Skip to main content Accessibility help
×
×
Home

Contents:

Information:

  • Access

Actions:

MathJax
MathJax is a JavaScript display engine for mathematics. For more information see http://www.mathjax.org.
      • Send article to Kindle

        To send this article to your Kindle, first ensure no-reply@cambridge.org is added to your Approved Personal Document E-mail List under your Personal Document Settings on the Manage Your Content and Devices page of your Amazon account. Then enter the ‘name’ part of your Kindle email address below. Find out more about sending to your Kindle. Find out more about sending to your Kindle.

        Note you can select to send to either the @free.kindle.com or @kindle.com variations. ‘@free.kindle.com’ emails are free but can only be sent to your device when it is connected to wi-fi. ‘@kindle.com’ emails can be delivered even when you are not connected to wi-fi, but note that service fees apply.

        Find out more about the Kindle Personal Document Service.

        Ergodic measures on spaces of infinite matrices over non-Archimedean locally compact fields
        Available formats
        ×

        Send article to Dropbox

        To send this article to your Dropbox account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your <service> account. Find out more about sending content to Dropbox.

        Ergodic measures on spaces of infinite matrices over non-Archimedean locally compact fields
        Available formats
        ×

        Send article to Google Drive

        To send this article to your Google Drive account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your <service> account. Find out more about sending content to Google Drive.

        Ergodic measures on spaces of infinite matrices over non-Archimedean locally compact fields
        Available formats
        ×
Export citation

Abstract

Let $F$ be a non-discrete non-Archimedean locally compact field and ${\mathcal{O}}_{F}$ the ring of integers in $F$ . The main results of this paper are the classification of ergodic probability measures on the space $\text{Mat}(\mathbb{N},F)$ of infinite matrices with entries in $F$ with respect to the natural action of the group $\text{GL}(\infty ,{\mathcal{O}}_{F})\times \text{GL}(\infty ,{\mathcal{O}}_{F})$ and the classification, for non-dyadic $F$ , of ergodic probability measures on the space $\text{Sym}(\mathbb{N},F)$ of infinite symmetric matrices with respect to the natural action of the group $\text{GL}(\infty ,{\mathcal{O}}_{F})$ .

1 Introduction

Given a group action on a topological space, it is natural to try to describe the corresponding space of ergodic invariant probability measures. For some very classical actions, such as, for example, that of the shift on the space of infinite binary sequences, the space of ergodic measures is huge and does not seem to admit a reasonable description. On the other hand, for a number of natural actions of infinite-dimensional groups, a complete classification is possible. For example, De Finetti’s Theorem (1937) claims that for the action of the infinite symmetric group on the space of infinite binary sequences, all ergodic probability measures are Bernoulli, and Schoenberg’s Theorem (1951) claims that for the action of the infinite orthogonal group on the space of infinite $\mathbb{R}$ -valued sequences, all ergodic probability measures are Gaussian. In both these examples, the space of ergodic probability measures is one dimensional. Pickrell [Pic87, Pic90, Pic91] and, by a different method, Olshanski and Vershik [OV96], classified all ergodic unitarily invariant measures on the space of infinite Hermitian matrices. In this case, an ergodic measure is determined by infinitely many parameters.

In this paper, we study classification of ergodic measures for actions related to the following inductive limit group

(1) $$\begin{eqnarray}\displaystyle \operatorname{GL}(\infty ,{\mathcal{O}}_{F}):=\lim _{\longrightarrow }\operatorname{GL}(n,{\mathcal{O}}_{F}), & & \displaystyle\end{eqnarray}$$

where ${\mathcal{O}}_{F}$ is the ring of integers in a non-discrete locally compact non-Archimedean field $F$ and $\operatorname{GL}(n,{\mathcal{O}}_{F})$ is the compact group of invertible $n\times n$ matrices over ${\mathcal{O}}_{F}$ . Denote by $\operatorname{Mat}(\mathbb{N},F)$ (respectively $\operatorname{Sym}(\mathbb{N},F)$ ) the space of infinite matrices (respectively infinite symmetric matrices) over $F$ . Our main results are:

  1. (i) the classification of the ergodic probability measures for the group action of $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ on $\operatorname{Mat}(\mathbb{N},F)$ defined by

    $$\begin{eqnarray}\displaystyle ((g_{1},g_{2}),M)\mapsto g_{1}Mg_{2}^{-1},\quad g_{1},g_{2}\in \operatorname{GL}(\infty ,{\mathcal{O}}_{F}),\quad M\in \operatorname{Mat}(\mathbb{N},F); & & \displaystyle \nonumber\end{eqnarray}$$

  2. (ii) the classification of the ergodic probability measures for the group action of $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ on $\operatorname{Sym}(\mathbb{N},F)$ defined by

    $$\begin{eqnarray}\displaystyle (g,M)\mapsto gMg^{t},\quad g\in \operatorname{GL}(\infty ,{\mathcal{O}}_{F}),\quad S\in \operatorname{Sym}(\mathbb{N},F), & & \displaystyle \nonumber\end{eqnarray}$$
    where $g^{t}$ is the transposition of  $g$ .

We proceed to the precise formulation. Let $F$ be a non-discrete locally compact non-Archimedean field (for example, the field of $p$ -adic numbers). Let $|\cdot |$ be the absolute value on $F$ . The ring ${\mathcal{O}}_{F}$ of integers in $F$ is given by $\{x\in F:|x|\leqslant 1\}$ . The unique maximal and principal ideal of ${\mathcal{O}}_{F}$ is given by $\{x\in F:|x|<1\}$ . Throughout the paper, we fix any generator $\unicode[STIX]{x1D71B}$ of $\{x\in F:|x|<1\}$ , that is, $\{x\in F:|x|<1\}=\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}$ . The quotient ${\mathcal{O}}_{F}/\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}$ is a finite field with $q$ elements.

Define the inductively compact group $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ by (1). Set

$$\begin{eqnarray}\displaystyle \operatorname{Mat}(\mathbb{N},F):=\{X=(X_{ij})_{i,j\in \mathbb{N}}\mid X_{ij}\in F\}. & & \displaystyle \nonumber\end{eqnarray}$$

Let $\operatorname{Mat}(\infty ,F)$ denote the subspace of $\operatorname{Mat}(\mathbb{N},F)$ consisting of matrices whose all but a finite number of coefficients are zero. Define also

$$\begin{eqnarray}\displaystyle \operatorname{Sym}(\mathbb{N},F):=\{X\in \operatorname{Mat}(\mathbb{N},F)\mid X_{ij}=X_{ji},\forall i,j\in \mathbb{N}\}, & & \displaystyle \nonumber\end{eqnarray}$$

and let $\operatorname{Sym}(\infty ,F):=\operatorname{Sym}(\mathbb{N},F)\cap \operatorname{Mat}(\infty ,F)$ .

Throughout the paper, by the usual identification $\operatorname{Mat}(\mathbb{N},F)\simeq F^{\mathbb{N}\times \mathbb{N}}$ , we equip $\operatorname{Mat}(\mathbb{N},F)$ with the Tychonoff’s product topology. The set $\operatorname{Sym}(\mathbb{N},F)$ is equipped with the induced topology by the natural embedding $\operatorname{Sym}(\mathbb{N},F)\subset \operatorname{Mat}(\mathbb{N},F)$ .

1.1 Classification of ergodic measures on $\operatorname{Mat}(\mathbb{N},F)$

Let $\unicode[STIX]{x1D6E5}$ be the set of non-increasing sequences in $\mathbb{Z}\cup \{-\infty \}$ , that is,

(2) $$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D6E5}:=\{\Bbbk =(k_{j})_{j=1}^{\infty }\mid k_{j}\in \mathbb{Z}\cup \{-\infty \};k_{1}\geqslant k_{2}\geqslant \cdots \,\}. & & \displaystyle\end{eqnarray}$$

By the inclusion $\unicode[STIX]{x1D6E5}\subset (\mathbb{Z}\cup \{-\infty \})^{\mathbb{N}}$ , we equip $\unicode[STIX]{x1D6E5}$ with the induced topology of the Tychonoff’s product topology on $(\mathbb{Z}\cup \{-\infty \})^{\mathbb{N}}$ .

To each sequence $\Bbbk \in \unicode[STIX]{x1D6E5}$ , we now assign an ergodic $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measure on $\operatorname{Mat}(\mathbb{N},F)$ . Let

$$\begin{eqnarray}\displaystyle X_{i}^{(n)},\quad Y_{i}^{(n)},\quad Z_{ij},\quad i,j,n=1,2,\ldots & & \displaystyle \nonumber\end{eqnarray}$$

be independent random variables, each sampled with respect to the normalized Haar measure on the compact additive group ${\mathcal{O}}_{F}$ . In what follows, we use the convention $\unicode[STIX]{x1D71B}^{\infty }=0$ .

Definition 1.1. Given an element $\Bbbk \in \unicode[STIX]{x1D6E5}$ , let

$$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D707}_{\Bbbk }:={\mathcal{L}}(M_{\Bbbk }) & & \displaystyle \nonumber\end{eqnarray}$$

be the probability distribution of the infinite random matrix $M_{\Bbbk }$ defined as follows. Write $k:=\lim _{n\rightarrow \infty }k_{n}\in \mathbb{Z}\cup \{-\infty \}$ and set

$$\begin{eqnarray}\displaystyle M_{\Bbbk }:=\biggl[\mathop{\sum }_{n:\,k_{n}>k}\unicode[STIX]{x1D71B}^{-k_{n}}X_{i}^{(n)}Y_{j}^{(n)}+\unicode[STIX]{x1D71B}^{-k}Z_{ij}\biggr]_{i,j\in \mathbb{N}}. & & \displaystyle \nonumber\end{eqnarray}$$

Remark 1.2. For the constant sequence $\Bbbk =(-\infty ,\ldots ,-\infty ,\ldots )\in \unicode[STIX]{x1D6E5}$ , using the convention $\unicode[STIX]{x1D71B}^{\infty }=0$ , the corresponding matrix $M_{\Bbbk }$ defined in Definition 1.1 is the zero matrix $O\in \operatorname{Mat}(\mathbb{N},F)$ and hence the measure $\unicode[STIX]{x1D707}_{\Bbbk }$ is the Dirac measure at the point $O\in \operatorname{Mat}(\mathbb{N},F)$ .

Let ${\mathcal{P}}_{\text{erg}}(\operatorname{Mat}(\mathbb{N},F))$ be the space of ergodic $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measures on $\operatorname{Mat}(\mathbb{N},F)$ , endowed with the induced weak topology. The classification of ${\mathcal{P}}_{\text{erg}}(\operatorname{Mat}(\mathbb{N},F))$ is given by the following.

Theorem 1.3. The map $\Bbbk \mapsto \unicode[STIX]{x1D707}_{\Bbbk }$ is a homeomorphism between $\unicode[STIX]{x1D6E5}$ and ${\mathcal{P}}_{\text{erg}}(\operatorname{Mat}(\mathbb{N},F))$ .

Remark 1.4. By Theorem 1.3, the space ${\mathcal{P}}_{\text{erg}}(\operatorname{Mat}(\mathbb{N},F))$ is weakly closed in the space of all Borel measures on $\operatorname{Mat}(\mathbb{N},F)$ and is $\unicode[STIX]{x1D70E}$ -compact; moreover, any measure $\unicode[STIX]{x1D707}_{\Bbbk }\in {\mathcal{P}}_{\text{erg}}(\operatorname{Mat}(\mathbb{N},F))$ is compactly supported.

Let us explain Theorem 1.3 in more detail. We have the following elementary ergodic measures.

  1. (Haar type measures) For any $k\in \mathbb{Z}$ , the normalized Haar measure on $\operatorname{Mat}(\mathbb{N},\unicode[STIX]{x1D71B}^{-k}{\mathcal{O}}_{F})$ is $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -ergodic.

  2. (Non-symmetric Wishart type measures) Let $X_{1},Y_{1},X_{2},Y_{2},\ldots$ be independent and uniformly distributed on ${\mathcal{O}}_{F}$ . For any $k\in \mathbb{Z}$ , the probability distribution of the random matrix

    $$\begin{eqnarray}\displaystyle [\unicode[STIX]{x1D71B}^{-k}X_{i}Y_{j}]_{i,j\in \mathbb{N}} & & \displaystyle \nonumber\end{eqnarray}$$
    is $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -ergodic.

