Pages

Thursday, February 23, 2017

Notes on Algebraic Number Theory (II: Some Explicit Examples for Integral Basis)

In this note, we will mention some examples about the ring of integers and its integral basis.

1. Cyclotomic fields. We will fix some notions: $p$ is an odd prime number, $\zeta=e^{2\pi i/p}$ is the primitive $p^{\text{th}}$ root of unity, $K=\mathbb{Q}(\zeta)$. Recall that $\mathscr{O}_K=\mathbb{Z}[\zeta]$, and $\Delta_K=(-1)^{(p-1)/2}p^{p-2}$.

One can see that $\zeta+\zeta^{-1}=e^{2\pi i/p}+e^{-2\pi i/p}=2\cos2\pi/p\in \mathbb{R}$. Hence, $L:=\mathbb{Q}(\zeta+\zeta^{-1})$ is a real subfield of $K$. We will prove that $[K:L]=2$. We first see that $G:=Gal(K/\mathbb{Q})\cong \mathbb{F}_P^\times$, a cyclic group of order $p-1$, an even number. And hence, $G$ has only one subgroup of order 2. Furthermore, the automorphism $K\to K$ sending $\zeta\mapsto \zeta^{-1}$ is an element of order 2 in $G$, and $L$ is fixed by this automorphism.

On the other hand, $\cos 2\pi/p\ne \cos 2\pi k/p$, for any integer $k$ such that $1<k<p-1$, i.e. $\zeta+\zeta^{-1}\ne \zeta^k+\zeta^{-k}$ for all $1<k<p-1$. Say another words, $L$ is not fixed by any automorphism of $K$ sending $\zeta$ to $\zeta^k$. Hence, by the Galois' correspondence, $Gal(K/L)$ corresponds to the field extension $K/L$. This yields $[K:L]=2$.

Also, one can see that for any real subfield $F$ of $K$, $F$ is fixed by the automorphism $\zeta\mapsto \zeta^{p-1}$. By the Galois' correspondence again, $F\subset L$. We have the following

Proposition 1. Let $K,L$ be fields as above, then $L$ is the maximal real subfield of $K$. Also, $\mathscr{O}_L=\mathbb{Z}[\zeta+\zeta^{-1}]$.

Proof. Let $\psi:=\zeta+\zeta^{-1}. $It is easy to see that $\mathbb{Z}[\psi]\subset \mathscr{O}_L$, because $\psi$ is integral over $\mathbb{Z}$. Assume that we have the strict inclusion, and there exists $\alpha:=a_0+a_1\psi+...+a_n\psi^{n}\in \mathscr{O}_L, n\le (p-1)/2$, such that $\forall j, 1\le j\le n, a_j\in\mathbb{Q}$, and there exists $a_i\notin \mathbb{Z}$. If $a_n\in \mathbb{Z}$, we can subtract $\alpha-a_n\psi^n$ to obtain another element in $\mathscr{O}_L$, and we continue this process to eliminate all leading coefficients in $\mathbb{Z}$. Hence, without loss of generality, we can assume $a_n\notin \mathbb{Z}$. Now,
$$\zeta^n\alpha=\zeta^n(a_0+a_1(\zeta+\zeta^{-1})+...+a_n(\zeta+\zeta^-1)^n)=...+a_n\zeta^{2n}$$
That means, the leading coefficient of $\zeta^n\alpha$ by representing it as a polynomial of variable $\zeta$ is $a_n\zeta^{2n}$. However, $\mathscr{O}_L\subset\mathscr{O}_K=\mathbb{Z}[\zeta]$, and hence, both $\zeta,\alpha\in \mathbb{Z}[\zeta]$. This yields $a_n\in \mathbb{Z}$, a contradiction to our assumption. Therefore, $\mathscr{O}_L=\mathbb{Z}[\zeta+\zeta^{-1}]$.

