5
$\begingroup$

I was wondering if there were general "tactics" to show if an ideal in the ring of integers $\mathcal{O}_K$, where $K/\mathbb{Q}$ is a number field of degree $n$.

For example, consider $\mathcal{O}_K = \mathbb{Z}[\sqrt{-5}]$ and $P = (2,1+\sqrt{5})$. One method to show this is a proper idea is to argue that if $a + b\sqrt{-5} \in P,$ then $a+b\sqrt{-5} = 2x + y(1+\sqrt{-5})$ for integers $x,y$ and comparing coefficients, we get $a - b \equiv 0 \bmod 2,$ so $P$ cannot be the whole ring. However, this method seems quite laborious, and this may just be a special case when modding out by a prime actually works.

For me, the following isomorphism seems quite intuitive: as $\mathbb{Z}[\sqrt{-5}] \cong \mathbb{Z}[x]/(x^2+5),$ we should have that \begin{align*} \mathbb{Z}[\sqrt{-5}]/(2,1+\sqrt{5}) &\cong \mathbb{Z}[x]/(x^2+5,2,1+x) \\ &\cong \mathbb{Z}/(6,2) \text{ by mapping } x \mapsto -1 \\ &\cong \mathbb{Z}/2\mathbb{Z} \end{align*} so the ideal $P$ is indeed proper. But I can't seem to set up an explicit isomorphism at the moment, so I was wondering if this works generally?

Furthermore, are there any useful techniques one might use to show an ideal is indeed proper?

$\endgroup$
  • 1
    $\begingroup$ Could you please explain what you mean when you say that you can't set up an explicit isomorphism? Doesn't your isomorphism send a coset $a + b\sqrt{-5} + (2, 1+\sqrt{-5})$ in $\mathbb Q [\sqrt{-5}]/(2,1+\sqrt{5})$ to the element $a - b \in \mathbb Z / 2 \mathbb Z$ (i.e. we think of $\sqrt{5}$ as $x$, then map $x \mapsto -1$, like in your second method)? And doesn't this agree perfectly with your first method? $\endgroup$ – Kenny Wong May 31 '17 at 23:43
  • 1
    $\begingroup$ There are other ways to see that $(2, 1+\sqrt{-5})$ is a proper ideal. For example, we can stare at the ideal-theoretical equation $(2)(3) = (1+\sqrt{-5})(1 - \sqrt{-5})$ and deduce that any prime ideal that divides $(2)$ must also divide $(1+\sqrt{-5})$ or $(1-\sqrt{-5})$, and hence, any such prime ideal divides $(2, 1+\sqrt{-5}) = (2, 1- \sqrt{-5})$. Or we can apply Dedekind's factorization theorem to show that $(2) = (2, 1+\sqrt{-5})^2$. But admittedly, these methods don't work for every ideal in every ring of integers. $\endgroup$ – Kenny Wong May 31 '17 at 23:47
  • $\begingroup$ your isomorphism isn't intuitive at all considering that if you map $x$ to $2$ instead you get something completely different. $\endgroup$ – mercio Jun 1 '17 at 9:20
  • $\begingroup$ @user1952009 I'm sure you know much more about this than me! For the second method, we're essentially testing whether or not $2$, $1+x$ and $x^2+1$ have g.c.d. $=1$, and it would be natural to use the Euclidean algorithm. I wonder if this is essentially equivalent to using Grobner bases? (I can't remember how the Grobner basis algorithm works off-hand...) $\endgroup$ – Kenny Wong Jun 1 '17 at 9:23
  • $\begingroup$ @mercio Could you please explain why we would choose to map $x \mapsto 2$ here? The reason we map $x \mapsto -1$ is that $x+ 1$ is in the ideal, so we're quotienting by $x + 1$. (We could just as well map $x \mapsto 85$, because $x + 85 = (x+1)+42\times 2$ is in the ideal, and ultimately we'll get the same answer. But I don't think we can map $x \mapsto 2$...) $\endgroup$ – Kenny Wong Jun 1 '17 at 9:25
2
$\begingroup$

I'm sure there is a much better way, but yes it works generally, assuming you know the multiplication law in $\mathcal{O}_K$ seen as a free $\mathbb{Z}$ module.

