Lectures in Representation Theory

Arun Ram
Department of Mathematics and Statistics
University of Melbourne
Parkville, VIC 3010 Australia
aram@unimelb.edu.au

Last update: 12 August 2013

Lecture 4

Definition 1.38 An algebra A is semisimple if it is isomorphic to a direct sum

A≅⨁λ∈Aˆ Mdλ(ℂ)

of matrix algebras where Aˆ is a finite index set and dλ is a positive integer for all λ∈Aˆ.

Let A be semisimple algebra, and let ϕ:A⟶⨁λ∈AˆMdλ(ℂ) be an isomorphism. Then a basis of matrix units for A is

{ eijλ= ϕ-1(Eijλ)  | λ∈Aˆ,1≤ i,j≤dλ } .

For λ∈Aˆ, the minimal central idempotent zλ of A indexed by λ is given by

zλ= ∑i=1dλ ei,iλ.

The minimal ideal Iλ of A indexed by λ is given by

Iλ=ϕ-1 (Mdλ(ℂ)),

and the irreducible representation Wλ of A indexed by λ is given by λ∈Aˆ and given by

Wλ: A ⟶ Mdλ(ℂ) a ⟼ ϕ(a)λ

where ϕ(a)λ denotes the λth block of the matrix ϕ(a). In terms of modules, the irreducible module Wλ, is

Wλ≅A eijλ.

Since Wλ(a)=ϕ(a)λ for all a∈A and λ∈Aˆ, we let Wijλ(a) denote the i,j-entry of the matrix Wλ(a). Then the irreducible character χλ corresponding to λ∈Aˆ is given by

χλ(a)= ∑i=1dλ Wiiλ(a).

Moreover, it should be noted that

Wijλ (ersμ)= δλ,μ δi,r ∂j,s.

Let t→ be a trace on the semisimple algebra A. Then t→ is determined by the weight vector (tλ)λ∈Aˆ which satisfies the following:

(a) t→(a)= ∑λ∈Aˆtλχλ(a), for all a∈A,
(b) t→(eiiλ) =tλ.

We suppose that t→ has the property that tλ≠0 for all λ∈Aˆ (we will see that this means that t→ is non-degenerate) and define a bilinear form on A by

⟨a,b⟩= t→(ab), for a,b∈A.

The form has the following properties:

(a) symmetric: ⟨a,b⟩= t→(ab)= t→(ba)= ⟨b,a⟩,
(b) associative: ⟨ab,c⟩= t→(abc)= ⟨a,bc⟩.
If A has a non-degenerate associative form, then A is a Frobenius algebra. If A has a symmetric, associative non-degenerate form, then A is a symmetric algebra. Many of the results of this chapter hold in these more general algebras.

Let {g1,g2,…,gd} be a basis for A. A dual basis with respect to the bilinear form ⟨,⟩ is the basis {g1*,g2*,…,gd*} having the property that

⟨gi,gj*⟩ =δij.

The basis {eijλ | λ∈Aˆ,1≤i,j≤dλ} of matrix units in A has as its dual basis the set {eijλ/tλ | λ∈Aˆ,1≤i,j≤dλ}. To verify this, we see that

⟨eijλ,esrμ/tμ⟩ =tr(eijλesrμtμ)= δλ,μδjstr (eirλtλ) =δλ,μδjs δir.

Theorem 1.39 Let t→=(tλ)λ∈Aˆ be a trace on A such that tλ≠0 for all λ∈Aˆ.

(a) Fourier Inversion Formula. If λ∈Aˆ and 1≤i,j≤dλ then eijλ= ∑μ∈Aˆ1≤r,s≤dμ tλWijλ (esrμtμ) ersμ.
(b) Central Idempotents. If λ∈Aˆ then zλ=∑μ∈Aˆ1≤r,s≤dμ tλχλ (esrμtμ) ersμ.
(c) Orthogonality of Characters. If λ,ν∈Aˆ, then ∑μ∈Aˆ1≤r,s≤dμ χλ (esrμtμ) χν(ersμ) =δλ,ν dλtλ.

Proof.

For (a) we have Wijλ(esrμtμ)=δλ,μδj,sδi,r, and the result follows.

For (b), we see that

zλ= ∑i=1dλ eiiλ = ∑i=1dλ ∑μ∈Aˆ1≤r,s≤dμ tλWiiλ (esrμtμ) ersμ ⏟Fourier inversion = ∑μ∈Aˆ1≤r,s≤dμ tλ ∑i=1dλ Wiiλ (esrμtμ) ⏟χλ(esrμtμ) ersμ

Part (c) follows by taking the character χλ of each side the equation in (b).

□

Our next goal is to show that these formulas are independent of the basis used. To this end we let 𝒜={g1,…,gd} and ℬ={h1,…,hd} be bases of A and let 𝒜*={g1*,…,gd*} and ℬ*={h1*,…,hd*} be their duals. Let S=(sij) be the transition matrix between 𝒜 and ℬ. That is

hi=∑j=1d sijgj.

If T is the transition matrix between 𝒜* and ℬ*, then

δim= ⟨hi,hm*⟩ = ⟨ ∑j=1d sijgj, ∑n=1d tmngn* ⟩ = ∑j,nsij tmn ⟨gj,gn*⟩ ⏟ =δj,n = ∑j=1d sijtmj.

Therefore, I=STt.

Now consider the orthogonality of characters formula. We have

∑jχλ (hj*)χν (hj) = ∑i=1d χλ (∑ntingn*) χν (∑msimgm) = ∑m,n ( ∑iti,nsi,m ⏟δm,n ) χλ(gn*) χν(gm) = ∑mχλ (gn*) χν(gm).

Thus, orthogonality of characters holds for any basis. In fact, the same argument proves that all three formulas in 1.39 are basis independent. Therefore,

Theorem 1.40 Let A be a semisimple algebra with non-degenerate trace t→=(tλ)λ∈Aˆ (tλ≠0), and let 𝒜={g1,…,gd} be a basis of A. Then

(a) Fourier Inversion Formula. The elements eijλ= ∑g∈𝒜tλ Wijλ(g*)g form a complete set of matrix units for A.
(b) Central Idempotents. The central idempotents of A are given by zλ=∑g∈𝒜 tλχλ(g*) g.
(c) Orthogonality of Characters. If λ,ν∈Aˆ, then ∑g∈𝒜χλ (g*)χν (g)=δλ,ν dλtλ, where dλ is the dimension of the irreducible A-representation Wλ.

Notes and References

This is a copy of lectures in Representation Theory given by Arun Ram, compiled by Tom Halverson, Rob Leduc and Mark McKinzie.

page history