(Q.E.D)

2. Non-monogenic ring of integers. One can see that both $\mathscr{O}_K, \mathscr{O}_L$ are monogenic, i.e. they have the form $\mathbb{Z}[\alpha]$ for some $\alpha$ integral in $K$. But this does not hold in general. We will consider a counter example by Dedekind.

Proposition 2. Let $\alpha$ be a root of an equation $f(x):=x^3 - x^2-2x-8=0$, and $K:=\mathbb{Q}(\alpha)$, then $\mathscr{O}_K$ is not monogenic.

Proof. Denote $d(\mathbb{Z}[\alpha])=D(1,\alpha,\alpha^2)$, then it can be seen that $d(\mathbb{Z}[\alpha])=-4.503$, and because $503$ is a prime, one can see by our Part 1 that the possible primes that are ramified in $\mathscr{O}_K$ are $2$ or $503$. By using Kummer-Dedekind's theorem, we can detect that $2$ is singular on $\mathbb{Z}[\alpha]$. By our previous note, we also know that $d(\mathbb{Z}[\alpha])=m^2\Delta_K$, where $m$ is the index $[\mathscr{O}_K:\mathbb{Z}[\alpha]]$ as $\mathbb{Z}$-module. Hence, $\Delta_K=-503$. From this, one can see $2$ is unramified in $\mathscr{O}_K$. Because $N(2)=8$, we have three cases:
(1) $2$ is inert in $\mathscr{O}_K$.
(2) $(2) = \mathfrak{p}_2\mathfrak{p}_4$ with $\mathfrak{p}_i$ denotes the prime ideal of norm $i$
(3) $(2) =\mathfrak{p}_2\mathfrak{q}_2\mathfrak{r}_2$, i.e. product of some prime ideals of norm 2.

Note that the minimum polynomial of $\alpha+1$ is $(x+1)^3-(x+1)^2-2(x+1)-8=0$, hence, $N(\alpha+1)=-10$, i.e. it can be factorized as $\mathfrak{p_2}{q}_5$, because $\mathscr{O}_K/(\alpha+1)=N(\alpha+1)=10$, and no finite field has 10 element. That means, there exists $\mathfrak{p}_2$ lying over $2$, and (1) cannot happen. Assume that (2) happens, i.e. $(2)=\mathfrak{p}_2\mathfrak{p}_4$, and they are all prime ideals lying over 2 in $\mathscr{O}_K$. We can also see that $N(\alpha)=8$, and $(\alpha)$ cannot be a prime ideal, since otherwise, $\mathfrak{p}=(\alpha)$ lies over 2, with $N(\mathfrak{p})=N(2)$, i.e. (2) is a prime ideal, that cannot happen.  Hence, we must have $(\alpha)=\mathfrak{p}_2^3$, or $\alpha=\mathfrak{p}_2\mathfrak{p}_4$. In any case, we have $\gcd((\alpha),(\alpha+1))=\mathfrak{p}_2$, a contradiction, because the two ideal are co-prime. This yields (1) and (2) cannot happen.

And hence, $(2)$ is a product of three distinct prime ideals of norm 2 in $\mathscr{O}_K$. If there exists $\beta\in \mathscr{O}_K$, such that $\mathscr{O}_K=\mathbb{Z}[\beta]$, we can see, by Kummer-Dedekind's theorem that if $f_\beta(x)$ is the minimum polynomial of $\beta$, then $f_\beta$ will have three distinct root in $\mathbb{F}_2$, which is absurd, because $\mathbb{F}_2$ just has 2 elements. This shows $\mathscr{O}_K$ is not monogenic.

(Q.E.D)


Actually, we can point out the integral basis for $\mathscr{O}_K$, by noting that $\beta:=(\alpha+\alpha^2)/2$ is in $\mathbb{O}_K$, and $D(1,\alpha,\beta)=-503=\Delta_K$. Hence, $\mathbb{Z}[1,\alpha,\beta]=\mathscr{O}_K$.

