2.10. Lecture 10

In this lecture we will finally define the last of the machinery needed to finish the proof of Theorem 2.8.7, and then we will finish said proof.

2.10.1. Universal projections

Recall that for any representation (π,V)(\pi,V), the sum of matrices 1|G|gGπ(g)\frac{1}{|G|}\sum_{g\in G}\pi(g) acts on VV as the orthogonal projection onto VGV^{G}, the subspace of vectors in VV that are GG-invariant, i.e. the subrepresentation of (π,V)(\pi,V) that is isomorphic to dim(VG)\operatorname{dim}(V^{G}) copies of the irreducible trivial representation. We now generalise this. To start with, recall that for any element of the group algebra gGzggG\sum_{g\in G}z_{g}g\in{\mathbb{C}}G, we define

π(gGzgg):=gGzgπ(g).\pi\left(\sum_{g\in G}z_{g}g\right):=\sum_{g\in G}z_{g}\pi(g).

The space of {\mathbb{C}}-valued functions on GG is isomorphic to G{\mathbb{C}}G; one possible choice of isomorphism is

f1|G|gGf(g)¯g.f\mapsto\frac{1}{|G|}\sum_{g\in G}\overline{f(g)}g.

Composition then lets us define π(f)\pi(f):

π(f):=1|G|gGf(g)¯π(g).\pi(f):=\frac{1}{|G|}\sum_{g\in G}\overline{f(g)}\pi(g).
Lemma 2.10.1.

If fCF(G)f\in CF(G), then π(f)HomG(V)\pi(f)\in\operatorname{Hom}_{G}(V).

Proof.

This is Problem 2.10.3 below. ∎

Proposition 2.10.2.

Let fCF(G)f\in CF(G) and (ρ,W)(\rho,W) be an irreducible representation of GG. Then

ρ(f)=1dim(W)χρ,fGId.\rho(f)=\frac{1}{\operatorname{dim}(W)}\langle\chi_{\rho},f\rangle_{G}\,{\,% \mathrm{Id}}.
Proof.

By Schur’s Lemma 1.4.1, HomG(W)=Id\operatorname{Hom}_{G}(W)={\mathbb{C}}\,{\,\mathrm{Id}}. For any fCF(G)f\in CF(G), Lemma 2.10.1 then gives ρ(f)=λId\rho(f)=\lambda{\,\mathrm{Id}} for some λ\lambda\in{\mathbb{C}}. We compute this λ\lambda by taking the trace of ρ(f)\rho(f):

λdim(W)=tr(ρ(f))=1|G|gGf(g)¯tr(ρ(g))=1|G|gGf(g)¯χρ(g)=χρ,f.\lambda\operatorname{dim}(W)=\operatorname{tr}\big{(}\rho(f)\big{)}=\frac{1}{|% G|}\sum_{g\in G}\overline{f(g)}\operatorname{tr}\big{(}\rho(g)\big{)}=\frac{1}% {|G|}\sum_{g\in G}\overline{f(g)}\chi_{\rho}(g)=\langle\chi_{\rho},f\rangle.

Proof of Theorem 2.8.7, continued:.

We want to prove that {χρ}ρIrr(G)\{\chi_{\rho}\}_{\rho\in\mathrm{Irr}(G)} spans the space CF(G)CF(G). To do this, we assume that fCF(G)f\in CF(G) is orthogonal to all χρ\chi_{\rho}; and need to show that then f=0f=0.

Given any representation (π,V)(\pi,V), we may decompose (π,V)(\pi,V) into irreducible subrepresentations:

(π,V)=(π,W1)(π,Wr).(\pi,V)=(\pi,W_{1})\oplus\ldots\oplus(\pi,W_{r}).

Now, π(f)\pi(f) preserves each irreducible component, hence π(f)=iπ(f)|Wi\pi(f)=\sum_{i}\pi(f)|_{W_{i}}. By Proposition 2.10.2, π(f)|Wi=χρi,fGdim(Wi)Id|Wi\pi(f)|_{W_{i}}=\frac{\langle\chi_{\rho_{i}},f\rangle_{G}}{\operatorname{dim}(% W_{i})}{\,\mathrm{Id}}|_{W_{i}}, where π|Wiρi\pi|_{W_{i}}\cong\rho_{i}. By assumption, χρi,fG=0\langle\chi_{\rho_{i}},f\rangle_{G}=0 for all ii, hence π(f)=0\pi(f)=0

Thus: π(f)=0\pi(f)=0 for any representation (π,V)(\pi,V) of GG. In particular, considering the regular representation (λ,G)(\lambda,{\mathbb{C}}G), we have

𝟎=λ(f)(e)=1|G|gGf(g)¯λ(g)e=1|G|gGf(g)¯g.{\bf 0}=\lambda(f)(e)=\frac{1}{|G|}\sum_{g\in G}\overline{f(g)}\lambda(g)e=% \frac{1}{|G|}\sum_{g\in G}\overline{f(g)}g.

