Math 202B: Lecture 23

*** Problems in this lecture due March 15 ***

*** Please complete a course evaluation ***

Let \mathbf{Rep}(G) be the category of finite-dimensional unitary representations (V,\varphi) of a finite group G.

Problem 23.1. One-dimensional unitary representations of G are the same thing as multiplicative characters of G. Explain this sentence carefully.

We also discussed a way to associate a scalar-valued function on G to every unitary representation (V,\varphi) of G. Namely, the character of (V,\varphi) is the function on G defined by

\chi^V(g)= \mathrm{Tr}\ \varphi(g), \quad g \in G.

After objects come morphisms. We have been working on our understanding of the hom-spaces in \mathbf{Rep}(G), which are vector spaces of linear transformations. Namely, for any two unitary representations (V,\varphi) and (W,\psi) of G, the corresponding set of morphisms \mathrm{Hom}_G(V,W) is the subspace of \mathrm{Hom}(V,W) consisting of those linear transformations which satisfy

T\circ \varphi(g) = \psi(g) \circ T, \quad g \in G.

Transformations T \in \mathrm{Hom}_G(V,W) are said to intertwine the actions \varphi and \psi. Equivalently, \mathrm{Hom}_G(V,W) is the space of G-equivariant linear maps from V to W. The first thing we did was prove a dimension formula, which we restate now for ease of reference.

Theorem 23.1. We have

\dim \mathrm{Hom}_G(V,W) = \frac{1}{|G|} \langle \chi^V,\chi^W \rangle.

This is at least reminiscent of our main result about multiplicative characters of a finite abelian group: they form an orthogonal basis of \mathcal{C}(G). So, we want to see some sort of orthogonality emerging for characters of unitary representations which, once again, are higher-dimensional generalizations of multiplicative characters. We saw that the more general version of character orthogonality we are trying to develop is related to the concept of irreducibility.

Problem 23.2. Prove that if G is abelian, all its irreducible representations are one-dimensional.

Last lecture, we proved the following.

Theorem 23.2 (Schur’s Lemma). If (V,\varphi) and (W,\psi) are irreducible unitary representations of G, then every nonzero map T \in \mathrm{Hom}_G(V,W) is an isomorphism.

As a consequence of Theorems 23.1 and 23.2 we have the following: if (V,\varphi) and (W,\psi) are irreducible and non-isomorphic, their characters are orthogonal: we have

\chi^V,\chi^W \rangle = |G| \dim \mathrm{Hom}_G(V,W) =0.

In particular, this generalizes character orthogonality as we know it for finite abelian groups. This leads to the following.

Conjecture 23.3. If (V,\varphi) is an irreducible unitary representation of G, then

\langle \chi^V,\chi^V\rangle = |G|.

The goal of this lecture is to prove Conjecture 23.3. First of all, because of Theorem 23.1, for any unitary representation (V,\varphi) of G we have

\langle \chi^V,\chi^V\rangle = |G|\dim \mathrm{Hom}_G(V,V).

So Conjecture 23.3 is equivalent to the following.

Conjecture 23.4. If (V,\varphi) is an irreducible unitary representation of G, then \mathrm{End}_G(V) = \mathrm{Hom}_G(V,V) is one-dimensional.

First of all, let us clarify that \mathrm{End}_G(V) is the space of all linear operators T \in \mathrm{End}(V) such that

T\rho(g)=\rho(g)T, \quad g \in G.

As we have seen, this is in fact a subalgebra of \mathrm{End}(V), and e know that \mathrm{End}_G(V) and every algebra has a unique one-dimensional subalgebra. So, Conjecture 23.4 is equivalent to the following.

Theorem 23.5. If (V,\varphi) is an irreducible unitary representation of G, then \mathrm{End}_G(V) = \mathbb{C}I.

Proof: Let T \in \mathrm{End}_G(V) and let \alpha be an eigenvalue of T. Then, T - \alpha I is again in the algebra \mathrm{End}_G(V), but at the same time \mathrm{Ker}(T-\alpha I) is a nontrivial G-stable subspace of V. Since (V,\varphi) is irreducible, \mathrm{Ker}(T-\alpha I)=V, which is the same as T=\alpha I. \square

The above argument relies on something we all take for granted. For any algebra \mathcal{A} and any element A \in \mathcal{A}, the spectrum of A is the set

\mathrm{Spec}(A) = \{\alpha \in \mathbb{C} \colon A-\alpha I \text{ not invertible}\}.

Theorem 23.6 (Fundamental Theorem of Algebra). Every T \in \mathrm{End}(V) has nonempty spectrum \mathrm{Spec}(T) \subset \mathbb{C}.

Problem 23.3. Assuming the above, prove that the spectrum of any element of any finite-dimensional von Neumann algebra is nonempty.

The following generalizes Theorem 23.6.

Theorem 23.7 (Burnside’s Theorem). The only subalgebra \mathcal{A} \leq \mathrm{End}(V) which admits no proper non-trivial stable subspace W \leq V is \mathcal{A}=\mathrm{End}(V).

Let (V,\varphi) be any unitary representation of G, and consider the corresponding linear representation of \mathcal{C}(G) defined by

\Phi(E_g) = \varphi(g), \quad g \in G.

Problem 23.4. Prove that (V,\varphi) is irreducible if and only if (V,\Phi) is irreducible.

Combining this problem with Burnside’s theorem, we have that if (V,\varphi) and hence (V,\Phi) is irreducible, then the image of \mathcal{C}(G) in \mathrm{End}(V) is \mathrm{End}(V). Consequently, \mathrm{End}_G(V) is the center of \mathrm{End}(V) and a second proof of Theorem 23.5 follows if we prove directly that the center of an endomorphism algebra is minimal.

Theorem 23.8. The center of \mathrm{End}(V) is \mathbb{C}I.

Proof: Let X \subset V be an orthonormal basis and let \{E_{yx} \colon x,y \in X\} be the corresponding elementary basis of \mathrm{End}(V). Suppose that A \in \mathrm{End}(V) commutes with all elementary operators. Then, for any p,q \in X we have

E_{pp}AE_{qq} = \sum\limits_{x,y \in X} \langle y,Ax\rangle E_{pp}E_{yx}E_{qq} = \sum\limits_{y \in X} \langle p,Ax\rangle E_{px}E_{qq} = \langle p,Aq \rangle E_{pq},

and likewise

E_{qq}AE_{pp} = \langle q,Ap \rangle E_{qp},

so that

p \neq q \implies \langle p,Aq\rangle = \langle q,Ap \rangle =0.

Furthermore, we have

E_{pq}AE_{qp} = \sum\limits_{x,y \in X} \langle y,Ax\rangle E_{pq}E_{yx}E_{qp} =\sum\limits_{x \in X} \langle q,Ax\rangle E_{px}E_{qp} = \langle q,Aq\rangle E_{pp},

and also

E_{pq}E_{qp}A = E_{pp}A = \sum\limits_{x,y \in X} \langle y,Ax \rangle E_{pp}E_{yx} = \sum\limits_{x \in X} \langle p,Ax\rangle E_{px} = \langle p,Ap\rangle E_{pp},

which forces \langle p,Ap\rangle = \langle q,Aq\rangle. \square

Leave a Reply