3. Short note on integral basis of cubic fields. We begin this section by an interesting

Lemma 3.1. Let $f(x)=x^n+a_{n-1}x^{n-1}+...+a_1x+a_0\in\mathbb{Z}[x]$ be an Eisentein polynomial, i.e., there exists a prime number $p$ such that $p|a_i$ for all $0\le i\le n-1$, and $p^2\nmid a_0$, then $p$ is not singular in $\mathbb{Z}[\alpha]$, where $\alpha$ is a root of $f(x)=0$, or equivalently, $p\nmid [\mathscr{O}_K:\mathbb{Z}[\alpha]]$.

Proof. It follows directly from the Kummer-Dedekind's theorem (Q.E.D)

Now, we turn to the integral basis of a cubic field $\mathbb{Q}(\alpha)$, where $\alpha$ is a root of $f(x):=x^3 -r$, for some $r=ab^2$ is an integer, where $ab$ is square free (which does imply $a,b$ are square free and $\gcd(a,b)=1$), and if $3|r$, then $3|a$.

It can be seen first that the discriminant of $f$, i.e. $D(1,\alpha,\alpha^2)=N(f'(\alpha))=-27r^2=-3^3r^2$. And hence, $-27r^2=m^2\Delta_K$, where $m$ is the index $[\mathscr{O}_K:\mathbb{Z}[\alpha]$. First, if $p|a$, then $f(x)$ is an Eisentein polynomial respect to $p$, i.e. $p\nmid m$, by the previous lemma, and hence, $p^2|\Delta_K$, which yields, $a^2|\Delta_K$. If $3|r$, then $3|a$, and hence, $27a^2|\Delta_K$.

Because $\alpha^3 = ab^2$, we have $\alpha^6 = a^2b^4$, i.e. $\frac{\alpha^2}{b}=a^2b$. Let $\beta:=\frac{\alpha^2}{b}$, $\beta$ has the minimum polynomial $g(x):=x^3 - a^2b$. Hence, $\beta\in \mathscr{O}_K$. Let $m'$ be the index $[\mathscr{O}_K:\mathbb{Z}[\beta]]$, we have $d(\mathbb{Z}[\beta])=-27a^4b^2=m'^2\Delta_K$. For any $p|b$, $g(x)$ is an Esisentein polynomial with respect to $p$, by Lemma 3.1, we have $p\nmid m'$, and $p^2|\Delta_K$, which yields $b^2|\Delta_K$.

In short, if $3|r$, we have $\Delta_K = -27r^2=d(\mathbb{Z}[\alpha])$, or equivalently, $\mathscr{O}_K=\mathbb{Z}[\alpha]$, when $r\equiv 0\mod 3$. Otherwise, $\Delta_K=-3r^2$, or $\Delta_K=-27r^2$. That means, in any case, $3$ is ramified in $\mathscr{O}_K$, and $(3)=\prod_{i=1}^n\mathfrak{p}_i^{e_i}$, with some $e_i>1$. Furthermore, in the first and the third case, we can see $m=1$, and in the second case $m=3$, i.e. 3 is singular on $\mathbb{Z}[\alpha]$. And one can realize that to compute explicitly $\Delta_K$ by given $r$, we must look at 3 in $\mathbb{Z}[\alpha]$. The polynomial $f(x) = x^3 - r\in \mathbb{F}_3[x]$ can be factorized as $x^3, (x-1)^2(x+1)$ or $(x+1)^2(x-1)$, with respect to three cases $r\equiv 0,1,2\mod 3$. The case $3|r$ is treated by our above argument. We consider $r\equiv 1,-1\mod 3$. If $r\equiv 1\mod 3$, then by Kummer-Dedekind's theorem, the remainder when we divide $x^3 - r$ by $x-1$ is $r-1$, and $3$ is singular on $\mathbb{Z}[\alpha]$ iff $r\equiv1\mod 9$. Similarly, when $r\equiv -1\mod 3$, then $3$ is sunglar on $\mathbb{Z}[\alpha]$ iff $r\equiv -1\mod 9$.

