Lecture 7

Problem 43. Let (π,W0)(\pi,W_{0}) denote the usual n1n-1 dimensional irreducible subrepresentation of the permutation representation of SnS_{n} on n{\mathbb{C}}^{n}. Compute the character χ(π,W0)\chi_{(\pi,W_{0})} of this subrepresentation.

Hint: combine Example 7.7 with Proposition 7.9 (iii).

Solution: We saw in class that the character of the 11-dimensional trivial representation (Id,)({\,\mathrm{Id}},{\mathbb{C}}) of a group GG is given by

χ(Id,)(g)=1gG.\chi_{({\,\mathrm{Id}},{\mathbb{C}})}(g)=1\qquad\forall g\in G.

We also saw that for the permutation representation (π,n)(\pi,{\mathbb{C}}^{n}) of SnS_{n}, we have

χ(π,n)(σ)=#{j{1,,n}|σ(j)=j}σSn.\chi_{(\pi,{\mathbb{C}}^{n})}(\sigma)=\#\big{\{}j\in\{1,\ldots,n\}\,|\,\sigma(% j)=j\big{\}}\qquad\forall\sigma\in S_{n}.

Since (π,n)=(π,(1,1,,1))(π,W0)(\pi,{\mathbb{C}}^{n})=(\pi,{\mathbb{C}}(1,1,\ldots,1))\oplus(\pi,W_{0}), by Proposition 7.9 iii),

χ(π,n)=χ(π,(1,,1))+χ(π,W0)\chi_{(\pi,{\mathbb{C}}^{n})}=\chi_{(\pi,(1,\ldots,1))}+\chi_{(\pi,W_{0})}

Now, (π,(1,,1))(\pi,(1,\ldots,1)) is isomorphic to (Id,)({\,\mathrm{Id}},{\mathbb{C}}), so χ(π,(1,,1))(σ)=1\chi_{(\pi,(1,\ldots,1))}(\sigma)=1 for all σSn\sigma\in S_{n}, giving

χ(π,W0)(σ)=χ(π,n)(σ)1=#{j{1,,n}|σ(j)=j}1σSn.\chi_{(\pi,W_{0})}(\sigma)=\chi_{(\pi,{\mathbb{C}}^{n})}(\sigma)-1=\#\big{\{}j% \in\{1,\ldots,n\}\,|\,\sigma(j)=j\big{\}}-1\qquad\forall\sigma\in S_{n}.

Problem 44. Let gGg\in G be such that gg and g1g^{-1} are both in the same conjugacy class. Show that χπ(g)\chi_{\pi}(g) is real-valued for any representation π\pi of GG.

Solution: Let (π,V)(\pi,V) be a representation of GG. If gg and g1g^{-1} are in the same conjugacy class of GG then there exists a hGh\in G such that hgh1=g1hgh^{-1}=g^{-1}. Thus, by Proposition 7.9(ii), χπ(g)=χπ(g1)\chi_{\pi}(g)=\chi_{\pi}(g^{-1}). However, Proposition 7.9(iv), gives us that χπ(g1)=χπ(g)¯\chi_{\pi}(g^{-1})=\overline{\chi_{\pi}(g)} and thus χπ(g)=χπ(g)¯\chi_{\pi}(g)=\overline{\chi_{\pi}(g)}. We therefore have that χπ(g)\chi_{\pi}(g) is real valued and as (π,V)(\pi,V) was arbitrary this is true for all representations of GG.

Problem 45. Let (π,V)(\pi,V) be a representation of group GG. Show that

  1. (a)

    |χπ(g)|dim(π)gG|\chi_{\pi}(g)|\leq\operatorname{dim}(\pi)\quad\forall g\in G.

  2. (b)

    kerπ=χπ1({dim(π)}).\ker\pi=\chi_{\pi}^{-1}\big{(}\{\operatorname{dim}(\pi)\}\big{)}.

Solution:

  1. (a)

    Let {λi}i=1,,dim(π)\{\lambda_{i}\}_{i=1,\ldots,\operatorname{dim}(\pi)} be the eigenvalues of π(g)\pi(g). Then

    |χπ(g)|=|λ1++λdim(π)||λ1|++|λdim(π)|=dim(π)|\chi_{\pi}(g)|=|\lambda_{1}+\ldots+\lambda_{\operatorname{dim}(\pi)}|\leq|% \lambda_{1}|+\ldots+|\lambda_{\operatorname{dim}(\pi)}|=\operatorname{dim}(\pi)

    (since each λi\lambda_{i} is a root of unity).

  2. (b)

    If gkerπg\in\ker\pi, then χπ(g)=tr(π(g))=tr(Id)=dim(π)\chi_{\pi}(g)=\operatorname{tr}(\pi(g))=\operatorname{tr}({\,\mathrm{Id}})=% \operatorname{dim}(\pi). On the other hand, if χπ(g)=dim(π)\chi_{\pi}(g)=\operatorname{dim}(\pi), we then have

    χπ(g)=λ1++λdim(π)=|λ1|++|λdim(π)|.\chi_{\pi}(g)=\lambda_{1}+\ldots+\lambda_{\operatorname{dim}(\pi)}=|\lambda_{1% }|+\ldots+|\lambda_{\operatorname{dim}(\pi)}|.

    However, since equality in the triangle inequality is only obtained for complex numbers in the {\mathbb{R}}-span of each other, we must then have λ1=λ2==λdim(π)=1\lambda_{1}=\lambda_{2}=\ldots=\lambda_{\operatorname{dim}(\pi)}=1, hence π(g)=Id\pi(g)={\,\mathrm{Id}}.

Problem 46. Let the finite group GG act on the finite set XX, and consider the representation (π,(X))(\pi,{\mathbb{C}}(X)) as in Definition 1.9. Show that

χπ(g)=#{xX|gx=x}.\chi_{\pi}(g)=\#\{x\in X\,|\,g\cdot x=x\}.

Solution: Let ={x}xX{\mathcal{B}}=\{x\}_{x\in X} denote the basis of (X){\mathbb{C}}(X). Note that π(g)x=gx\pi(g)x=g\cdot x. Choosing some ordering of the basis elements, and writing down the coordinate vectors of the π(g)xi\pi(g){x_{i}} gives:

[π(g)xi]=(00100),[\pi(g){x_{i}}]_{{\mathcal{B}}}=\left(\begin{matrix}0\\ \vdots\\ 0\\ 1\\ 0\\ \vdots\\ 0\end{matrix}\right),

where the one is in row jj iff gxi=xjg\cdot x_{i}=x_{j}. Now, as seen previously, we have

χπ(g)=trπ(g)=tr[π(g)]=tr([π(g)x1][π(g)xn]);\chi_{\pi}(g)=\operatorname{tr}\pi(g)=\operatorname{tr}[\pi(g)]_{{\mathcal{B}}% }=\operatorname{tr}\left(\begin{matrix}\vdots&&\vdots\\ [\pi(g){x_{1}}]_{{\mathcal{B}}}&\cdots&[\pi(g){x_{n}}]_{{\mathcal{B}}}\\ \vdots&&\vdots\end{matrix}\right);

and each column has a one on the diagonal iff gxi=xig\cdot x_{i}=x_{i}, and a zero otherwise. Thus

χπ(g)=#{i|gxi=xi}=#{xX|gx=x}.\chi_{\pi}(g)=\#\{i\,|\,g\cdot x_{i}=x_{i}\}=\#\{x\in X\,|\,g\cdot x=x\}.