1 Representation theory of finite groups

1.5 Maschke’s theorem and complete reducibility

1.5.1 Maschke’s theorem

Theorem 1.40.

Suppose that G is a finite group and that V is a representation of G over a field of characteristic {not dividing |G|}. Suppose that W is a subrepresentation of V. Then there is a subrepresentation W of V such that

VWW.
Proof.

First we make a useful definition.

Definition 1.41.

If WV are vector spaces, then a linear map π:VW is a projection if

π(w)=w

for all wW.

Exercise 1.42.

If π:VW is a projection, then

V=Wker(π).

We will construct a projection π:VW that is a G-homomorphism (we could call this a G-projection). Given such a π, the proof is easy: let W=ker(π). This is a subrepresentation as π is a G-homomorphism, and by exercise we have V=WW as required.

It remains to construct π. Let π0:VW be any linear map such that π0|W=idW (to construct it, choose a basis for W and extend it to a basis for V. Then define π to be the identity on the basis of W and whatever you like on the other basis vectors). This might not be a G-homomorphism, but we turn it into one using an ‘averaging trick’: define

π(v)=1|G|gGg-1π0(gv).

Then this is a G-homomorphism: for any hG,

h-1π(hv) =1|G|gGh-1g-1π0(ghv)
=1|G|kGk-1π0(kv) writing k=gh
=π(v),

whence π(hv)=hπ(v).

Finally, if vW then π(v)=1|G|gGg-1gv=v, so π is a projection. ∎

Maschke’s Theorem and Schur’s Lemma both hold in the situation that k= and G is finite. From now on, we assume that this is the case, and that all representations are finite-dimensional.

1.5.2 Complete reducibility

The following corollary of Maschke’s theorem says that any (finite-dimensional) representation of a (finite) group can be written as a direct sum of irreducible representations, in an essentially unique way. So irreducible representations are the ’prime numbers’ of representation theory.

Corollary 1.43.

Let V be a representation of G. Then

VW1W2Wr

for some irreducible representations W1,,Wr.

Moreover, the number of times each isomorphism class of irreducible representation shows up in the above decomposition is independent of the exact choice of decomposition.

Proof.

The existence of such a decomposition follows from Maschke’s theorem and induction: let W1 be an irreducible subrepresentation, write V=W1W1 by Maschke, repeat starting with W1.

If W is an irreducible representation and VW1Wr, then

dimHomG(W,V)=dimHomG(W,Wi)=#{i:WWi}

by Schur’s Lemma (specifically, part (3) of 1.32). This only depends on V and W, not on the choice of decomposition of V. ∎

In the previous proof, we used an easy, but important, property of Hom:

Lemma 1.44.

If V,V,W,W are representations of G, then

HomG(V,WW) HomG(V,W)HomG(V,W)
HomG(VV,W) HomG(V,W)HomG(V,W).

We isolate the following part of the proof of 1.43 for later use.

Lemma 1.45.

If ρ is an irreducible representation of G and σ is some other representation of V, then the number of times ρ appears in the irreducible decomposition of σ is exactly

dimHomG(ρ,σ).
Example 1.46.

We work out the projections constructed in the proof of Maschke’s theorem when V is the permutation representation of S3 on 3. Then V has a subspace V0={(x,y,z):x=y=z}. The map

V V0
(x,y,z) 13(x+y+z)(e1+e2+e3)

is a G-equivariant projection. As in the proof of Maschke’s theorem, its kernel V1={(x,y,z):x+y+z=0} is a complement:

V=V0V1.

The representation V1 is irreducible (under the isomorphism S3D3, it is ρ1). We can also write down a G-equivariant projection VV1:

(x,y,z)13(2x-y-z,2y-x-z,2z-x-y).

The kernel of this projection is V0.

Remark 1.47.

The decomposition of Corollary 1.43 is not unique: if G is the trivial group, then 2 can be written as in infinitely many ways, simply by choosing any two distinct lines. However, if we have an irreducible representation (ρ,W) of G, then for any representation V the subspace V(ρ) spanned by all the subrepresentations of V isomorphic to ρ is uniquely determined: it is called the ρ-isotypic component of V. The key example to keep in mind is, if GCn is generated by g, and χ is a character with χ(g)=ω, then V(ω) is just the ω-eigenspace of g acting on V.

1.5.3 Unitarizability

We didn’t cover this section, but it is recommended reading.

Recall that a (Hermitian) inner product on a complex vector space V is a map (v,w):V×V such that

(v+v,w) =(v,w)+(v,w),
(v,w+w) =(v,w)+(v,w),
(λv,μw) =λ¯μ(v,w),and
(w,v) =(v,w)¯

for all v,v,w,wV, λ,μ, and that is positive definite, meaning that (v,v)>0 for all nonzero vV. The standard example is

((w1,,wn),(z1,,zn))=w¯1z1++w¯nzn

on V=n. A (Hermitian) inner product space is a vector space together with a chosen Hermitian inner product.

Remark 1.48.

If (,) is a Hermitian inner product on n, and 𝐰 and 𝐳n are written as column vectors, then we can write

(𝐰,𝐳)=𝐰H𝐳

where 𝐰 is the complex conjugate of the transpose of 𝐰. The matrix H will satisfy

H=H

and be diagonalizable with positive real eigenvalues.

If (,) is a Hermitian inner product on V and W is a subspace, then the orthogonal complement is

W={vV:(v,w)=0 for all wW}.

As vector spaces, we have V=WW.

Definition 1.49.

A representation V of G is unitarizable if there is a G-invariant inner product on V; that is, a Hermitian inner product such that

(gv,gw)=(v,w)

for all gG.

If such an inner product is chosen, then the representation V is said to be unitary.

Exercise 1.50.

Suppose that V is a unitary representation of G and that W is a subrepresentation. Then W is a subrepresentation of V.

Theorem 1.51.

If V is a complex representation of G, then V is unitarizable.

Proof.

Start with any Hermitian inner product, (,), not necessarily G-invariant. Then average it over G. ∎

Exercise 1.52.

Fill in the details of the proof of Theorem 1.51, and use it together with exercise 1.50 to give a second proof of Maschke’s theorem.