Schur’s orthogonality relations for the free group
Guillaume Delord · 2025
Read Schur’s orthogonality relations for the free group by Guillaume Delord (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-03924.txt.
Introduction
Guillaume Delord works on Schur's orthogonality relations for the free group, in a mathematics paper from 2025. The paper deals with the quasi-regular representation of a finitely generated free group. Such relations state how matrix coefficients of a representation pair up when averaged over the group.
Delord shows that integrals of these products over larger and larger balls in the group settle to an explicit limit once they are scaled in the right way. The result gives a precise large-scale picture of the coefficients. Readers in group theory and representation theory get a clean convergence statement for a group where many sums fail to behave so well.
Catalog record
| CREATOR | Guillaume Delord |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 20 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-03924
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Schur’s orthogonality relations for the free group Thanks: Thanks to Ch. Pittet, A. Boyer and A. Pinochet Lobos for their great help and generosity Guillaume Delord Abstract . We prove an explicit convergence of suitably normalized integrals on balls where the integrand is the product of coefficients of the quasi-regular representation of the finitely generated free group. This result follows from the fact that the quasi-regular representation of the free group is c-tempered in the sens of Kazhdan and Yom Din [ KYD22 , def 2.1] . The convergence can be summed up by the formula: lim n → ∞ 1 n 3 ∫ B n ⟨ π ( g ) ψ 1 , ψ 2 ⟩ ⟨ π ( g ) ψ 3 , ψ 4 ⟩ ¯ 𝑑 μ ( g ) = 3 q ( q + 1 ) ( q − 1 ) 2 ⟨ ψ 1 , ψ 3 ⟩ ⟨ ψ 2 , ψ 4 ⟩ ¯ \underset{n\to\infty}{\lim}\frac{1}{n^{3}}\int_{B_{n}}\langle\pi(g)\psi_{1},\psi_{2}\rangle\overline{\langle\pi(g)\psi_{3},\psi_{4}\rangle}d\mu(g)=\frac{3q(q+1)}{(q-1)^{2}}\langle\psi_{1},\psi_{3}\rangle\overline{\langle\psi_{2},\psi_{4}\rangle} where q + 1 q+1 is twice the rank of the free group, B n B_{n} is the ball of radius n n with respect to the canonical word metric and ψ 1 , ψ 2 , ψ 3 , ψ 4 \psi_{1},\psi_{2},\psi_{3},\psi_{4} are square summable functions on the boundary of the free group. 1. Introduction. Given an irreducible unitary representation of a compact group \rho:G\to\mathcal{U}(\mathcal{V}) , we know (see for instance [ BBP23 ] ) that the Hilbert space \mathcal{V} is finite dimensional (say d:=dim_{\mathbf{C}}(\mathcal{V}) ) and, for all v_{1},v_{2},v_{3} and v_{4} in \mathcal{V} , we have: \int_{G}\langle\rho(g)v_{1},v_{2}\rangle\overline{\langle\rho(g)v_{3},v_{4}\rangle}dg=\frac{1}{d}\langle v_{1},v_{3}\rangle\overline{\langle v_{2},v_{4}\rangle} where dg denotes the normalised Haar measure on G . This formula (Schur’s orthogonality relations) can be seen as a generalisation of the fact that characters of finite groups are unitary in l^{2}(G,\mathbf{C}) but does not make sense when G is not compact and the coefficients not square summable. In this paper, we compute an equivalent asymptotic formula for the boundary representation of the free group. In [ BG16 , First theorem] , one can find similar results for Gromov hyperbolic groups when the metric is non arithmetic . 