The set {g}gG\{g\}_{g\in G} forms a basis of G{\mathbb{C}}G, so the linear combination is only zero if and only if f(g)¯=0\overline{f(g)}=0 for all gGg\in G, i.e. f=0f=0. ∎

We collect two more useful facts regarding maps π(f)\pi(f):

Corollary 2.10.3.

If (π,V)(\pi,V) and (ρ,W)(\rho,W) are two irreducible representations, then

π(χρ)={1dim(V)Idif(π,V)(ρ,W),0otherwise.\pi(\chi_{\rho})=\begin{cases}\frac{1}{\operatorname{dim}(V)}{\,\mathrm{Id}}% \quad&\mathrm{if}\;(\pi,V)\cong(\rho,W),\\ 0\quad&\mathrm{otherwise}.\end{cases}
Proof.

This follows directly from Proposition 2.10.2 and Theorem 2.8.7. ∎

Corollary 2.10.4.

Let (π,V)(\pi,V) be a representation of GG with decomposition into irreducible subrepresentations

(π,V)=(π,W1)(π,W2)(π,Wr).(\pi,V)=(\pi,W_{1})\oplus(\pi,W_{2})\oplus\ldots\oplus(\pi,W_{r}).

For each ρIrr(G)\rho\in\mathrm{Irr}(G), define

(π,Vρ):=iπ|Wiρ(π,Wi).(\pi,V_{\rho}):=\bigoplus_{\underset{\pi|_{W_{i}}\cong\rho}{i}}(\pi,W_{i}).

Then dim(ρ)π(χρ)\operatorname{dim}(\rho)\,\pi(\chi_{\rho}) is a projection onto VρV_{\rho}.

Proof.

This follows directly from the previous corollary. ∎

The subrepresentation (π,Vρ)(\pi,V_{\rho}) is called the ρ\rho-isotopic component of π\pi, and dim(ρ)π(χρ)\operatorname{dim}(\rho)\pi(\chi_{\rho}) is called the ρ\rho-isotopic projector. Observe that since dim(ρ)π(χρ)\operatorname{dim}(\rho)\,\pi(\chi_{\rho}) is independent of the choice of decomposition of (π,V)(\pi,V) into irreducibles, (π,Vρ)(\pi,V_{\rho}) must also be independent of this choice. This implies that the only choices in the decomposition are given by the different decompositions of each (π,Vρ)(\pi,V_{\rho}) into copies of ρ\rho.

Example 2.10.5.

We consider the permutation representation (π,3)(\pi,{\mathbb{C}}^{3}) of S3S^{3}, and compute the matrices of the isotopic projectors for all the irreducible representations of S3S_{3}:

π(χtriv)=16(π(e)+π(12)+π(13)+π(23)+π(123)+π(132))=16(222222222).\pi(\chi_{{\mathrm{triv}}})=\frac{1}{6}\big{(}\pi(e)+\pi(12)+\pi(13)+\pi(23)+% \pi(123)+\pi(132)\big{)}=\frac{1}{6}\left(\begin{matrix}2&2&2\\ 2&2&2\\ 2&2&2\end{matrix}\right).
π(χsgn)=16(π(e)π(12)π(13)π(23)+π(123)+π(132))=16(000000000).\pi(\chi_{\mathrm{sgn}})=\frac{1}{6}\big{(}\pi(e)-\pi(12)-\pi(13)-\pi(23)+\pi(% 123)+\pi(132)\big{)}=\frac{1}{6}\left(\begin{matrix}0&0&0\\ 0&0&0\\ 0&0&0\end{matrix}\right).
2π(χ(π,W0))=13(2π(e)π(123)π(132))=13(211121112).2\,\pi(\chi_{(\pi,W_{0})})=\frac{1}{3}\big{(}2\pi(e)-\pi(123)-\pi(132)\big{)}=% \frac{1}{3}\left(\begin{matrix}2&-1&-1\\ -1&2&-1\\ -1&-1&2\end{matrix}\right).

2.10.2. Example: the character table of S4S_{4}.

Example 2.10.6.

We determine the character table of S4S_{4}. First, we have the 1-dimensional trivial and sign representations:

e(12)(12)(34)(123)(1234)16386triv11111sgn11111\begin{array}[h]{c|rrrrr}&e&(12)&(12)(34)&(123)&(1234)\\ &1&6&3&8&6\\ \hline\cr{\mathrm{triv}}&1&1&1&1&1\\ \mathrm{sgn}&1&-1&1&1&-1\end{array}

Next, we have the 3-dimensional irreducible permutation representation (π,W0)(\pi,W_{0}) of S4S_{4} on {1,2,3,4}\{1,2,3,4\}, whose character χπ\chi_{\pi} satisfies

χπ(g)=#{fixed points of g}1.\chi_{\pi}(g)=\#\{\text{fixed points of $g$}\}-1.