Find $r \in I \cap \mathbb{Z}$, list all the elements of the finite ring $$R_0 = \mathcal{O}_K/(r)$$ then write $I = (u_1,\ldots,u_n)$ and list all the elements of the finite quotient rings $$R_1= R_0/(u_1), \quad R_2= R_1/(u_2), \quad R_{m+1} = R_m/(u_{m+1})$$ You'll get $$I = \mathcal{O}_K \qquad \Longleftrightarrow \qquad \mathcal{O}_K / I = R_n = \{0\}$$


The more general setup is $$\mathcal{O}_K = \mathbb{Z}[X_1,\ldots,X_k]/J$$ for some ideal $J$, and we want to know if $$\mathcal{O}_K= I \qquad \Longleftrightarrow \qquad (I,J) = \mathbb{Z}[X_1,\ldots,X_k]$$ there is an algorithm for that using Gröbner basis

$\endgroup$
3
$\begingroup$

A "proper" ideal is any ideal "which is strictly smaller than the whole ring," right?

Then it's enough to show that the given ideal does not contain 1. If the ideal contains a unit, it must also contain 1 and must therefore be the whole ring.

In your example, we have $\mathfrak P = \langle 2, 1 + \sqrt{-5} \rangle$, which is simply the set of all numbers in this domain of the form $2x + y(1 + \sqrt{-5}) = a + b \sqrt{-5}$. The contribution of $2x$ to $a$ and $b$ is even. We can certainly choose $y$ so that $a = 1$, but then $b \neq 0$ like we want, since in fact $b$ must be odd for $a$ to be odd.

I can't set up an isomorphism either, that's a deficiency of mine, but it's kind of overkill if you just want to show that an ideal is proper.

$\endgroup$
3
$\begingroup$

To know is an ideal $\mathfrak{a}$ is proper or not isn't easier if we just compute the size of $\mathcal{O}_K/ \mathfrak{a}$ thinking of $\mathcal{O}_K$ as a lattice?.

As long as you know a $\mathbb{Z}$-basis for $\mathcal{O}_K$ say $\{ \gamma_1, \ldots, \gamma_n \}$ and you're given $\mathfrak{a}$ is terms of $\mathcal{O}_K$-generators say $ (\beta_1,\ldots,\beta_m)$ you can esaily find a $\mathbb{Z}$-basis $\{ \alpha_1, \ldots, \alpha_n \}$ for $\mathfrak{a}$ ( reducing the $\mathbb{Z}$-generators of $\mathfrak{a}$ $ \{\beta_i \gamma_j \}$ to a $\mathbb{Z}$-basis, for example by hermite form). Then you simply write $\alpha_i=\sum a_{ij} \cdot \gamma_j$ and $ | \mathcal{O}_K/ \mathfrak{a} |=|\text{det}(a_{ij})|$ would be $1$ iff $\mathcal{O}_K=\mathfrak{a}$.

In your example $\{1, \sqrt{-5} \}$ is a $\mathbb{Z}$-basis for $\mathcal{O}_K$, $\{2,2\sqrt{-5},1+\sqrt{-5}, (1+\sqrt{-5})\sqrt{-5} \}$ reduces to the $\mathbb{Z}$-basis $ \{1+\sqrt{-5},2 \sqrt{-5}\}$ for $P$, because the hermite form of $$ \begin{bmatrix} 2 & 0 \\ 0 & 2 \\ 1 & 1\\ -5 & 1 \end{bmatrix} $$ is the matrix

$$ \begin{bmatrix} 1 & 1 \\ 0 & 2 \\ 0 & 0\\ 0 & 0 \end{bmatrix} $$

and so $ | \mathcal{O}_K/ P |=\text{det}\begin{bmatrix} 1 & 1 \\ 0 & 2 \end{bmatrix}$$=2\neq 1$.

$\endgroup$
  • $\begingroup$ this is what does sagemath (see for example the function __cmp__ where self.pari_rhnf is the Hermite normal form of the ideal) $\endgroup$ – reuns Jun 2 '17 at 17:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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