1.1. Settings and notations. Let G be the free group with N generators, X the Cayley graph associated to right multiplication in G and x_{0} a base point in X . Then X is an homogeneous tree of degree 2N=:q+1 equipped with the unique distance d which gives the value 1 to any pair of adjacent vertices (see figure 1 for an example where N=2 and x_{0}=e the neutral element of G ). We denote by [x,y] the unique geodesic joining x to y in X , S_{k}:=S(x_{0},k):=\{x\in X||x|:=d(x_{0},x)=k\} the sphere centered at x_{0} with radius k\in\mathbf{N} (for an element g of G , we write also g\in S_{k} whenever gx_{0}\in S(x_{0},k) as an element of X ). We also denote by B_{n}:=B(x_{0},n) the ball with radius n\in\mathbf{N} . Figure 1. Example for N=2 and x_{0}=e A point \omega of the boundary \Omega can be seen as a direction to infinity or, more precisely, as an infinite geodesics [x_{0},\omega) starting at x_{0} . We equip \Omega with a topological structure declaring its basis of open sets to be all the shadows \Omega_{x}=\{\omega\in\Omega|[x_{0},x]\subset[x_{0},\omega)\} , where x is in X , which makes \Omega a compact topological space (for more details, see the introduction of [ KS92 ] where Kuhn embeds X\bigsqcup\Omega in a cartesian product of compact spaces). We also equip \Omega with a Borel probability measure \nu which satisfy: \forall x\in X,\nu(\Omega_{x})=\frac{1}{|S_{|x|}|}=\frac{1}{(q+1)q^{|x|-1}}. The isometric left action of G on X clearly extends to a left action on \Omega and one can show that \nu is quasi-invariant under this action (see [ BOU95 , Corollary 2.6.3] where Bourdon shows that G acts by conformal maps on \Omega ). In particular, for all g in G , g_{*}\nu<<\nu . One can show that \frac{dg_{*}\nu}{d\nu}(\omega)=P(g^{-1},\omega):=q^{\beta_{\omega}(x_{0},gx_{0})} where \beta is the Busemann function (see also [ BOU95 ] ). Let \mathcal{H} be the Hilbert space L^{2}(\Omega,\nu,\mathbf{C}) . We define the unitary representation \pi:G\to\mathcal{U}(\mathcal{H}) by: \forall g\in G,\psi\in\mathcal{H},\omega\in\Omega,\pi(g)\psi(\omega):=P(g^{-1},\omega)^{\frac{1}{2}}\psi(g^{-1}\omega). It is well known \pi is irreducible (see for example [ BL17 , thm 1.2] ). In particular, by Schur’s lemma: (1) Hom_{G}(\pi,\pi)=\{T\in\mathcal{B}(\mathcal{H})\text{ such that }T\circ\pi(g)=\pi(g)\circ T,\forall g\in G\}=\mathbf{C}Id_{\mathcal{H}}. 1.2. Organisation of the paper. In section 2 we compute the values of the Harish-Chandra function and prove that it is spherical using partitions of the boundary where the Busemann function is constant. In section 3 , we show (using results from [ BL17 ] ) that the representation is c-tempered in the sense of [ KYD22 ] (using results from [ Kuh94 ] and [ Haa78 ] about the extension of our representation and the regular one). To conclude, in section 4 , we adapt and detail the proof one can find in [ KYD22 ] to obtain our asymptotic orthogonality relations by showing some intermediate results based on functional analysis from [ RS81 ] and explain how we obtain our initially mentioned formula. 2. Computing the Harish-Chandra function. 2.1. Partitions where the Busemann function is constant. Lemma 2.1 . \textit{Let }x\in S_{n}\textit{ such that }[x_{0},x]=(x_{0},x_{1},\dots,x=x_{n}) . One defines the following sets: E_{k}(x):=\begin{cases}\Omega_{x}&\textit{ if }k=n\\ \Omega_{x_{k}}-\Omega_{x_{k+1}}&\textit{ if }0\leq k<n\end{cases} Then \{E_{0}(x),\dots,E_{n}(x)\} is a partition of \Omega (see Figure 2 for an example where n=2 ). Moreover, \omega\mapsto\beta_{\omega}(x_{0},x) is constantly equal to 2k-n on E_{k}(x) . Figure 2. Example with [x_{0},x]=(x_{0},x_{1},x_{2}) illustrating