Notice that we also have an operation on representations known as twisting: if (ρ,V)(\rho,V) is any irreducible representation of GG and ψ\psi is a 1-dimensional character of GG, then we can define a new representation, ψρ\psi\rho, of GG on VV by the formula (ψρ)(g)=ψ(g)ρ(g)(\psi\rho)(g)=\psi(g)\rho(g). It has character ψχρ\psi\chi_{\rho}. In this case, we can look at the representation (sgnπ,W0)(\mathrm{sgn}\pi,W_{0}) and see that it will also be irreducible:

e(12)(12)(34)(123)(1234)16386π31101sgnπ31101\begin{array}[h]{c|rrrrr}&e&(12)&(12)(34)&(123)&(1234)\\ &1&6&3&8&6\\ \hline\cr\pi&3&1&-1&0&-1\\ \mathrm{sgn}\pi&3&-1&-1&0&1\\ \end{array}

Since χπS42=sgnχπS42=1\|\chi_{\pi}\|_{S_{4}}^{2}=\|\mathrm{sgn}\chi_{\pi}\|_{S_{4}}^{2}=1, these are both irreducible. The characters of the two representations are different, hence they are non-isomorphic.

There is one more row of the table to find. This can be done using orthonormality with the first column, this is Corollary 2.8.8. If we recall that the dimension must be a positive integer we can test candidate cases and determine that it must be 22. Thus we obtain that the full character table is

e(12)(12)(34)(123)(1234)16386triv11111sgn11111π31101sgnπ31101ρ20210\begin{array}[h]{c|rrrrr}&e&(12)&(12)(34)&(123)&(1234)\\ &1&6&3&8&6\\ \hline\cr{\mathrm{triv}}&1&1&1&1&1\\ \mathrm{sgn}&1&-1&1&1&-1\\ \pi&3&1&-1&0&-1\\ \mathrm{sgn}\pi&3&-1&-1&0&1\\ \rho&2&0&2&-1&0\end{array}

Notice that we constructed the character of the final representation without constructing the representation itself!

2.10.3. Exercises

.

Problem 58. Prove Lemma 2.10.1.

Problem 59. Let (π,V)(\pi,V) and (ρ,U)(\rho,U) be two representations of a group GG. Show that (π,V)(ρ,U)(\pi,V)\cong(\rho,U) if and only if

dimHomG(V,W)=dimHomG(U,W)\operatorname{dim}\operatorname{Hom}_{G}(V,W)=\operatorname{dim}\operatorname{% Hom}_{G}(U,W)

for all representations (σ,W)(\sigma,W) of GG.

Problem 60. Let (π,V)(\pi,V) be a representation of a group GG and HH a subgroup of GG. Show that

χπ|H(h)=χπ(h)hH.\chi_{\pi|_{H}}(h)=\chi_{\pi}(h)\qquad\forall\,h\in H.

Problem 61. Find the character table of A4A_{4}. Decompose the restriction of each irreducible representation of S4S_{4} to A4A_{4} into irreducibles.

Remark: When we say “decompose a representation into irreducibles”, we simply mean find out how many times each irreducible representation occurs in the decomposition. This is in contrast to if we were to ask for a “decomposition into irreducible subrepresentations”, which involves finding irreducible subrepresentations of the given representation and showing that their direct sum is the whole representation - this is a much harder task.

Problem 62. Decompose [Q8]{\mathbb{C}}[Q_{8}] into irreducible isotopic components.

Problem 63. Let XX be a finite set on which a group GG acts. Consider the permutation representation (π,(X))(\pi,{\mathbb{C}}(X)) of GG on the free vector space of XX.

  1. (a)

    Show that dim((X)G)\operatorname{dim}({\mathbb{C}}(X)^{G}) (recall that (X)G={𝐯(X)|π(g)𝐯=𝐯gG}{\mathbb{C}}(X)^{G}=\{{\bf v}\in{\mathbb{C}}(X)\,|\,\pi(g){\bf v}={\bf v}\,% \forall g\in G\}) is equal to the number of GG-orbits in XX.

  2. (b)

    Compute χπ,χIdG\langle\chi_{\pi},\chi_{{\,\mathrm{Id}}}\rangle_{G}, and use it to show Burnside’s lemma: the number of GG-orbits in XX is equal to

    1|G|gG#{xX|gx=x},\frac{1}{|G|}\sum_{g\in G}\#\{x\in X\,|\,g\cdot x=x\},

    i.e. the average number of fixed points of the group elements.