Let $\Gamma$ be an arithmetic lattice in a linear algebraic $\mathbb{Q}$-group $\mathbf{G}$, that is, $\Gamma$ is a subgroup of $\mathbf{G}(\mathbb{Q})$ that is commensurable with $\mathbf{G}(\mathbb{Z})$.

For a prime $p$, we can consider $\Gamma$ as a subspace of $\mathbf{G}(\mathbb{Q}_p)$. My question is:

What does the closure of $\Gamma$ in $\mathbf{G}(\mathbb{Q}_p)$ with respect to the $p$-adic topology look like?

The closure of $\mathbb{Z}$ in $\mathbb{Q}_p$ is the ring of $p$-adic integers $\mathbb{Z}_p$. So it seems plausible to me that, for example, the closure of the lattice $\Gamma = SL(n,\mathbb{Z})$ in the $p$-adic topology of $SL(n,\mathbb{Q}_p)$ would be $SL(n,\mathbb{Z}_p)$. Is this correct?

Also, what about other lattice, for example, what is the closure of a congruence subgroup $$\Gamma(c) := \{ g \in SL(n,\mathbb{Z}) : g - I_n \equiv 0 \;\text{ mod } c\, \} \subset SL(n,\mathbb{Z})$$ in the $p$-adic topology of $SL(n,\mathbb{Q}_p)$?


Suppose $G$ is $\mathbb Q$ simple (i.e. has no connected normal algebraic subgroups which are defined over $\mathbb Q$) and is simply connected (i.e. $G(\mathbb C)$ is simply connected). Assume also that $G(\mathbb R)$ is not compact. With these assumptions, the closure of an arithmetic lattice in $G({\mathbb Z}_p)$ is an open subgroup. This statement is known as strong approximation.

More generally, if $G(\mathbb Z)$ is Zariski dense in $G$, and $G(\mathbb C)$ is connected and simply connected, then the closure of a finite index subgroup of $G(\mathbb Z)$ is open in $G(\mathbb Z _p)$.

Examples are $G=SL_n$ and $Sp_{2n}$. But not $PGL_n$ (this is not simply connected).

In your example, the closure of $SL(n,\mathbb Z)$ is indeed $SL(n,\mathbb Z _p)$; this can be proved by using the fact that $SL(n,\mathbb Z), SL(n,\mathbb Z _p)$ are generated by unipotent elements.


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.