Lemma 2.1 where the family (E_{0},E_{1},E_{2}) is a partition of the boundary. One computes: (2) \nu(E_{k}(x))=\begin{cases}\nu(\Omega_{x})=\frac{1}{(q+1)q^{n-1}}&($if $k=n)\\ \nu(\Omega_{x_{k}})-\nu(\Omega_{x_{k+1}})=\frac{q-1}{(q+1)q^{k}}$ $&($if $k\in[1,n-1])\\ 1-\nu(\Omega_{x_{1}})=\frac{q}{q+1}&($if $k=0)\end{cases} 2.2. Computation: The Harish-Chandra function is spherical. Here we define \Xi and show that it is constant on S_{n} by computing its value. \begin{array}[]{cccc}\Xi:&G&\to&\mathbf{R}^{+*}\\ &g&\mapsto&\langle\pi(g)1_{\Omega},1_{\Omega}\rangle\end{array} Note that \Xi(0):=\Xi(e)=1 Now, let n\geq 1 , and g such that g\in S_{n} and compute \displaystyle\Xi(g) \displaystyle=\int_{\Omega}\pi(g)1_{\Omega}\overline{1_{\Omega}}d\nu=\int_{\Omega}P(g^{-1},\omega)^{\frac{1}{2}}d\nu(\omega)=\sum_{k=0}^{n}\int_{E_{k}(gx_{0})}q^{\frac{\beta_{\omega}(x_{0},gx_{0})}{2}}d\nu(\omega) \displaystyle=\sum_{k=0}^{n}q^{\frac{2k-n}{2}}\nu(E_{k}(gx_{0}))\qquad(\text{by Lemma }\ref{partition of finite geodesic}) \displaystyle=q^{-\frac{n}{2}}\frac{q}{q+1}+\sum_{k=1}^{n-1}q^{\frac{2k-n}{2}}\frac{q-1}{(q+1)q^{k}}+q^{\frac{n}{2}}\frac{1}{(q+1)q^{n-1}}\quad(\text{from Equation }\ref{E_k(x) measure}) \displaystyle=\bigg[1+\big(\frac{q-1}{q+1}\big)n\bigg]q^{-\frac{n}{2}}. So, \Xi is spherical and we can define: \begin{array}[]{ccccc}\Xi:&\mathbf{N}&\to&\mathbf{R}^{+*}&\\ &n&\mapsto&\Xi(g)&(\text{where }g\text{ is any element of }S_{n})\end{array} 3. \pi is c-tempered. Here we will prove that \pi is c-tempered, in the sense of Kazhdan and Yom Din in section 2 «Notion of c-temperedness» [ KYD22 , def 2.1] . For all subset L\subset G and all \psi_{1},\psi_{2}\in\mathcal{H} , one can define the quantity: M_{\psi_{1},\psi_{2}}(L):=\int_{L}|\langle\pi(g)\psi_{1},\psi_{2}\rangle|^{2}d\mu(g) In particular, we have M_{1_{\Omega},1_{\Omega}}(L)=\int_{L}\Xi(g)^{2}d\mu(g) Using the spherical property of \Xi , one computes: (3) M_{1_{\Omega},1_{\Omega}}(S_{k})=\frac{q+1}{q}\left(1+\frac{q-1}{q+1}k\right)^{2}\quad(k>0). and (4) \displaystyle M_{1_{\Omega},1_{\Omega}}(B_{n}) \displaystyle=\sum_{k=0}^{n}M_{1_{\Omega},1_{\Omega}}(S_{k}) \displaystyle=1+\frac{q+1}{q}\bigg[\frac{n(n+1)(2n+1)}{6}\left(\frac{q-1}{q+1}\right)^{2}+n(n+1)\frac{q-1}{q+1}+n\bigg] \displaystyle\underset{n\to\infty}{\sim}\frac{n^{3}}{K}\text{ where }K=\frac{3q(q+1)}{(q-1)^{2}}. Considering the sequence \{B_{n}\}_{n\in\mathbf{N}} of balls in G and our unit vector 1_{\Omega}\in\mathcal{H} . Then if the two following conditions are satisfied: (5) \forall\psi_{1},\psi_{2}\in\mathcal{H},\quad\underset{n\to\infty}{\limsup}\frac{M_{\psi_{1},\psi_{2}}(B_{n})}{M_{1_{\Omega},1_{\Omega}}(B_{n})}<\infty; (6) \forall g,h\in G\text{ and }\psi_{1},\psi_{2}\in\mathcal{H},\quad\underset{n\to\infty}{\lim}\frac{M_{\psi_{1},\psi_{2}}(B_{n}\Delta h^{-1}B_{n}g)}{M_{1_{\Omega},1_{\Omega}}(B_{n})}=0. One says that \pi is c-tempered with Følner sequence \{B_{n}\}_{n\in\mathbf{N}} (see [ KYD22 , def 2.1] ). Remark 3.1 . The condition ( 6 ) is equivalent to the second condition in the definition [ KYD22 , def 2.1] because G is discret, therefore its compacts are the finite sets. Lemma 3.2 . \pi satisfies the first condition ( 5 ). Proof. By [ Kuh94 ] , we know that \pi is weakly contained in the regular representation \pi_{reg} . In particular, the extensions on l^{1}(G) of these representations: \rho^{ext}(\sum_{g\in G}a_{g}\delta_{g})=\int_{G}a_{g}\rho(g)dg=\sum_{g\in G}a_{g}\rho(g) (where \rho\in\{\pi,\pi_{reg}\} )