Theorem 1.3 implies that any ergodic $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measure on $\operatorname{Mat}(\mathbb{N},F)$ can be obtained as a possibly infinite convolution of the above two types of elementary ones.

1.2 Classification of ergodic measures on $\operatorname{Sym}(\mathbb{N},F)$

In what follows, when dealing with the ergodic measures on $\operatorname{Sym}(\mathbb{N},F)$ , we always assume that the field $F$ is non-dyadic, that is, the cardinality of the field of residue class ${\mathcal{O}}_{F}/\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}$ is not a power of 2.

The group of units of ${\mathcal{O}}_{F}$ is given by ${\mathcal{O}}_{F}^{\times }:=\{x\in F:|x|=1\}$ . Denote by $({\mathcal{O}}_{F}^{\times })^{2}$ the subgroup of ${\mathcal{O}}_{F}^{\times }$ defined by

$$\begin{eqnarray}\displaystyle ({\mathcal{O}}_{F}^{\times })^{2}:=\{x\in {\mathcal{O}}_{F}^{\times }:\text{there exists }a\in F\text{ such that }x=a^{2}\}. & & \displaystyle \nonumber\end{eqnarray}$$

If $F$ is non-dyadic, then the quotient ${\mathcal{O}}_{F}^{\times }/({\mathcal{O}}_{F}^{\times })^{2}$ has two elements. Throughout the paper, we fix a non-square unit $\unicode[STIX]{x1D700}\in {\mathcal{O}}_{F}^{\times }\,\setminus \,({\mathcal{O}}_{F}^{\times })^{2}$ .

We now explicitly describe the parametrization of ergodic $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measures on $\operatorname{Sym}(\mathbb{N},F)$ .

Recall the definition (2) of the set $\unicode[STIX]{x1D6E5}$ of non-increasing sequences in $\mathbb{Z}\cup \{-\infty \}$ . A sequence $(k_{j})_{j\in \mathbb{N}}\in \unicode[STIX]{x1D6E5}$ is called finite if $k_{j}=-\infty$ for all sufficiently large $j$ . In this case, either $k_{1}=-\infty$ , and then we identity the sequence with an empty sequence, or $j_{0}:=\max \{j\mid k_{j}\in \mathbb{Z}\}\in \mathbb{N}$ , and then we identify $(k_{j})_{j\in \mathbb{N}}$ with $(k_{j})_{j=1}^{j_{0}}$ and $j_{0}$ is called the length of the sequence. Conversely, for any finite non-increasing sequence $(k_{j})_{j=1}^{n}$ in $\mathbb{Z}$ , we identify it with the element in $\unicode[STIX]{x1D6E5}$ by adding infinitely many $-\infty$ at the end of $(k_{j})_{j=1}^{n}$ .

For any $k\in \mathbb{Z}$ , let $\mathbb{Z}_{{>}k}$ denote the set of integers strictly larger than $k$ . We introduce the following four subsets of $\unicode[STIX]{x1D6E5}$ :

  1. $\unicode[STIX]{x1D6E5}[k]$ , the set of non-increasing sequences of finite length in $\mathbb{Z}_{{>}k}$ ;

  2. $\unicode[STIX]{x1D6E5}^{\sharp }[k]$ , the set of strictly decreasing sequences in $\mathbb{Z}_{{>}k}$ (which are automatically of finite length);

  3. $\unicode[STIX]{x1D6E5}[-\infty ]$ , the set of non-increasing sequences in $\mathbb{Z}$ of finite length or of infinite length tending to $-\infty$ ;

  4. $\unicode[STIX]{x1D6E5}^{\sharp }[-\infty ]$ , the set of strictly decreasing sequences of finite or infinite length in $\mathbb{Z}$ .

Note that for any $k\in \mathbb{Z}\cup \{-\infty \}$ , the following relations hold:

$$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D6E5}^{\sharp }[k]\subset \unicode[STIX]{x1D6E5}[k],\quad \unicode[STIX]{x1D6E5}^{\sharp }[k]\subset \unicode[STIX]{x1D6E5}^{\sharp }[-\infty ]\quad \text{and}\quad \unicode[STIX]{x1D6E5}^{\sharp }[k]\subset \unicode[STIX]{x1D6E5}[-\infty ]. & & \displaystyle \nonumber\end{eqnarray}$$

We introduce the parameter space

(3) $$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D6FA}:=\bigsqcup _{k\in \mathbb{Z}\cup \{-\infty \}}\{k\}\times \unicode[STIX]{x1D6E5}[k]\times \unicode[STIX]{x1D6E5}^{\sharp }[k], & & \displaystyle\end{eqnarray}$$

where $\{k\}$ is the singleton with a single element $k$ . The space $\unicode[STIX]{x1D6FA}$ is equipped with the topology induced by the inclusion:

$$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D6FA}\subset (\mathbb{Z}\cup \{-\infty \})\times (\mathbb{Z}\cup \{-\infty \})^{\mathbb{N}}\times (\mathbb{Z}\cup \{-\infty \})^{\mathbb{N}}. & & \displaystyle \nonumber\end{eqnarray}$$

To each element $h\in \unicode[STIX]{x1D6FA}$ , we assign an ergodic $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measure on $\operatorname{Sym}(\mathbb{N},F)$ as follows. Let

$$\begin{eqnarray}\displaystyle X_{i}^{(n)},\quad Y_{i}^{(n)},\quad H_{ij},\quad i\leqslant j,n=1,2,\ldots & & \displaystyle \nonumber\end{eqnarray}$$

be independent random variables uniformly distributed on ${\mathcal{O}}_{F}$ . In particular, define

(4) $$\begin{eqnarray}\displaystyle H=[H_{ij}]_{i,j\in \mathbb{N}} & & \displaystyle\end{eqnarray}$$

by setting $H_{ij}=H_{ji}$ if $i>j$ , then $H$ is an infinite symmetric random matrix sampled uniformly from $\operatorname{Sym}(\mathbb{N},{\mathcal{O}}_{F})$ .

Definition 1.5. For any $h\in \unicode[STIX]{x1D6FA}$ given by

$$\begin{eqnarray}\displaystyle h=(k;\Bbbk ,\Bbbk ^{\prime }),\quad \text{with }k\in \mathbb{Z}\cup \{-\infty \},~\Bbbk \in \unicode[STIX]{x1D6E5}[k],~\Bbbk ^{\prime }\in \unicode[STIX]{x1D6E5}^{\sharp }[k], & & \displaystyle \nonumber\end{eqnarray}$$

we define

$$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D708}_{h}:={\mathcal{L}}(S_{h}), & & \displaystyle \nonumber\end{eqnarray}$$

as the probability distribution of the infinite symmetric random matrix $S_{h}$ defined as follows. First let

$$\begin{eqnarray}\displaystyle W_{\Bbbk }:=\mathop{\biggl[\mathop{\sum }_{n=1}^{\infty }\unicode[STIX]{x1D71B}^{-k_{n}}X_{i}^{(n)}X_{j}^{(n)}\biggr]}\nolimits_{i,j\in \mathbb{N}},\quad W_{\Bbbk ^{\prime }}:=\mathop{\biggl[\mathop{\sum }_{n=1}^{\infty }\unicode[STIX]{x1D71B}^{-k_{n}^{\prime }}Y_{i}^{(n)}Y_{j}^{(n)}\biggr]}\nolimits_{i,j\in \mathbb{N}}, & & \displaystyle \nonumber\end{eqnarray}$$

and then set

$$\begin{eqnarray}\displaystyle S_{h}:=W_{\Bbbk }+\unicode[STIX]{x1D700}W_{\Bbbk ^{\prime }}+\unicode[STIX]{x1D71B}^{-k}H. & & \displaystyle \nonumber\end{eqnarray}$$

Remark 1.6. For the element $h=(-\infty ;\emptyset ,\emptyset )$ , where $\emptyset$ means the empty sequence in $\mathbb{Z}$ , the corresponding matrix $S_{h}$ defined in Definition 1.5 is the zero matrix $O\in \operatorname{Sym}(\mathbb{N},F)$ and hence the measure $\unicode[STIX]{x1D708}_{h}$ is the Dirac measure at the point $O\in \operatorname{Sym}(\mathbb{N},F)$ .

Remark 1.7. The strictly decreasing assumption on the sequences $\Bbbk ^{\prime }\in \unicode[STIX]{x1D6E5}^{\sharp }[k]$ is imposed for the uniqueness of parametrization. The reason is the following:

(5) $$\begin{eqnarray}\displaystyle {\mathcal{L}}\biggl(\mathop{\biggl[\mathop{\sum }_{n=1}^{2}X_{i}^{(n)}X_{j}^{(n)}\biggr]}\nolimits_{i,j\in \mathbb{N}}\biggr)={\mathcal{L}}\biggl(\unicode[STIX]{x1D700}\mathop{\biggl[\mathop{\sum }_{n=1}^{2}X_{i}^{(n)}X_{j}^{(n)}\biggr]}\nolimits_{i,j\in \mathbb{N}}\biggr). & & \displaystyle\end{eqnarray}$$

For the detail, see Remark 4.8 and the proof of Proposition 5.3 below.

Let ${\mathcal{P}}_{\text{erg}}(\operatorname{Sym}(\mathbb{N},F))$ be the space of ergodic $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measures on $\operatorname{Sym}(\mathbb{N},F)$ , endowed with the induced weak topology. The classification of ${\mathcal{P}}_{\text{erg}}(\operatorname{Sym}(\mathbb{N},F))$ is given by the following.

Theorem 1.8. Assume that $F$ is non-dyadic. Then the map $h\mapsto \unicode[STIX]{x1D708}_{h}$ is a homeomorphism between $\unicode[STIX]{x1D6FA}$ and ${\mathcal{P}}_{\text{erg}}(\operatorname{Sym}(\mathbb{N},F))$ .

Remark 1.9. By Theorem 1.8, the space ${\mathcal{P}}_{\text{erg}}(\operatorname{Sym}(\mathbb{N},F))$ is weakly closed in the space of all Borel measures on $\operatorname{Sym}(\mathbb{N},F)$ and is $\unicode[STIX]{x1D70E}$ -compact. Moreover, any measure $\unicode[STIX]{x1D708}_{h}\in {\mathcal{P}}_{\text{erg}}(\operatorname{Sym}(\mathbb{N},F))$ is compactly supported.

Theorem 1.8 can be explained as follows. We have the following elementary ergodic measures.

  1. (Haar type measures) For any $k\in \mathbb{Z}$ , the normalized Haar measure on $\operatorname{Sym}(\mathbb{N},\unicode[STIX]{x1D71B}^{-k}{\mathcal{O}}_{F})$ is $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -ergodic.

  2. (Symmetric Wishart type measures) Let $X_{1},X_{2},\ldots$ be independent copies of $F$ -valued random variables, all of which are uniformly distributed on ${\mathcal{O}}_{F}$ . For any $k\in \mathbb{Z}$ , the distributions of the infinite rank-one random matrices

    $$\begin{eqnarray}[\unicode[STIX]{x1D71B}^{-k}X_{i}X_{j}]_{i,j\in \mathbb{N}}\quad \text{and}\quad [\unicode[STIX]{x1D700}\unicode[STIX]{x1D71B}^{-k}X_{i}X_{j}]_{i,j\in \mathbb{N}}\end{eqnarray}$$
    are $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -ergodic.

Theorem 1.8 implies that any ergodic $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measure on $\operatorname{Sym}(\mathbb{N},F)$ can be obtained as a possibly infinite convolution of the above two types of elementary ones.