In short, what we have done so far can be summarized as follows. If $r\not\equiv\pm1\mod 9$, then $\mathscr{O}_K=\mathbb{Z}[\alpha]$, and the integral basis of $\mathscr{O}_K$ in this case is just $1,\alpha,\alpha^2$. Otherwise, if $r\equiv\pm 1$, then $\Delta_K=-3r^2$, and $(3)$ is singular on $\mathbb{Z}[\alpha]$.

If $r\equiv 1\mod 9$, then it can be seen that
$$\beta:=\frac{1+\alpha+\alpha^2}{3}=\frac{\alpha^3-1}{3(\alpha-1)}=\frac{r-1}{3(\alpha-1)}$$
is in $\mathscr{O}_K$. And it is easy to see that $D(1,\alpha,\beta)=\Delta_K$, and in the integral basis in this case is $(1,\alpha,\frac{1+\alpha+\alpha^2}{3})$.

If $r\equiv-1\mod 9$, then it can be seen that $\frac{1-\alpha+\alpha^2}{3}$ is in $\mathscr{O}_K$, and the integral basis in this case is $(1,\alpha,\frac{1-\alpha+\alpha^2}{3})$.

4. The integral basis of multi-quadratic fields. If $m$ is a square free integers, and $K:=\mathbb{Q}(\sqrt{m})$ be a quadratic field, it can be easily proved that $\mathscr{O}_K=\mathbb{Z}[\sqrt{m}]$ if $m\equiv 2,3\mod 4$, and $\mathscr{O}_K=\mathbb{Z}[\frac{1+\sqrt{m}}{2}]$ if $m\equiv 1\mod 4$. Let $n$ be a square free integer, what is the integral basis for $\mathscr{O}_L$, for $L=\mathbb{Q}(\sqrt{m},\sqrt{n})$? We will partly answer this question in this section.  We begin with the following important

Proposition 4.1. Let $K, L$ be two number fields, and $KL$ the smallest field containing both $K, L$. Let $d=\gcd(\Delta_K,\Delta_L)$, then $\mathscr{O}_{KL}\subset \frac{1}{d}\mathscr{O}_K\mathscr{O}_L$.

Proof.

Via this proposition, we obtain

Corollary 4.2. Let $K,L, KL$ be the same in Proposition, and $(\alpha_1,...,\alpha_n)$ the integral basis for $\mathscr{O}_K$, and $(\beta_1,...,\beta_m)$ the integral basis for $\mathscr{O}_L$, then $(\alpha_i\beta_j)_{ij}$ is the integral basis for $\mathscr{O}_{KL}$ if $\gcd(\Delta_K,\Delta_L)=1$.

Proof. Easily deduce from the proof of Proposition 4.1.

And we easily obtain the following

Corollary 4.3. Let $p,q$ be two distinct prime numbers, with $p,q\equiv 1\mod 4$, then $\mathscr{O}_{\mathbb{Q}(\sqrt{p},\sqrt{q})}=\mathbb{Z}[\alpha,\beta,\alpha\beta]$, where $\alpha=\frac{1+\sqrt{p}}{2}, \beta=\frac{1+\sqrt{q}}{2}$.

Proof. It can be easily seen that $\mathscr{O}_{\mathbb{Q}(\sqrt{p})}=\mathbb{Z}[\alpha]$, and $\mathscr{O}_{\mathbb{Q}(\sqrt{q})}=\mathbb{Z}[\beta]$. Furthermore, $\Delta_{\mathbb{Q}(\sqrt{p})}=p$, and $\Delta_{\mathbb{Q}(\sqrt{q})}=q$, and hence, $\gcd(p,q)=1$. It follows by Corollary 4.2 that $\mathscr{O}_{\mathbb{Q}(\sqrt{p},\sqrt{q})}=\mathbb{Z}[\alpha,\beta,\alpha\beta]$.

(Q.E.D)

No comments:

Post a Comment