Integration over the general linear group over $\mathbb Q_p$

52 Views Asked by At

Before asking, I apologize if this question is duplicated. It seems a sort of basic stuff and I tried to find the reference, but I couldn't find it(only the statement without the proof appears in the textbook of Goldfeld and Hundley, Automorphic representations and L-functions for general linear groups).

Let $G:=GL(n,\mathbb Q_p)$, which is a topological group and $K_j:=I_n+p^j M_n(\mathbb Z_p)$. It is known that $\{K_j\}_{j \geq 1}$ forms a base of neighborhoods of $I_n$ (this comes from the fact that $p^j \mathbb Z_p$ form a base of $0$ in $\mathbb Q_p$. $M_n(\mathbb Q_p)$ is just a product space of $\mathbb Q_p$, and $G$ is its subspace). It is also known that $K_j$'s are open, closed, and compact (which is not hard to verify).

The Haar measure $\mu$ on $G$ is also well-defined by setting that $$ \mu(gK_j)=\mu(K_j g)=\frac{1}{[K_0:K_j]}\mu(K_0) $$ for any $g \in G$, where $K_0=GL(n,\mathbb Z_p)$.

If $f:G \to \mathbb C$ is a smooth(i.e., locally constant) function, then we may define the integral of $f$ over $G$ $$ \int_G f(y)d\mu(y), $$ which is left and right invariant.

I would like to check that this integration always converges. Since $f$ is locally constant, we may choose $K_{j_x}$ for each $x \in G$ and $c_x$ such that if $y\in xK_{j_x}$ then $f(y)=c_x$.

We may assume that the union $\bigcup_x xK_{j_x}$ are disjoint. Indeed, if $xK_{j_x}$ and $yK_{j_y}$ have the non-empty intersection, then $xK_{j_x} \subseteq yK_{j_y}$ without loss of generality so that $c_x=c_y$.

Therefore, the integral of $f$ over $G$ is by definition $$ \sum_x c_x\mu(K_{j_x}). $$

However, I don't know how one can prove that this sum converges. I guess that $[K_j:K_{j+1}]=[K_{j-1}:K_j]$ for $j \geq 1$ so that the sum is something like a geometric series, but how I show that $\mu_{K_{j_x}}$ is really 'shrinking' if one properly order the sum?

Thank you so much for reading and attending to this question.

1

There are 1 best solutions below

1
On

Aphelli pointed out that we need the 'compactly supported' condition. If we have, then the sum is just a finite sum so the assertion is obvious.

However, the assertion is not true in general since we can easily construct a counter-example by taking large $c_x$. The question is closed.