1.3 Characteristic functions of ergodic measures

Let $\unicode[STIX]{x1D712}\in \widehat{F}$ be a fixed character of $F$ such that

(6) $$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D712}|_{{\mathcal{O}}_{F}}\equiv 1\quad \text{and}\quad \unicode[STIX]{x1D712}\text{ is not constant on }\unicode[STIX]{x1D71B}^{-1}{\mathcal{O}}_{F}. & & \displaystyle\end{eqnarray}$$

Given a Borel probability measure $\unicode[STIX]{x1D707}$ on $\operatorname{Mat}(\mathbb{N},F)$ , its characteristic function, or Fourier transform, $\widehat{\unicode[STIX]{x1D707}}$ is defined on $\operatorname{Mat}(\infty ,F)$ by the formula

$$\begin{eqnarray}\displaystyle \widehat{\unicode[STIX]{x1D707}}(A):=\int _{\operatorname{Mat}(\mathbb{N},F)}\unicode[STIX]{x1D712}(\operatorname{tr}(AM))\,\unicode[STIX]{x1D707}(dM),\quad A\in \operatorname{Mat}(\infty ,F). & & \displaystyle \nonumber\end{eqnarray}$$

Similarly, given a Borel probability measure $\unicode[STIX]{x1D708}$ on $\operatorname{Sym}(\mathbb{N},F)$ , its characteristic function $\widehat{\unicode[STIX]{x1D708}}$ is defined on $\operatorname{Sym}(\infty ,F)$ by the formula

$$\begin{eqnarray}\displaystyle \widehat{\unicode[STIX]{x1D708}}(A):=\int _{\operatorname{Sym}(\mathbb{N},F)}\unicode[STIX]{x1D712}(\operatorname{tr}(AS))\,\unicode[STIX]{x1D708}(dS),\quad A\in \operatorname{Sym}(\infty ,F). & & \displaystyle \nonumber\end{eqnarray}$$

Let $\operatorname{Mat}(n,F)$ be the space of $n\times n$ matrices with entries in $F$ . Every $A\in \operatorname{Mat}(n,F)$ can be written (see Lemma 2.4 below) in the form

$$\begin{eqnarray}\displaystyle A=a\cdot \operatorname{diag}(\unicode[STIX]{x1D71B}^{-k_{1}},\ldots ,\unicode[STIX]{x1D71B}^{-k_{n}})\cdot b,\quad a,b\in \operatorname{GL}(n,{\mathcal{O}}_{F}),k_{i}\in \mathbb{Z}\cup \{-\infty \}. & & \displaystyle \nonumber\end{eqnarray}$$

These numbers $k_{1},k_{2},\ldots ,k_{n}$ , taken with multiplicities, are uniquely determined by $A$ and are called the singular numbers of $A$ . The collection of the singular numbers of the matrix $A$ is denoted by $\text{Sing}(A)$ .

After the computation of characteristic functions for the probability measures in the list of measures defined in Definition 1.1 (see Proposition 4.1 below), Theorem 1.3 can be reformulated in the following form.

Theorem 1.10. The characteristic functions of ergodic $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measures on $\operatorname{Mat}(\mathbb{N},F)$ are exactly of the form

$$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D711}(A)=\mathop{\prod }_{\ell \in \text{Sing}(A)}\exp \biggl(-\text{log}\,q\cdot \mathop{\sum }_{j=1}^{\infty }(k_{j}+\ell )\unicode[STIX]{x1D7D9}_{\{k_{j}+\ell \geqslant 1\}}\biggr),\quad A\in \operatorname{Mat}(\infty ,F), & & \displaystyle \nonumber\end{eqnarray}$$

where $\Bbbk =(k_{j})_{j\in \mathbb{N}}\in \unicode[STIX]{x1D6E5}$ is the parameter sequence.

For formulating a similar statement in the symmetric case, we need to introduce a function $\unicode[STIX]{x1D703}:F\rightarrow \mathbb{C}$ by

(7) $$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D703}(x):=\int _{{\mathcal{O}}_{F}}\unicode[STIX]{x1D712}(z^{2}\cdot x)\,dz. & & \displaystyle\end{eqnarray}$$

Properties of the function $\unicode[STIX]{x1D703}$ are summarized in Proposition 4.5 below.

Let $\operatorname{Sym}(n,F)$ be the space of $n\times n$ symmetric matrices with entries in $F$ . Note that if the field $F$ is non-dyadic, then any $A\in \operatorname{Sym}(n,F)$ can be written (see Lemma 2.7 below) in the form

$$\begin{eqnarray}\displaystyle A=g\cdot \operatorname{diag}(x_{1},\ldots ,x_{n})\cdot g^{t},\quad g\in \operatorname{GL}(n,{\mathcal{O}}_{F}). & & \displaystyle \nonumber\end{eqnarray}$$

After the computation of characteristic functions for the probability measures in the list of measures defined in Definition 1.5 (see Proposition 4.4), Theorem 1.8 can be reformulated in the following form.

Theorem 1.11. Assume that $F$ is non-dyadic. Then the characteristic functions of the ergodic $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measures on $\operatorname{Sym}(\mathbb{N},F)$ are exactly given by

$$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D6F7}(\operatorname{diag}(x_{1},\ldots ,x_{m},0,\ldots ))=\mathop{\prod }_{i=1}^{m}\biggl[\unicode[STIX]{x1D7D9}_{{\mathcal{O}}_{F}}(\unicode[STIX]{x1D71B}^{-k}x_{i})\mathop{\prod }_{j=1}^{\infty }\unicode[STIX]{x1D703}(\unicode[STIX]{x1D71B}^{-k_{j}}x_{i})\mathop{\prod }_{j=1}^{\infty }\unicode[STIX]{x1D703}(\unicode[STIX]{x1D700}\unicode[STIX]{x1D71B}^{-k_{j}^{\prime }}x_{i})\biggr], & & \displaystyle \nonumber\end{eqnarray}$$

where $h=(k;(k_{j})_{j\in \mathbb{N}},(k_{j}^{\prime })_{j\in \mathbb{N}})$ is a parameter in $\unicode[STIX]{x1D6FA}$ introduced in (3).

1.4 Spherical representations

Our classification theorems, Theorems 1.3 and 1.8, can equivalently be formulated as a classification of spherical representations of analogues, in our context, of the infinite dimensional Cartan motion groups

$$\begin{eqnarray}\displaystyle \operatorname{Mat}(\infty ,F)\rtimes (\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F}))\quad \text{and}\quad \operatorname{Sym}(\infty ,F)\rtimes \operatorname{GL}(\infty ,{\mathcal{O}}_{F}) & & \displaystyle \nonumber\end{eqnarray}$$

respectively. We explain this in more detail for $\operatorname{Sym}(\infty ,F)\rtimes \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ .

Recall that $\operatorname{Sym}(n,F)\rtimes \operatorname{GL}(n,{\mathcal{O}}_{F})$ is the semi-direct product of the additive group $\operatorname{Sym}(n,F)$ and the general linear group $\operatorname{GL}(n,{\mathcal{O}}_{F})$ . Elements of $\operatorname{Sym}(n,F)\rtimes \operatorname{GL}(n,{\mathcal{O}}_{F})$ are pairs $(A,g),A\in \operatorname{Sym}(n,F),g\in \operatorname{GL}(n,{\mathcal{O}}_{F})$ and the rule of multiplication is given by

$$\begin{eqnarray}(A,g)\cdot (B,h)=(A+gBg^{t},gh).\end{eqnarray}$$

The group $\operatorname{Sym}(\infty ,F)\rtimes \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ is defined in a similar way and is of course the inductive limit of the sequence $\operatorname{Sym}(n,F)\rtimes \operatorname{GL}(n,{\mathcal{O}}_{F})$ . The groups $\operatorname{Sym}(\infty ,F)$ and $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ are canonically identified with subgroups of $\operatorname{Sym}(\infty ,F)\rtimes \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ by the following embeddings

$$\begin{eqnarray}A\mapsto (A,1)\quad \text{and}\quad g\mapsto (0,g),\end{eqnarray}$$

where $A\in \operatorname{Sym}(\infty ,F)$ and $g\in \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ .

A unitary representation $\unicode[STIX]{x1D70C}$ of $\operatorname{Sym}(\infty ,F)\rtimes \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ in a Hilbert space $H(\unicode[STIX]{x1D70C})$ is called spherical if it is irreducible and the subspace $H(\unicode[STIX]{x1D70C})^{\operatorname{GL}(\infty ,{\mathcal{O}}_{F})}$ of $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant vectors in $H(\unicode[STIX]{x1D70C})$ is non-trivial; in which case, by irreducibility, the subspace $H(\unicode[STIX]{x1D70C})^{\operatorname{GL}(\infty ,{\mathcal{O}}_{F})}$ has dimension one. A vector $h\in H(\unicode[STIX]{x1D70C})^{\operatorname{GL}(\infty ,{\mathcal{O}}_{F})}$ of norm 1 is called a spherical vector of $\unicode[STIX]{x1D70C}$ and the function

$$\begin{eqnarray}\unicode[STIX]{x1D711}_{\unicode[STIX]{x1D70C}}(g):=(\unicode[STIX]{x1D70C}(g)h,h),\quad g\in \operatorname{GL}(\infty ,{\mathcal{O}}_{F})\end{eqnarray}$$

is called the spherical function of $\unicode[STIX]{x1D70C}$ . The spherical function $\unicode[STIX]{x1D711}_{\unicode[STIX]{x1D70C}}$ is an invariant of the spherical representation $\unicode[STIX]{x1D70C}$ and it uniquely determines $\unicode[STIX]{x1D70C}$ . As a bi-invariant function with respect to the subgroup $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\subset \operatorname{Sym}(\infty ,F)\rtimes \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ , the function $\unicode[STIX]{x1D711}_{\unicode[STIX]{x1D70C}}$ is uniquely determined by its restriction $\unicode[STIX]{x1D711}_{\unicode[STIX]{x1D70C}}|_{\operatorname{Sym}(\infty ,F)}$ .

Given an ergodic $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measure $\unicode[STIX]{x1D708}$ on $\operatorname{Sym}(\mathbb{N},F)$ , one may define a spherical representation $\unicode[STIX]{x1D70C}_{\unicode[STIX]{x1D708}}$ in the Hilbert space $L^{2}(\operatorname{Sym}(\mathbb{N},F),\unicode[STIX]{x1D708})$ as follows:

$$\begin{eqnarray}\displaystyle & (\unicode[STIX]{x1D70C}(g)\unicode[STIX]{x1D709})(S)=\unicode[STIX]{x1D709}(g^{-1}S),\quad g\in \operatorname{GL}(\infty ,{\mathcal{O}}_{F}), & \displaystyle \nonumber\\ \displaystyle & (\unicode[STIX]{x1D70C}(A)\unicode[STIX]{x1D709})(S)=\unicode[STIX]{x1D712}(\operatorname{tr}(AS))\unicode[STIX]{x1D709}(S),\quad A\in \operatorname{Sym}(\infty ,F), & \displaystyle \nonumber\end{eqnarray}$$

where $\unicode[STIX]{x1D709}\in L^{2}(\operatorname{Sym}(\mathbb{N},F),\unicode[STIX]{x1D708})$ and $S\in \operatorname{Sym}(\mathbb{N},F)$ . The spherical vector can be chosen as the constant function $\unicode[STIX]{x1D709}_{0}(S)\equiv 1$ .

Proposition 1.12. The map $\unicode[STIX]{x1D708}\mapsto \unicode[STIX]{x1D70C}_{\unicode[STIX]{x1D708}}$ defines a bijection between the set of ergodic $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measures on $\operatorname{Sym}(\mathbb{N},F)$ and the set of spherical representations of the group $\operatorname{Sym}(\infty ,F)\rtimes \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ .

The proof of Proposition 1.12 is the same as that of Olshanski and Vershik [OV96, Proposition 1.5].

1.5 An outline of the argument

Our argument relies on the Vershik–Kerov ergodic method in the spirit of Olshanski and Vershik [OV96]. The implementation of individual steps is, however, quite different. In the case of measures on $\operatorname{Mat}(\mathbb{N},F)$ and the case of measures on $\operatorname{Sym}(\mathbb{N},F)$ , the main steps are:

  1. explicit construction of ergodic measures, see Definitions 1.1 and 1.5;

  2. the asymptotic formulae for the analogues of Harish-Chandra–Izykson–Zuber orbital integrals, see Theorem 7.1, Theorem 7.4 in the case of measures on $\operatorname{Mat}(\mathbb{N},F)$ and Theorem 7.5, Theorem 7.6 in the case of measures on $\operatorname{Sym}(\mathbb{N},F)$ ;

  3. proof of completeness of the lists of ergodic measures, see Theorems 8.8 and 8.14 respectively.

We now explain our method in greater detail in the case of measures on $\operatorname{Sym}(\mathbb{N},F)$ .

(1) The Vershik–Kerov method: approximation of ergodic measures by orbital measures. While we follow the general scheme of Vershik and Kerov, a number of details are different.

