Discriminant of Orders in Quaternion Algebra "to divide is to contain"

75 Views Asked by At

I'm reading Bergeron's book "The Spectrum of Hyperbolic Surfaces". In chapter 8, he talks about orders of quaternion algebras. I am interested in Proposition 8.3 which states

Let $\mathcal{O}$ and $\mathcal{O}'$ be two orders of $A$ (a quaternion algebra over $\mathbb{Q}$). If $\mathcal{O}\supset\mathcal{O}'$, then $d(\mathcal{O})$ divides $d(\mathcal{O}')$. Moreover, if $\mathcal{O}\supset\mathcal{O}'$ and $d(\mathcal{O})=d(\mathcal{O}')$, then $\mathcal{O}=\mathcal{O}'$.

The proof is omitted but it is remarked that it follows from a previous proposition that says for an order $\mathcal{O}$, we have that $\mathcal{O}\cap\mathbb{Q}=\mathbb{Z}$.

In the above, we have that $d(\mathcal{O})$ is the reduced discriminant. It is shown that if $(e_1,e_2,e_3,e_4)$ is a basis for the order $\mathcal{O}$, and $P$ is the change of basis matrix from $(1,i,j,k)$ to $(e_1,e_2,e_3,e_4)$, then we have that $d(\mathcal{O})=2^2ab(\det P)$ where we have that the quaternion algebra $A=D_{a,b}(\mathbb{Q})$ (i.e. with $i^2=a,j^2=b$).

I am having trouble seeing a couple of things. First, they claim that since $\text{disc}(\mathcal{O})\in\mathbb{Z}$ and $\det P\in\mathbb{Q}$ this implies that $d(\mathcal{O})\in\mathbb{Z}$. It is clear to me why $\text{disc}(\mathcal{O})\in\mathbb{Z}$, but it isn't clear why $\det P\in\mathbb{Q}$ since I could see a world where this lies in some quadratic extension of $\mathbb{Q}$ since we have that $$ \text{disc}(\mathcal{O})=-2^4a^2b^2(\det P)^2 $$ I see that the left-hand side of this is an integer, so I see that $(\det P)^2\in\mathbb{Q}$, but it isn't clear to me why $\det P\in\mathbb{Q}$. If this is true, I then see why $d(\mathcal{O})=2^2ab(\det P)\in\mathbb{Z}$. It seems to me that for some reason $P$ being the change of basis into an order must be what forces $\det P$ to actually be rational, but if there is a more concrete explanation as to why this is true that would help clear that up for me.

The other thing I am having trouble seeing is the first part of the proposition, and why $\mathcal{O}\cap \mathbb{Q}=\mathbb{Z}$ implies this "to divide is to contain" sort of proposition. I can understand heuristically why it should be true in that we have that if $P$ is the change of basis matrix for $\mathcal{O}$ and $Q$ is the change of basis matrix for $\mathcal{O}'$, then the division criterion on the reduced discriminant is equivalent to saying that $\det P\mid \det Q$. I think the idea is that since $\mathcal{O}\supset\mathcal{O}'$ we will have that $\mathcal{O'}$ is the larger lattice, so it has a larger fundamental parallelogram, and the quotient of the determinants should be the index of the lattice.

If someone could help clear these two points up for me, I would greatly appreciate it.