^n\beta }{p} \in \mathcal{O}_K. \]

Now, since \(m_\alpha \) satisfies Eisenstein at \(p\), we see that \(\alpha ^n=pf(\alpha )\) for some \(f \in \mathbb {Z}[x]\) and therefore the above gives us that

\[ \frac{x_d\alpha ^{n-1}}{p}+\beta f(\alpha ) \in \mathcal{O}_K. \]

and thus

\[ \frac{x_d\alpha ^{n-1}}{p} \in \mathcal{O}_K. \]

Lets now calculate the norm of this:

\[ N_{K/\mathbb {Q}} \left(\frac{x_d\alpha ^{n-1}}{p} \right)=\frac{x_d^n N_{K/\mathbb {Q}}(\alpha )^{n-1}}{p^n}. \]

By Eisenstein the constant coefficient of \(m_\alpha \) is divisible by \(p\) but not \(p^2\), so since the constant coefficient of \(m_\alpha \) is \(N_{K/\mathbb {Q}}(\alpha )\) we see that \(N_{K/\mathbb {Q}}(\alpha )=p a\) where \(p \nmid a\). Therefore we have

\[ N_{K/\mathbb {Q}} \left(\frac{x_d\alpha ^{n-1}}{p} \right)=\frac{x_d^n p^{n-1}a^{n-1}}{p^n}= \frac{x_d^n a^{n-1}}{p}. \]

But this cant be in \(\mathbb {Z}\) since \(p\) doesn’t divide \(x_d\) or \(a\), and this gives us a contradiction since Proposition 2.1.19 says that the norm of an algebraic integer must be an integer. So \(\phi \) couldn’t have been an algebraic integer.

How do we use this? well let me show you one use.

Example 2.2.22
#

Let \(\alpha \) be a root of \(x^8-2\). Then as we saw in Example 2.2.20 we know that \(B=\{ 1,\alpha ,\dots ,\alpha ^7\} \) has discriminant \(-2^{31}\). So the we only need to check at \(p=2\). But \(m_\alpha \) satisfies Eisensteins criterion 1.2.15 with \(p=2\), therefore the Lemma 2.2.21 tells us that in fact this must be an integral basis, since dividing by \(2\) wont ever give us new algebraic integers.