Given $x\in \operatorname{Sym}(\mathbb{N},F)$ and $n\in \mathbb{N}$ , let $m_{\operatorname{GL}(n,{\mathcal{O}}_{F})}(x)$ denote the unique $\operatorname{GL}(n,{\mathcal{O}}_{F})$ -invariant probability measure on $\operatorname{Sym}(\mathbb{N},F)$ supported on the orbit $\operatorname{GL}(n,{\mathcal{O}}_{F})\cdot x\subset \operatorname{Sym}(\mathbb{N},F)$ . Let $\mathscr{ORB}_{\infty }(\operatorname{Sym}(\mathbb{N},F))$ be the class of probability measures $\unicode[STIX]{x1D708}$ on $\operatorname{Sym}(\mathbb{N},F)$ verifying the condition: there exists a sequence of positive integers $n_{1}<n_{2}<\cdots \,$ and a sequence $(\unicode[STIX]{x1D708}_{n_{k}})_{k\in \mathbb{N}}$ of probability measures with $\unicode[STIX]{x1D708}_{n_{k}}$ being a $\operatorname{GL}(n_{k},{\mathcal{O}}_{F})$ -orbital measure supported on $\operatorname{Sym}(n_{k},F)\subset \operatorname{Sym}(\mathbb{N},F)$ , such that $\unicode[STIX]{x1D708}_{n_{k}}$ converges weakly to $\unicode[STIX]{x1D708}$ .

As a variant of Vershik’s Theorem (see Theorem 6.3 below), we obtain the following inclusion:

$$\begin{eqnarray}\displaystyle {\mathcal{P}}_{\text{erg}}(\operatorname{Sym}(\mathbb{N},F))\subset \mathscr{ORB}_{\infty }(\operatorname{Sym}(\mathbb{N},F)). & & \displaystyle \nonumber\end{eqnarray}$$

Note that, a priori, we do not know whether the inverse inclusion holds.

(2) Main ingredients: Classification of $\mathscr{ORB}_{\infty }(\operatorname{Sym}(\mathbb{N},F))$ .

(i) Computation of orbital integrals. To describe $\mathscr{ORB}_{\infty }(\operatorname{Sym}(\mathbb{N},F))$ , we need to understand the asymptotic behaviour of the characteristic functions of orbital measures of the compact groups $\operatorname{GL}(n,{\mathcal{O}}_{F})$ . Recalling the assumption on the character $\unicode[STIX]{x1D712}\in \widehat{F}$ in (6), we obtain an asymptotic formula for the following orbital integral:

(8) $$\begin{eqnarray}\displaystyle \int _{\operatorname{GL}(n,{\mathcal{O}}_{F})}\unicode[STIX]{x1D712}(\operatorname{tr}(g\cdot \operatorname{diag}(x_{1},\ldots ,x_{n})\cdot g^{t}\cdot \operatorname{diag}(a_{1},\ldots ,a_{r},0,\ldots )))\,dg, & & \displaystyle\end{eqnarray}$$

where $dg$ is the normalized Haar measure of $\operatorname{GL}(n,{\mathcal{O}}_{F})$ . The formula we obtain for the integral (8) is uniformly asymptotically multiplicative, that is, the orbital integral (8) has the same asymptotic behaviour, uniformly on the choices of $x_{1},\ldots ,x_{n}$ , as the following product of much simpler orbital integrals:

$$\begin{eqnarray}\displaystyle \mathop{\prod }_{j=1}^{r}\int _{\operatorname{GL}(n,{\mathcal{O}}_{F})}\unicode[STIX]{x1D712}(\operatorname{tr}(g\cdot \operatorname{diag}(x_{1},\ldots ,x_{n})\cdot g^{t}\cdot \operatorname{diag}(a_{j},0,0,\ldots )))\,dg. & & \displaystyle \nonumber\end{eqnarray}$$

See Theorem 7.6 for the details.

Explicit computation of the above orbital integral requires some Fourier analysis on the field $F$ and quite a few combinatorial arguments in which we compute the cardinality of various sets of matrices over the finite field  $\mathbf{F}_{q}$ .

(ii) Multiplicativity of characteristic functions for limits of orbital measures. An immediate consequence of the uniform asymptotic multiplicativity for the orbital integral (8) is that for any $\unicode[STIX]{x1D708}\in \mathscr{ORB}_{\infty }(\operatorname{Sym}(\mathbb{N},F))$ , its characteristic function $\widehat{\unicode[STIX]{x1D708}}$ possesses an exact multiplicativity property, that is, for any $r\in \mathbb{N}$ and any $x_{1},\ldots ,x_{r}\in F$ ,

(9) $$\begin{eqnarray}\displaystyle \widehat{\unicode[STIX]{x1D708}}(\operatorname{diag}(x_{1},\ldots ,x_{r},0,0,\ldots ))=\mathop{\prod }_{j=1}^{r}\widehat{\unicode[STIX]{x1D708}}(x_{j}e_{11}), & & \displaystyle\end{eqnarray}$$

where $e_{11}$ is the elementary matrix whose $(1,1)$ -coefficient is $1$ . This multiplicativity result implies in particular that the classification of the class $\mathscr{ORB}_{\infty }(\operatorname{Sym}(\mathbb{N},F))$ is reduced to the classification of the class of functions on $F$ defined by $x\mapsto \widehat{\unicode[STIX]{x1D708}}(xe_{11})$ .

(3) Ergodicity: Proof of the inclusion:

$$\begin{eqnarray}\displaystyle \mathscr{ORB}_{\infty }(\operatorname{Sym}(\mathbb{N},F))\subset {\mathcal{P}}_{\text{erg}}(\operatorname{Sym}(\mathbb{N},F)). & & \displaystyle \nonumber\end{eqnarray}$$

Our direct proof of ergodicity for all measures in $\mathscr{ORB}_{\infty }(\operatorname{Sym}(\mathbb{N},F))$ uses an argument of Okounkov and Olshanski [OO98]: ergodicity for measures in $\mathscr{ORB}_{\infty }(\operatorname{Sym}(\mathbb{N},F))$ is derived from the De Finetti Theorem, see Theorem 3.10 below for the details. This approach of proving ergodicity can be also applied to different situations, such as that of Olshanski and Vershik in [OV96].

(4) Proof of the equality

(10) $$\begin{eqnarray}\displaystyle {\mathcal{P}}_{\text{erg}}(\operatorname{Sym}(\mathbb{N},F))=\{\unicode[STIX]{x1D708}_{h}:h\in \unicode[STIX]{x1D6FA}\}. & & \displaystyle\end{eqnarray}$$

By comparing the characteristic functions $\widehat{\unicode[STIX]{x1D708}_{h}}$ for all the measures $\unicode[STIX]{x1D708}_{h}$ with $h\in \unicode[STIX]{x1D6FA}$ and that of measures in $\mathscr{ORB}_{\infty }(\operatorname{Sym}(\mathbb{N},F))$ , we obtain the equality

$$\begin{eqnarray}\displaystyle \mathscr{ORB}_{\infty }(\operatorname{Sym}(\mathbb{N},F))=\{\unicode[STIX]{x1D708}_{h}:h\in \unicode[STIX]{x1D6FA}\}. & & \displaystyle \nonumber\end{eqnarray}$$

Combining this equality with the results obtained in the previous steps, we finally get the desired equality (10).

1.6 Organization of the paper

The exposition, which we tried to make essentially self-contained, is organized as follows.

In § 2, we recall the definition of ergodic measures and the necessary definitions related to non-discrete locally compact non-Archimedean fields, linear groups over them and Fourier transforms in this setting.

In § 3, we prove that all the measures on $\operatorname{Mat}(\mathbb{N},F)$ from the family $\{\unicode[STIX]{x1D707}_{\Bbbk }\mid \Bbbk \in \unicode[STIX]{x1D6E5}\}$ introduced in Definition 1.1 are $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant and ergodic and that all the measures on $\operatorname{Sym}(\mathbb{N},F)$ from the family $\{\unicode[STIX]{x1D708}_{h}\mid h\in \unicode[STIX]{x1D6FA}\}$ introduced in Definition 1.5 are $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant and ergodic.

In § 4, we give explicit formulae for characteristic functions of measures from the two families $\{\unicode[STIX]{x1D707}_{\Bbbk }:\Bbbk \in \unicode[STIX]{x1D6E5}\}$ and $\{\unicode[STIX]{x1D708}_{h}:h\in \unicode[STIX]{x1D6FA}\}$ .

In § 5, we prove that the parametrization maps $\Bbbk \mapsto \unicode[STIX]{x1D707}_{\Bbbk }$ from $\unicode[STIX]{x1D6E5}$ to ${\mathcal{P}}_{\text{erg}}(\operatorname{Mat}(\mathbb{N},F))$ and $h\mapsto \unicode[STIX]{x1D708}_{h}$ from $\unicode[STIX]{x1D6FA}$ to ${\mathcal{P}}_{\text{erg}}(\operatorname{Sym}(\mathbb{N},F))$ are injective.

In § 6, we introduce orbital measures and recall the Vershik–Kerov ergodic method for dealing with ergodic measures for inductively compact groups.

In § 7, we obtain the asymptotic formula for orbital integrals of the type (8).

In § 8, we complete the classifications by proving that the parametrization maps $\Bbbk \mapsto \unicode[STIX]{x1D707}_{\Bbbk }$ and $h\mapsto \unicode[STIX]{x1D708}_{h}$ are surjective.

In § 9, we show that the parametrization maps $\Bbbk \mapsto \unicode[STIX]{x1D707}_{\Bbbk }$ and $h\mapsto \unicode[STIX]{x1D708}_{h}$ are homeomorphisms between corresponding topological spaces.

Proofs of some routine technical lemmata are given in Appendix A.

2 Preliminaries

2.1 Ergodic measures

2.1.1 Ergodicity

Let ${\mathcal{X}}$ be a Polish space, that is, it is homeomorphic to a complete metric space that has a countable dense subset. Denote by ${\mathcal{P}}({\mathcal{X}})$ the set of Borel probability measures on ${\mathcal{X}}$ . Denote by $C_{b}({\mathcal{X}})$ the space of bounded continuous complex-valued functions on ${\mathcal{X}}$ . Recall that a sequence $(\unicode[STIX]{x1D707}_{n})_{n\in \mathbb{N}}$ in ${\mathcal{P}}({\mathcal{X}})$ is said to converge weakly to $\unicode[STIX]{x1D707}\in {\mathcal{P}}({\mathcal{X}})$ and is denoted by $\unicode[STIX]{x1D707}_{n}\Longrightarrow \unicode[STIX]{x1D707}$ if for any $f\in C_{b}({\mathcal{X}})$ , we have

$$\begin{eqnarray}\displaystyle \lim _{n\rightarrow \infty }\int _{{\mathcal{X}}}f(x)\,\unicode[STIX]{x1D707}_{n}(dx)=\int _{{\mathcal{X}}}f(x)\,\unicode[STIX]{x1D707}(dx). & & \displaystyle \nonumber\end{eqnarray}$$

Given a group action of a group $G$ on ${\mathcal{X}}$ , we denote by ${\mathcal{P}}_{\text{inv}}^{G}({\mathcal{X}})$ the set of $G$ -invariant Borel probability measures on ${\mathcal{X}}$ . By definition, a $G$ -invariant Borel probability measure $\unicode[STIX]{x1D707}\in {\mathcal{P}}_{\text{inv}}^{G}({\mathcal{X}})$ is ergodic, if for any $G$ -invariant Borel subset ${\mathcal{A}}\subset {\mathcal{X}}$ , either $\unicode[STIX]{x1D707}({\mathcal{A}})=0$ or $\unicode[STIX]{x1D707}({\mathcal{X}}\,\setminus \,{\mathcal{A}})=0$ . The totality of ergodic $G$ -invariant probability measures on ${\mathcal{X}}$ is denoted by ${\mathcal{P}}_{\text{erg}}^{G}({\mathcal{X}})$ . If the group action is clear from the context, we denote ${\mathcal{P}}_{\text{inv}}^{G}({\mathcal{X}})$ and ${\mathcal{P}}_{\text{erg}}^{G}({\mathcal{X}})$ simply by ${\mathcal{P}}_{\text{inv}}({\mathcal{X}})$ and ${\mathcal{P}}_{\text{erg}}({\mathcal{X}})$ respectively.

2.1.2 Indecomposability and ergodicity

Using the notation as before, a measure $\unicode[STIX]{x1D707}\in {\mathcal{P}}_{\text{inv}}^{G}({\mathcal{X}})$ is called indecomposable in ${\mathcal{P}}_{\text{inv}}^{G}({\mathcal{X}})$ if the equality $\unicode[STIX]{x1D707}=\unicode[STIX]{x1D6FC}\unicode[STIX]{x1D707}_{1}+(1-\unicode[STIX]{x1D6FC})\unicode[STIX]{x1D707}_{2}$ with $\unicode[STIX]{x1D6FC}\in (0,1)$ , $\unicode[STIX]{x1D707}_{1},\unicode[STIX]{x1D707}_{2}\in {\mathcal{P}}_{\text{inv}}^{G}({\mathcal{X}})$ , implies $\unicode[STIX]{x1D707}=\unicode[STIX]{x1D707}_{1}=\unicode[STIX]{x1D707}_{2}$ . Recall that a Borel subset ${\mathcal{A}}\subset {\mathcal{X}}$ is called almost $G$ -invariant with respect to a measure $\unicode[STIX]{x1D707}\in {\mathcal{P}}_{\text{inv}}^{G}({\mathcal{X}})$ if for every $g\in G$ , we have $\unicode[STIX]{x1D707}({\mathcal{A}}\unicode[STIX]{x1D6E5}g.{\mathcal{A}})=0$ , where $g.{\mathcal{A}}=\{g.x:x\in {\mathcal{A}}\}$ and ${\mathcal{A}}\unicode[STIX]{x1D6E5}g.{\mathcal{A}}=({\mathcal{A}}\,\setminus \,g.{\mathcal{A}})\cup (g.{\mathcal{A}}\,\setminus \,{\mathcal{A}})$ .

A Borel probability measure $\unicode[STIX]{x1D707}\in {\mathcal{P}}_{\text{inv}}^{G}({\mathcal{X}})$ is indecomposable in ${\mathcal{P}}_{\text{inv}}^{G}({\mathcal{X}})$ if and only if any Borel subset ${\mathcal{A}}\subset {\mathcal{X}}$ which is almost $G$ -invariant with respect to the measure $\unicode[STIX]{x1D707}$ satisfies either $\unicode[STIX]{x1D707}({\mathcal{A}})=0$ or $\unicode[STIX]{x1D707}({\mathcal{X}}\,\setminus \,{\mathcal{A}})=0$ (see [Bog69, Fom50] and [Buf14, Proposition 1]).

Indecomposable Borel probability measures in ${\mathcal{P}}_{\text{inv}}^{G}({\mathcal{X}})$ are a fortiori ergodic. For actions of general groups, ergodic probability measures may fail to be indecomposable, see [Buf14, § 5] for a counterexample of Kolmogorov. Nevertheless, it is proved in [Buf14, Proposition 2] that for action of inductively compact groups, and thus including the actions in our main Theorems 1.3 and 1.8, etc, the notions of indecomposability and ergodicity coincide.

Note that for the action of countable groups, the notions of indecomposability and ergodicity coincide. Indeed, if $G$ is a countable group and $\unicode[STIX]{x1D707}\in {\mathcal{P}}_{\text{erg}}^{G}({\mathcal{X}})$ , then for any $G$ -almost invariant Borel subset ${\mathcal{A}}\subset {\mathcal{X}}$ with respect to the measure $\unicode[STIX]{x1D707}$ , the Borel subset $\bigcap _{g\in G}g.{\mathcal{A}}$ is $G$ -invariant and hence, by the ergodicity of $\unicode[STIX]{x1D707}$ , we have $\unicode[STIX]{x1D707}(\bigcap _{g\in G}g.{\mathcal{A}})\in \{0,1\}$ . Since ${\mathcal{A}}$ is $G$ -almost invariant with respect to $\unicode[STIX]{x1D707}$ and since $G$ is countable, we also have $\unicode[STIX]{x1D707}({\mathcal{A}}\unicode[STIX]{x1D6E5}\bigcap _{g\in G}g.{\mathcal{A}})=0$ . Therefore, $\unicode[STIX]{x1D707}({\mathcal{A}})\in \{0,1\}$ . This proves that $\unicode[STIX]{x1D707}$ is indecomposable.

2.2 Fields and integers

Let $F$ be a non-discrete locally compact non-Archimedean field. The classification of local fields (see, e.g., Ramakrishnan and Valenza’s book [RV99, Theorem 4-12]) implies that $F$ is isomorphic to one of the following fields:

  1. a finite extension of the field $\mathbb{Q}_{p}$ of $p$ -adic numbers for some prime $p$ ;

  2. the field of formal Laurent series over a finite field.

Let $|\cdot |$ be the absolute value on $F$ and denote by $d(\cdot ,\cdot )$ the ultrametric on $F$ defined by $d(x,y)=|x-y|$ . The ring of integers in $F$ is given by

$$\begin{eqnarray}\displaystyle {\mathcal{O}}_{F}:=\{x\in F:|x|\leqslant 1\}. & & \displaystyle \nonumber\end{eqnarray}$$

The subset $\mathfrak{m}:=\{x\in F:|x|<1\}$ is the unique maximal ideal of ${\mathcal{O}}_{F}$ . The ideal $\mathfrak{m}$ is principal. Any generator of $\mathfrak{m}$ is called a uniformizer of $F$ . Throughout the paper, we fix any uniformizer $\unicode[STIX]{x1D71B}$ of $F$ , that is, $\mathfrak{m}=\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}$ . The field ${\mathcal{O}}_{F}/\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}$ is finite with $q=p^{f}$ elements for a prime number $p$ and a positive integer $f\in \mathbb{N}$ . If $q=2^{f}$ , then we say that $F$ is dyadic, otherwise, we say that $F$ is non-dyadic.

We write $\mathbf{F}_{q}:={\mathcal{O}}_{F}/\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}$ . The quotient map is denoted by

$$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D70B}:{\mathcal{O}}_{F}\rightarrow \mathbf{F}_{q}={\mathcal{O}}_{F}/\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}. & & \displaystyle \nonumber\end{eqnarray}$$

Fix a complete set of representatives ${\mathcal{C}}_{q}\subset {\mathcal{O}}_{F}$ of cosets of $\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}$ in ${\mathcal{O}}_{F}$ and assume that $0\in {\mathcal{C}}_{q}$ . The restriction of the quotient map $\unicode[STIX]{x1D70B}$ on the finite set ${\mathcal{C}}_{q}$ is a bijection:

(11) $$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D70B}:{\mathcal{C}}_{q}\xrightarrow[{}]{\text{bijection}}\mathbf{F}_{q}. & & \displaystyle\end{eqnarray}$$

Any element of $F$ is uniquely expanded as a convergent series in $F$ :

(12) $$\begin{eqnarray}\displaystyle x=\mathop{\sum }_{n=v}^{\infty }a_{n}\unicode[STIX]{x1D71B}^{n}\quad (v\in \mathbf{Z},a_{n}\in {\mathcal{C}}_{q},a_{v}\neq 0). & & \displaystyle\end{eqnarray}$$

If $x\in F$ is given by the series (12), then we define the $F$ -valuation of $x$ by $\text{ord}_{F}(x):=v$ . By convention, we set $\text{ord}_{F}(0)=\infty$ . The absolute value and the $F$ -valuation of any element $x\in F$ are related by the formula $|x|=q^{-\text{ord}_{F}(x)}$ .

2.3 Group actions

Let $\operatorname{GL}(n,F)$ and $\operatorname{GL}(n,{\mathcal{O}}_{F})$ denote the groups of invertible $n\times n$ matrices over $F$ and ${\mathcal{O}}_{F}$ respectively. The group $\operatorname{GL}(n,{\mathcal{O}}_{F})$ is embedded naturally into $\operatorname{GL}(n+1,{\mathcal{O}}_{F})$ by

(13) $$\begin{eqnarray}\displaystyle a\in \operatorname{GL}(n,{\mathcal{O}}_{F})\mapsto \left(\begin{array}{@{}cc@{}}a & 0\\ 0 & 1\end{array}\right)\in \operatorname{GL}(n+1,{\mathcal{O}}_{F}). & & \displaystyle\end{eqnarray}$$

Define an inductive limit group

$$\begin{eqnarray}\displaystyle \operatorname{GL}(\infty ,{\mathcal{O}}_{F}):=\lim _{\longrightarrow }\operatorname{GL}(n,{\mathcal{O}}_{F}). & & \displaystyle \nonumber\end{eqnarray}$$

Equivalently, $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ is the group of infinite invertible matrices $g=(g_{ij})_{i,j\in \mathbb{N}}$ over ${\mathcal{O}}_{F}$ such that $g_{ij}=\unicode[STIX]{x1D6FF}_{ij}$ if $i+j$ is large enough.

Let $\operatorname{Mat}(n,F)$ and $\operatorname{Mat}(n,{\mathcal{O}}_{F})$ denote the spaces of all $n\times n$ matrices over $F$ and ${\mathcal{O}}_{F}$ respectively. Define

$$\begin{eqnarray}\operatorname{Mat}(\mathbb{N},F):=\{X=(X_{ij})_{i,j\in \mathbb{N}}\mid X_{ij}\in F\}.\end{eqnarray}$$

Let $\operatorname{Mat}(\infty ,F)$ denote the subspace of $\operatorname{Mat}(\mathbb{N},F)$ consisting of matrices whose all but a finite number of coefficients are zeros. Define also

$$\begin{eqnarray}\displaystyle \operatorname{Sym}(n,F) & := & \displaystyle \{X\in \operatorname{Mat}(n,F)\mid X_{ij}=X_{ji},\forall 1\leqslant i,j\leqslant n\},\nonumber\\ \displaystyle \operatorname{Sym}(\mathbb{N},F) & := & \displaystyle \{X\in \operatorname{Mat}(\mathbb{N},F)\mid X_{ij}=X_{ji},\forall i,j\in \mathbb{N}\}.\nonumber\end{eqnarray}$$

Set $\operatorname{Sym}(\infty ,F):=\operatorname{Sym}(\mathbb{N},F)\cap \operatorname{Mat}(\infty ,F)$ .

Two natural group actions under consideration in this paper are:

  1. the group action of $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ on $\operatorname{Mat}(\mathbb{N},F)$ defined by

    $$\begin{eqnarray}\displaystyle ((g_{1},g_{2}),M)\mapsto g_{1}Mg_{2}^{-1},\quad g_{1},g_{2}\in \operatorname{GL}(\infty ,{\mathcal{O}}_{F}),\quad M\in \operatorname{Mat}(\mathbb{N},F); & & \displaystyle \nonumber\end{eqnarray}$$

  2. the group action of $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ on $\operatorname{Sym}(\mathbb{N},F)$ defined by

    $$\begin{eqnarray}\displaystyle (g,M)\mapsto gMg^{t},\quad g\in \operatorname{GL}(\infty ,{\mathcal{O}}_{F}),\quad S\in \operatorname{Sym}(\mathbb{N},F), & & \displaystyle \nonumber\end{eqnarray}$$
    where $g^{t}$ is the transposition of $g$ .

2.4 Conventions

Given a finite set $B$ , we denote by $\#B$ its cardinality.

Let $(\unicode[STIX]{x1D6F4},{\mathcal{B}},m)$ be a measurable space equipped with a positive measure $m$ and let $f$ be a real or complex valued integrable function defined on $\unicode[STIX]{x1D6F4}$ . If $A\subset \unicode[STIX]{x1D6F4}$ is measurable and $0<m(A)<\infty$ , then we write

(14) $$\begin{eqnarray}\displaystyle \unicode[STIX]{x2A0D}_{A}f(x)\,dm(x):=\frac{1}{m(A)}\int _{A}f(x)\,dm(x). & & \displaystyle\end{eqnarray}$$

For any random variable $Y$ , we denote its distribution by ${\mathcal{L}}(Y)$ .

We establish conventions concerning the empty set $\emptyset$ . Let $(r_{i})_{i\in I}$ be a family of real numbers (or complex numbers for the last two formulae). We set

$$\begin{eqnarray}\inf _{i\in \emptyset }r_{i}=+\infty ,\quad \sup _{i\in \emptyset }r_{i}=-\infty ,\quad \mathop{\sum }_{i\in \emptyset }r_{i}=0,\quad \mathop{\prod }_{i\in \emptyset }r_{i}=1.\end{eqnarray}$$

The following conventions will also be used.

  1. As elements in $F$ , $\unicode[STIX]{x1D71B}^{\infty }=\unicode[STIX]{x1D71B}^{+\infty }=0\in F$ .

  2. As elements in $\mathbb{R}\cup \{+\infty \}$ , $q^{\infty }=q^{+\infty }=+\infty$ and $q^{-\infty }=0\in \mathbb{R}$ .

2.5 Haar measure on $\operatorname{GL}(n,{\mathcal{O}}_{F})$

For any $n\in \mathbb{N}$ , denote by $d\text{vol}_{n}$ the Haar measure on $F^{n}$ normalized by the condition $\operatorname{vol}_{n}({\mathcal{O}}_{F}^{n})=1$ . If there is no confusion, we will use the simplified notation $\operatorname{vol}(\cdot )$ for $\operatorname{vol}_{n}(\cdot )$ .

Remark 2.1. The Haar measure $\operatorname{vol}_{n}$ on $F^{n}$ is preserved by any linear map represented by a matrix from the group $\operatorname{GL}(n,{\mathcal{O}}_{F})$ .

For any $n$ , we fix a Haar measure $\operatorname{vol}(\cdot )$ on $\operatorname{Mat}(n,F)$ normalized by $\operatorname{vol}(\operatorname{Mat}(n,{\mathcal{O}}_{F}))=1$ . Up to a multiplicative constant, the Haar measure on the locally compact group $\operatorname{GL}(n,F)$ is uniquely given (see, e.g., Neretin [Ner13]) by

(15) $$\begin{eqnarray}\displaystyle |\text{det}(M)|^{-n}\cdot \operatorname{vol}(dM). & & \displaystyle\end{eqnarray}$$

Let $\operatorname{GL}(n,\mathbf{F}_{q})$ be the group of invertible $n\times n$ matrices over $\mathbf{F}_{q}$ . Set

$$\begin{eqnarray}\displaystyle \operatorname{GL}(n,{\mathcal{C}}_{q}):=\{t=(t_{ij})_{1\leqslant i,j\leqslant n}\in \operatorname{GL}(n,{\mathcal{O}}_{F}):t_{ij}\in {\mathcal{C}}_{q}\}. & & \displaystyle \nonumber\end{eqnarray}$$

Proposition 2.2. A standard partition of $\operatorname{GL}(n,{\mathcal{O}}_{F})$ is given by

(16) $$\begin{eqnarray}\displaystyle \operatorname{GL}(n,{\mathcal{O}}_{F})=\bigsqcup _{t\in \operatorname{GL}(n,{\mathcal{C}}_{q})}(t+\operatorname{Mat}(n,\unicode[STIX]{x1D71B}{\mathcal{O}}_{F})). & & \displaystyle\end{eqnarray}$$

In particular, we have

(17) $$\begin{eqnarray}\displaystyle \operatorname{vol}(\operatorname{GL}(n,{\mathcal{O}}_{F}))=\mathop{\prod }_{j=1}^{n}(1-q^{-j}). & & \displaystyle\end{eqnarray}$$

Proof. By definition, $a=(a_{ij})_{1\leqslant i,j\leqslant n}\in \operatorname{GL}(n,{\mathcal{O}}_{F})$ implies that $a_{ij}\in {\mathcal{O}}_{F}$ and $|\text{det}(a)|=1$ . Now take any $x\in \operatorname{Mat}(n,\unicode[STIX]{x1D71B}{\mathcal{O}}_{F})$ . First, we have $a+x\in \operatorname{Mat}(n,{\mathcal{O}}_{F})$ . Second, write $x=\unicode[STIX]{x1D71B}y$ with $y\in \operatorname{Mat}(n,{\mathcal{O}}_{F})$ . By definition, there exists $z\in {\mathcal{O}}_{F}$ , such that $\det (a+x)=\det (a)+\unicode[STIX]{x1D71B}z$ . Since $|\unicode[STIX]{x1D71B}z|\leqslant q^{-1}$ , by ultrametricity, we obtain $|\text{det}(a+x)|=|\text{det}(a)+\unicode[STIX]{x1D71B}z|=1$ , whence $a+x\in \operatorname{GL}(n,{\mathcal{O}}_{F})$ and the set on the right-hand side of (16) is contained in $\operatorname{GL}(n,{\mathcal{O}}_{F})$ . Conversely, since ${\mathcal{C}}_{q}$ is a complete set of representatives of the cosets of $\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}$ in ${\mathcal{O}}_{F}$ , for any $A\in \operatorname{GL}(n,{\mathcal{O}}_{F})$ , there exists a unique $t\in \operatorname{GL}(n,{\mathcal{C}}_{q})$ , such that $A\equiv t(\operatorname{mod}\unicode[STIX]{x1D71B}{\mathcal{O}}_{F})$ . This completes the proof of (16).

Recalling that

$$\begin{eqnarray}\displaystyle \#\text{GL}(n,{\mathcal{C}}_{q})=\#\text{GL}(n,\mathbf{F}_{q})=\mathop{\prod }_{j=0}^{n-1}(q^{n}-q^{j}), & & \displaystyle \nonumber\end{eqnarray}$$

we arrive at (17). ◻

By Proposition 2.2, $\operatorname{GL}(n,{\mathcal{O}}_{F})$ is an open subgroup of $\operatorname{GL}(n,F)$ . It follows that the restriction on $\operatorname{GL}(n,{\mathcal{O}}_{F})$ of the Haar measure (15) on $\operatorname{GL}(n,F)$ is a Haar measure on $\operatorname{GL}(n,{\mathcal{O}}_{F})$ . Consequently, the normalized Haar measure on $\operatorname{GL}(n,{\mathcal{O}}_{F})$ is given by

(18) $$\begin{eqnarray}\displaystyle \frac{\operatorname{vol}(\cdot )}{\mathop{\prod }_{j=1}^{n}(1-q^{-j})}. & & \displaystyle\end{eqnarray}$$

Let $T(n)$ be sampled uniformly from the finite set $\operatorname{GL}(n,{\mathcal{C}}_{q})$ ; let $V(n)$ be sampled with respect to the normalized Haar measure on $\operatorname{Mat}(n,\unicode[STIX]{x1D71B}{\mathcal{O}}_{F})$ and independent of $T(n)$ .

Proposition 2.3. The random matrix $T(n)+V(n)$ is a Haar random matrix on $\operatorname{GL}(n,{\mathcal{O}}_{F})$ ; that is, the distribution law ${\mathcal{L}}(T(n)+V(n))$ coincides with the normalized Haar measure on $\operatorname{GL}(n,{\mathcal{O}}_{F})$ .

Proof. The proof follows immediately from Proposition 2.2 and the formula (18) for the normalized Haar measure on $\operatorname{GL}(n,{\mathcal{O}}_{F})$ .◻

2.6 Diagonalization in $\operatorname{Mat}(n,F)$

Lemma 2.4 (See, e.g., Neretin [Ner13, § 1.3]).

Every matrix $A\in \operatorname{Mat}(n,F)$ can be written in the form

(19) $$\begin{eqnarray}\displaystyle A=a\cdot \operatorname{diag}(\unicode[STIX]{x1D71B}^{-k_{1}},\unicode[STIX]{x1D71B}^{-k_{2}},\ldots ,\unicode[STIX]{x1D71B}^{-k_{n}})\cdot b,\quad (a,b\in \operatorname{GL}(n,{\mathcal{O}}_{F})), & & \displaystyle\end{eqnarray}$$

where $k_{1}\geqslant k_{2}\geqslant \cdots \geqslant k_{n}\geqslant -\infty$ and $\operatorname{diag}(\unicode[STIX]{x1D71B}^{-k_{1}},\unicode[STIX]{x1D71B}^{-k_{2}},\ldots ,\unicode[STIX]{x1D71B}^{-k_{n}})$ is the diagonal matrix with diagonal coefficients $\unicode[STIX]{x1D71B}^{-k_{1}},\unicode[STIX]{x1D71B}^{-k_{2}},\ldots ,\unicode[STIX]{x1D71B}^{-k_{n}}$ . Moreover, the $n$ -tuple $(k_{1},k_{2},\ldots ,k_{n})$ is uniquely determined by the matrix  $A$ .

2.7 Square-units in $F$

The group of units of the ring ${\mathcal{O}}_{F}$ is given by ${\mathcal{O}}_{F}^{\times }:=\{x\in F:|x|=1\}$ . Let $({\mathcal{O}}_{F}^{\times })^{2}$ be the subgroup of ${\mathcal{O}}_{F}^{\times }$ defined by

$$\begin{eqnarray}\displaystyle ({\mathcal{O}}_{F}^{\times })^{2}:=\{x\in {\mathcal{O}}_{F}^{\times }:\text{there exists }a\in F\text{ such that }x=a^{2}\}. & & \displaystyle \nonumber\end{eqnarray}$$

Let ${\mathcal{C}}_{q}^{\times }$ denote the set ${\mathcal{C}}_{q}\,\setminus \,\{0\}$ and define

$$\begin{eqnarray}\displaystyle ({\mathcal{C}}_{q}^{\times })^{2}:=\{a\in {\mathcal{C}}_{q}^{\times }\mid \text{there exists }y\in F\text{ such that }a=y^{2}\}. & & \displaystyle \nonumber\end{eqnarray}$$

Write

$$\begin{eqnarray}\displaystyle (\mathbf{F}_{q}^{\times })^{2}:=\{a\in \mathbf{F}_{q}^{\times }\mid \text{there exists }c\in \mathbf{F}_{q}^{\times }\text{such that }a=c^{2}\}. & & \displaystyle \nonumber\end{eqnarray}$$

Lemma 2.5. Assume that $F$ is non-dyadic. Then

(20) $$\begin{eqnarray}\displaystyle ({\mathcal{O}}_{F}^{\times })^{2}=\bigsqcup _{a\in ({\mathcal{C}}_{q}^{\times })^{2}}(a+\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}). & & \displaystyle\end{eqnarray}$$

Consequently, the map $\unicode[STIX]{x1D70B}$ in (11) induces a bijection:

(21) $$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D70B}:({\mathcal{C}}_{q}^{\times })^{2}\xrightarrow[{}]{\mathit{bijection}}(\mathbf{F}_{q}^{\times })^{2}. & & \displaystyle\end{eqnarray}$$

Proof. Since $F$ is non-dyadic, we have $|2|=1$ . For any $x=\unicode[STIX]{x1D6FC}^{2}\in ({\mathcal{O}}_{F}^{\times })^{2}$ , since ${\mathcal{C}}_{q}$ is a complete set of representatives for ${\mathcal{O}}_{F}/\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}$ , there exists $a\in {\mathcal{C}}_{q}^{\times }$ , such that $x\equiv a(\operatorname{mod}\unicode[STIX]{x1D71B}{\mathcal{O}}_{F})$ , that is,

(22) $$\begin{eqnarray}\displaystyle |x-a|<1. & & \displaystyle\end{eqnarray}$$

Take any $b\in {\mathcal{O}}_{F}^{\times }$ such that $|b-x|<1=|2\unicode[STIX]{x1D6FC}|^{2}$ . Then the polynomial $P_{b}(X)=X^{2}-b\in {\mathcal{O}}_{F}[X]$ satisfies

(23) $$\begin{eqnarray}\displaystyle |P_{b}(\unicode[STIX]{x1D6FC})|<|P_{b}^{\prime }(\unicode[STIX]{x1D6FC})|^{2}. & & \displaystyle\end{eqnarray}$$

By Hensel’s lemma (see Cassels [Cas86, pp. 49–51]), the inequality (23) implies that there exists $\unicode[STIX]{x1D6FD}\in F$ such that

(24) $$\begin{eqnarray}\displaystyle P_{b}(\unicode[STIX]{x1D6FD})=\unicode[STIX]{x1D6FD}^{2}-b=0\quad \text{and}\quad |\unicode[STIX]{x1D6FD}-\unicode[STIX]{x1D6FC}|\leqslant \frac{|P_{b}(\unicode[STIX]{x1D6FC})|}{|P_{b}^{\prime }(\unicode[STIX]{x1D6FC})|}<|P_{b}^{\prime }(\unicode[STIX]{x1D6FC})|=1. & & \displaystyle\end{eqnarray}$$

In particular, we have

(25) $$\begin{eqnarray}\displaystyle \{b\in {\mathcal{O}}_{F}^{\times }:|b-x|<1\}=x+\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}\subset ({\mathcal{O}}_{F}^{\times })^{2}. & & \displaystyle\end{eqnarray}$$

Combining (22) and (25), we get $a\in ({\mathcal{O}}_{F}^{\times })^{2}$ . Hence

$$\begin{eqnarray}\displaystyle ({\mathcal{O}}_{F}^{\times })^{2}\subset \bigsqcup _{a\in ({\mathcal{C}}_{q}^{\times })^{2}}(a+\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}). & & \displaystyle \nonumber\end{eqnarray}$$

Conversely, since $({\mathcal{C}}_{q}^{\times })^{2}\subset ({\mathcal{O}}_{F}^{\times })^{2}$ , for any $a\in ({\mathcal{C}}_{q}^{\times })^{2}$ , replacing $x$ by $a$ in the above argument, the inclusion (25) implies $a+\unicode[STIX]{x1D71B}{\mathcal{O}}_{F}\subset ({\mathcal{O}}_{F}^{\times })^{2}$ . Hence

$$\begin{eqnarray}\displaystyle \text{}\bigsqcup _{a\in ({\mathcal{C}}_{q}^{\times })^{2}}(a+\unicode[STIX]{x1D71B}{\mathcal{O}}_{F})\subset ({\mathcal{O}}_{F}^{\times })^{2}.\Box & & \displaystyle \nonumber\end{eqnarray}$$

Remark 2.6. Recall that if $F$ is non-dyadic, then the quotient group ${\mathcal{O}}_{F}^{\times }/({\mathcal{O}}_{F}^{\times })^{2}$ has two elements.

2.8 Diagonalization in $\operatorname{Sym}(n,F)$

In what follows, we fix a non-square unit $\unicode[STIX]{x1D700}\in {\mathcal{O}}_{F}^{\times }\,\setminus \,({\mathcal{O}}_{F}^{\times })^{2}$ . By Remark 2.6, the following set

(26) $$\begin{eqnarray}\displaystyle {\mathcal{T}}:=\{\unicode[STIX]{x1D71B}^{-k}\mid k\in \mathbb{Z}\}\sqcup \{\unicode[STIX]{x1D71B}^{-k}\unicode[STIX]{x1D700}\mid k\in \mathbb{Z}\}\sqcup \{0\} & & \displaystyle\end{eqnarray}$$

is a complete set of representatives for the quotient $F/({\mathcal{O}}_{F}^{\times })^{2}$ .

Lemma 2.7. Assume that $F$ is non-dyadic. Then any symmetric matrix $A\in \operatorname{Sym}(n,F)$ can be written in the form

$$\begin{eqnarray}\displaystyle A=g\cdot \operatorname{diag}(x_{1},\ldots ,x_{n})\cdot g^{t},\quad x_{1},\ldots ,x_{n}\in {\mathcal{T}},g\in \operatorname{GL}(n,{\mathcal{O}}_{F}). & & \displaystyle \nonumber\end{eqnarray}$$

Proof. Since ${\mathcal{T}}$ is a complete set of representatives for the quotient $F/({\mathcal{O}}_{F}^{\times })^{2}$ , it suffices to show that any symmetric matrix $A\in \operatorname{Sym}(n,F)$ is diagonalizable. Assume that $A$ is not a zero matrix. We claim that, up to passing $A$ to $gAg^{t}$ for some $g\in \operatorname{GL}(n,{\mathcal{O}}_{F})$ , we may assume that the $(1,1)$ -coefficient of $A$ has maximal absolute value.

Case 1: Assume first that there exists $1\leqslant i_{0}\leqslant n$ , such that $|A_{i_{0}i_{0}}|=\max _{1\leqslant i,j\leqslant n}|A_{ij}|$ . If $i_{0}=1$ , then there is nothing to prove. Otherwise, take $g=M_{(1i_{0})}$ , where $M_{(1i_{0})}$ is the permutation matrix associated with the transposition $(1i_{0})$ . Then the matrix $gAg^{t}$ has $A_{i_{0}i_{0}}$ as its $(1,1)$ -coefficient.

Case 2: Now assume that there exists $i_{0}<j_{0}$ such that

(27) $$\begin{eqnarray}\displaystyle |A_{i_{0}j_{0}}|=\max _{1\leqslant i,j\leqslant n}|A_{ij}|>\max _{1\leqslant i\leqslant n}|A_{ii}|. & & \displaystyle\end{eqnarray}$$

Let us do operations on the submatrix indexed by $\{i_{0},j_{0}\}\times \{i_{0},j_{0}\}$ as follows:

$$\begin{eqnarray}\displaystyle \left[\begin{array}{@{}cc@{}}2A_{i_{0}j_{0}}+A_{i_{0}i_{0}}+A_{j_{0}j_{0}} & A_{i_{0}j_{0}}+A_{j_{0}j_{0}}\\ A_{i_{0}j_{0}}+A_{j_{0}j_{0}} & A_{j_{0}j_{0}}\end{array}\right]=\left[\begin{array}{@{}cc@{}}1 & 1\\ 0 & 1\end{array}\right]\left[\begin{array}{@{}cc@{}}A_{i_{0}i_{0}} & A_{i_{0}j_{0}}\\ A_{i_{0}j_{0}} & A_{j_{0}j_{0}}\end{array}\right]\left[\begin{array}{@{}cc@{}}1 & 0\\ 1 & 1\end{array}\right]. & & \displaystyle \nonumber\end{eqnarray}$$

Since $F$ is non-dyadic and non-Archimedean, (27) implies

$$\begin{eqnarray}\displaystyle |2A_{i_{0}j_{0}}+A_{i_{0}i_{0}}+A_{j_{0}j_{0}}|=|A_{i_{0}j_{0}}+A_{j_{0}j_{0}}|=|A_{i_{0}j_{0}}|. & & \displaystyle \nonumber\end{eqnarray}$$

This shows that we can reduce the second case to the first case where a diagonal coefficient has maximal absolute value.

Now assume that

$$\begin{eqnarray}\displaystyle A=\left[\begin{array}{@{}cc@{}}x & c^{t}\\ c & A_{1}\end{array}\right],\quad c\in F^{n-1}, & & \displaystyle \nonumber\end{eqnarray}$$

such that $x$ attains the maximal absolute value of all coefficients of $A$ . Then $x^{-1}c\in {\mathcal{O}}_{F}^{n-1}$ and we have

$$\begin{eqnarray}\displaystyle \left[\begin{array}{@{}cc@{}}1 & 0\\ -x^{-1}c & 1\end{array}\right]\left[\begin{array}{@{}cc@{}}x & c^{t}\\ c & A_{1}\end{array}\right]\left[\begin{array}{@{}cc@{}}1 & -x^{-1}c^{t}\\ 0 & 1\end{array}\right]=\left[\begin{array}{@{}cc@{}}x & 0\\ 0 & A_{1}-x^{-1}cc^{t}\end{array}\right]. & & \displaystyle \nonumber\end{eqnarray}$$

By continuing the above procedure on the submatrix $A_{1}-x^{-1}cc^{t}$ , we prove finally that $A$ is diagonalizable.◻

Remark 2.8. The assumption that $F$ is non-dyadic is necessary in Lemma 2.7. Indeed, if

$$\begin{eqnarray}\displaystyle \left[\begin{array}{@{}cc@{}}\unicode[STIX]{x1D706}_{1} & 0\\ 0 & \unicode[STIX]{x1D706}_{2}\end{array}\right]=\left[\begin{array}{@{}cc@{}}a & b\\ c & d\end{array}\right]\left[\begin{array}{@{}cc@{}}0 & 1\\ 1 & 0\end{array}\right]\left[\begin{array}{@{}cc@{}}a & c\\ b & d\end{array}\right],\quad \left[\begin{array}{@{}cc@{}}a & c\\ b & d\end{array}\right]\in \operatorname{GL}(2,{\mathcal{O}}_{F}), & & \displaystyle \nonumber\end{eqnarray}$$

then

$$\begin{eqnarray}\displaystyle bc+ad=0,\quad ad-bc\in {\mathcal{O}}_{F}^{\times }. & & \displaystyle \nonumber\end{eqnarray}$$

It follows that $2ad\in {\mathcal{O}}_{F}^{\times }$ and hence $2\in {\mathcal{O}}_{F}^{\times }$ . This implies that $F$ is non-dyadic.

2.9 Characteristic functions

Denote by $\widehat{F}$ the Pontryagin dual of the additive group $F$ . Elements in $\widehat{F}$ are called characters of $F$ . Throughout the paper, we fix a non-trivial character $\unicode[STIX]{x1D712}\in \widehat{F}$ such that

(28) $$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D712}|_{{\mathcal{O}}_{F}}\equiv 1\quad \text{and}\quad \unicode[STIX]{x1D712}\text{ is not constant on }\unicode[STIX]{x1D71B}^{-1}{\mathcal{O}}_{F}. & & \displaystyle\end{eqnarray}$$

For any $y\in F$ , define a character $\unicode[STIX]{x1D712}_{y}\in \widehat{F}$ by $\unicode[STIX]{x1D712}_{y}(x)=\unicode[STIX]{x1D712}(yx)$ . The map $y\mapsto \unicode[STIX]{x1D712}_{y}$ from $F$ to $\widehat{F}$ defines a group isomorphism.

We write explicitly characteristic functions of probability measures in the following situations.

  1. (i) If $\unicode[STIX]{x1D707}$ is a Borel probability measure on $F^{m}$ , then $\widehat{\unicode[STIX]{x1D707}}$ is defined on $F^{m}$ by

    $$\begin{eqnarray}\displaystyle \widehat{\unicode[STIX]{x1D707}}(y):=\int _{F^{m}}\unicode[STIX]{x1D712}(x\cdot y)\unicode[STIX]{x1D707}\,(dx), & & \displaystyle \nonumber\end{eqnarray}$$
    where $x\cdot y:=\sum _{j=1}^{m}x_{j}y_{j}$ .

  2. (ii) If $\unicode[STIX]{x1D707}$ is a Borel probability measure on $\operatorname{Mat}(n,F)$ , then $\widehat{\unicode[STIX]{x1D707}}$ is defined on $\operatorname{Mat}(n,F)$ by

    $$\begin{eqnarray}\displaystyle \widehat{\unicode[STIX]{x1D707}}(A):=\int _{\operatorname{Mat}(n,F)}\unicode[STIX]{x1D712}(\operatorname{tr}(AM))\unicode[STIX]{x1D707}\,(dM). & & \displaystyle \nonumber\end{eqnarray}$$

  3. (iii) If $\unicode[STIX]{x1D707}$ is a Borel probability measure on $\operatorname{Mat}(\mathbb{N},F)$ , then $\widehat{\unicode[STIX]{x1D707}}$ is defined on $\operatorname{Mat}(\infty ,F)$ by

    (29) $$\begin{eqnarray}\displaystyle \widehat{\unicode[STIX]{x1D707}}(A):=\int _{\operatorname{Mat}(\mathbb{N},F)}\unicode[STIX]{x1D712}(\operatorname{tr}(AM))\unicode[STIX]{x1D707}\,(dM). & & \displaystyle\end{eqnarray}$$

  4. (iv) If $\unicode[STIX]{x1D708}$ is a Borel probability measure on $\operatorname{Sym}(n,F)$ , then $\widehat{\unicode[STIX]{x1D708}}$ is defined on $\operatorname{Sym}(n,F)$ by

    $$\begin{eqnarray}\displaystyle \widehat{\unicode[STIX]{x1D708}}(A):=\int _{\operatorname{Sym}(n,F)}\unicode[STIX]{x1D712}(\operatorname{tr}(AS))\unicode[STIX]{x1D708}(dS). & & \displaystyle \nonumber\end{eqnarray}$$

  5. (v) If $\unicode[STIX]{x1D708}$ is a Borel probability measure on $\operatorname{Sym}(\mathbb{N},F)$ , then $\widehat{\unicode[STIX]{x1D708}}$ is defined on $\operatorname{Sym}(\infty ,F)$ by

    $$\begin{eqnarray}\displaystyle \widehat{\unicode[STIX]{x1D708}}(A):=\int _{\operatorname{Sym}(\mathbb{N},F)}\unicode[STIX]{x1D712}(\operatorname{tr}(AS))\unicode[STIX]{x1D708}(dS). & & \displaystyle \nonumber\end{eqnarray}$$

Since the corresponding groups are locally compact, Theorem 31.5 in Hewitt and Ross [HR70, p. 212] implies that in cases (i), (ii) and (iv), the characteristic function $\widehat{\unicode[STIX]{x1D707}}$ determines $\unicode[STIX]{x1D707}$ uniquely. The same statement holds for cases (iii) and (v). Indeed, although the additive groups $\operatorname{Mat}(\mathbb{N},F)$ and $\operatorname{Mat}(\infty ,F)$ are not locally compact and we can not apply the result on locally compact groups directly, we may use the fact that any Borel probability measure $\unicode[STIX]{x1D707}$ on $\operatorname{Mat}(\mathbb{N},F)$ is uniquely determined by its finite dimensional projections $(\operatorname{Cut}_{n}^{\infty })_{\ast }(\unicode[STIX]{x1D707})$ and (29) contains all information for $\widehat{(\operatorname{Cut}_{n}^{\infty })_{\ast }(\unicode[STIX]{x1D707})}$ , $n=1,2,\ldots \,$ . Case (v) is treated similarly.

Remark 2.9. If $\unicode[STIX]{x1D707}$ is a probability measure on $\operatorname{Mat}(n,F)$ which is invariant under the action of the group $\operatorname{GL}(n,{\mathcal{O}}_{F})\times \operatorname{GL}(n,{\mathcal{O}}_{F})$ , then for any $a,b\in \operatorname{GL}(n,{\mathcal{O}}_{F})$ , we have

(30) $$\begin{eqnarray}\displaystyle \widehat{\unicode[STIX]{x1D707}}(a\cdot \operatorname{diag}(\unicode[STIX]{x1D71B}^{-k_{1}},\unicode[STIX]{x1D71B}^{-k_{2}},\ldots ,\unicode[STIX]{x1D71B}^{-k_{n}})\cdot b)=\widehat{\unicode[STIX]{x1D707}}(\operatorname{diag}(\unicode[STIX]{x1D71B}^{-k_{1}},\unicode[STIX]{x1D71B}^{-k_{2}},\ldots ,\unicode[STIX]{x1D71B}^{-k_{n}})). & & \displaystyle\end{eqnarray}$$

Similarly, if $\unicode[STIX]{x1D708}$ is a $\operatorname{GL}(n,{\mathcal{O}}_{F})$ -invariant probability measure on $\operatorname{Sym}(n,F)$ , then for any $g\in \operatorname{GL}(n,{\mathcal{O}}_{F})$ , we have

(31) $$\begin{eqnarray}\displaystyle \widehat{\unicode[STIX]{x1D708}}(g\cdot \operatorname{diag}(x_{1},\ldots ,x_{n})\cdot g^{t})=\widehat{\unicode[STIX]{x1D708}}(\operatorname{diag}(x_{1},\ldots ,x_{n})). & & \displaystyle\end{eqnarray}$$

Similar statements hold for $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measures on $\operatorname{Mat}(\mathbb{N},F)$ and for $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant probability measures on $\operatorname{Sym}(\mathbb{N},F)$ .

Let $m\in \mathbb{N}$ . Given any Borel probability measures $\unicode[STIX]{x1D707}_{1},\ldots ,\unicode[STIX]{x1D707}_{m}$ on $\operatorname{Mat}(\mathbb{N},F)$ (respectively $\operatorname{Sym}(\mathbb{N},F)$ ), their convolution $\unicode[STIX]{x1D707}_{1}\ast \cdots \ast \unicode[STIX]{x1D707}_{m}$ is defined as follows: let $M_{1},\ldots ,M_{m}$ be independent random matrices such that ${\mathcal{L}}(M_{i})=\unicode[STIX]{x1D707}_{i},i=1,\ldots ,m$ and set

$$\begin{eqnarray}\displaystyle \unicode[STIX]{x1D707}_{1}\ast \cdots \ast \unicode[STIX]{x1D707}_{m}:={\mathcal{L}}(M_{1}+\cdots +M_{m}). & & \displaystyle \nonumber\end{eqnarray}$$

The characteristic function of $\unicode[STIX]{x1D707}_{1}\ast \cdots \ast \unicode[STIX]{x1D707}_{m}$ is given by the formula

(32) $$\begin{eqnarray}\displaystyle (\unicode[STIX]{x1D707}_{1}\ast \cdots \ast \unicode[STIX]{x1D707}_{m})^{\wedge }=\mathop{\prod }_{i=1}^{m}\widehat{\unicode[STIX]{x1D707}_{i}}. & & \displaystyle\end{eqnarray}$$

3 Invariance and ergodicity

In this section, we prove that all the measures on $\operatorname{Mat}(\mathbb{N},F)$ from the family $\{\unicode[STIX]{x1D707}_{\Bbbk }={\mathcal{L}}(M_{\Bbbk })\mid \Bbbk \in \unicode[STIX]{x1D6E5}\}$ introduced in Definition 1.1 are $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant and ergodic and that all the measures on $\operatorname{Sym}(\mathbb{N},F)$ from the family $\{\unicode[STIX]{x1D708}_{h}={\mathcal{L}}(S_{h})\mid h\in \unicode[STIX]{x1D6FA}\}$ introduced in Definition 1.5 are $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant and ergodic.

3.1 $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariance for probability measures $\unicode[STIX]{x1D707}_{\Bbbk }$

Proposition 3.1. For any $\Bbbk \in \unicode[STIX]{x1D6E5}$ , the probability measure $\unicode[STIX]{x1D707}_{\Bbbk }$ on $\operatorname{Mat}(\mathbb{N},F)$ is $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant.

Recall that the normalized integral $\unicode[STIX]{x2A0D}$ is introduced in (14).

Remark 3.2. For any $n\geqslant 1$ , we have

(33) $$\begin{eqnarray}\displaystyle \unicode[STIX]{x2A0D}_{\unicode[STIX]{x1D71B}^{-n}{\mathcal{O}}_{F}}\unicode[STIX]{x1D712}(x)\,dx=0. & & \displaystyle\end{eqnarray}$$

Indeed, for any fixed $n\geqslant 1$ , the character $\unicode[STIX]{x1D712}$ defines a non-trivial character $\widetilde{\unicode[STIX]{x1D712}}$ of the finite group $\unicode[STIX]{x1D6E4}_{n}:=\unicode[STIX]{x1D71B}^{-n}{\mathcal{O}}_{F}/{\mathcal{O}}_{F}$ by

$$\begin{eqnarray}\displaystyle \widetilde{\unicode[STIX]{x1D712}}(\unicode[STIX]{x1D6FE}):=\unicode[STIX]{x1D712}(x)\quad (\text{if }\unicode[STIX]{x1D6FE}=x+{\mathcal{O}}_{F},x\in \unicode[STIX]{x1D71B}^{-n}{\mathcal{O}}_{F}). & & \displaystyle \nonumber\end{eqnarray}$$

By orthogonality of the character $\widetilde{\unicode[STIX]{x1D712}}$ and the trivial character, we have

$$\begin{eqnarray}\displaystyle \unicode[STIX]{x2A0D}_{\unicode[STIX]{x1D71B}^{-n}{\mathcal{O}}_{F}}\unicode[STIX]{x1D712}(x)\,dx=\frac{1}{\#\unicode[STIX]{x1D6E4}_{n}}\mathop{\sum }_{\unicode[STIX]{x1D6FE}\in \unicode[STIX]{x1D6E4}_{n}}\widetilde{\unicode[STIX]{x1D712}}(\unicode[STIX]{x1D6FE})=0, & & \displaystyle \nonumber\end{eqnarray}$$

and (33) is proved.

Lemma 3.3. For any $y\in F$ and any $l\in \mathbb{Z}$ , we have

(34) $$\begin{eqnarray}\displaystyle \unicode[STIX]{x2A0D}_{\unicode[STIX]{x1D71B}^{l}{\mathcal{O}}_{F}}\unicode[STIX]{x1D712}(xy)\,dx=\unicode[STIX]{x1D7D9}_{\unicode[STIX]{x1D71B}^{-l}{\mathcal{O}}_{F}}(y). & & \displaystyle\end{eqnarray}$$

Proof. First assume that $y\in \unicode[STIX]{x1D71B}^{-l}{\mathcal{O}}_{F}$ . Then for any $x\in \unicode[STIX]{x1D71B}^{l}{\mathcal{O}}_{F}$ , we have $xy\in {\mathcal{O}}_{F}$ . Consequently, by (28), we have $\unicode[STIX]{x2A0D}_{\unicode[STIX]{x1D71B}^{l}{\mathcal{O}}_{F}}\unicode[STIX]{x1D712}(xy)\,dx=1.$ Now assume that $y\notin \unicode[STIX]{x1D71B}^{-l}{\mathcal{O}}_{F}$ . Since the Haar measure $d\text{vol}$ on $F$ is invariant under the multiplication action by any element $u\in {\mathcal{O}}_{F}^{\times }$ , without loss of generality, we may assume that $y=\unicode[STIX]{x1D71B}^{k}$ with $k\leqslant -l-1$ . By (33), we have

$$\begin{eqnarray}\displaystyle \unicode[STIX]{x2A0D}_{\unicode[STIX]{x1D71B}^{l}{\mathcal{O}}_{F}}\unicode[STIX]{x1D712}(xy)\,dx=\unicode[STIX]{x2A0D}_{\unicode[STIX]{x1D71B}^{l}{\mathcal{O}}_{F}}\unicode[STIX]{x1D712}(\unicode[STIX]{x1D71B}^{k}x)\,dx=\unicode[STIX]{x2A0D}_{{\mathcal{O}}_{F}}\unicode[STIX]{x1D712}(\unicode[STIX]{x1D71B}^{k+l}z)\,dz=0. & & \displaystyle \nonumber\end{eqnarray}$$

This completes the proof of (34). ◻

Lemma 3.4. For any $m\in \mathbb{N}$ , the distribution of the random vector $(X_{i}^{(1)})_{i=1}^{m}$ is $\operatorname{GL}(m,{\mathcal{O}}_{F})$ -invariant.

Proof. Write $X=(X_{i}^{(1)})_{i=1}^{m}$ . It suffices to prove that for any $A\in \operatorname{GL}(m,{\mathcal{O}}_{F})$ and any $y=(y_{1},\ldots ,y_{m})\in F^{m}$ , we have

(35) $$\begin{eqnarray}\displaystyle \mathbb{E}[\unicode[STIX]{x1D712}(X\cdot y)]=\mathbb{E}[\unicode[STIX]{x1D712}((AX)\cdot y)]. & & \displaystyle\end{eqnarray}$$

By the independence between $X_{i}^{(1)}:i=1,\ldots ,m$ and Lemma 3.3, we have

$$\begin{eqnarray}\displaystyle \mathbb{E}[\unicode[STIX]{x1D712}(X\cdot y)]=\mathop{\prod }_{j=1}^{n}\mathbb{E}[\unicode[STIX]{x1D712}(X_{j}y_{j})]=\mathop{\prod }_{j=1}^{n}\unicode[STIX]{x1D7D9}_{{\mathcal{O}}_{F}}(y_{j})=\unicode[STIX]{x1D7D9}_{{\mathcal{O}}_{F}^{m}}(y). & & \displaystyle \nonumber\end{eqnarray}$$

Similarly,

$$\begin{eqnarray}\displaystyle \mathbb{E}[\unicode[STIX]{x1D712}((AX)\cdot y)]=\mathbb{E}[\unicode[STIX]{x1D712}(X\cdot (A^{t}y))]=\unicode[STIX]{x1D7D9}_{{\mathcal{O}}_{F}^{m}}(A^{t}y)=\unicode[STIX]{x1D7D9}_{{\mathcal{O}}_{F}^{m}}(y). & & \displaystyle \nonumber\end{eqnarray}$$

Hence we get (35). The proof of Lemma 3.4 is completed. ◻

Proof of Proposition 3.1.

It suffices to prove that the following probability measures

$$\begin{eqnarray}\displaystyle {\mathcal{L}}[X_{i}^{(1)}Y_{j}^{(1)}]_{i,j\in \mathbb{N}}\quad \text{and}\quad {\mathcal{L}}([Z_{ij}]_{i,j\in \mathbb{N}}) & & \displaystyle \nonumber\end{eqnarray}$$

are $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -invariant. The invariance of both measures follows immediately from Lemma 3.4.◻

3.2 $\operatorname{GL}(\infty ,{\mathcal{O}}_{F})\times \operatorname{GL}(\infty ,{\mathcal{O}}_{F})$ -ergodicity for probability measures $\unicode[STIX]{x1D707}_{\Bbbk }$

Theorem 3.5. For any $\Bbbk \in \unicode[STIX]{x1D6E5}$ , the probability measure $\unicode[STIX]{x1D707}_{\Bbbk }$ on $\operatorname{Mat}(\mathbb{N},F)$ is