Lecture 14

Problem 81. Let (π,V)(\pi,V) be a representation of H<GH<G. Define

Cπ(G,V)={f:GV|f(gh)=π(h1)f(g)gG,hH}.C_{\pi}(G,V)=\{f:G\rightarrow V\,|\,f(gh)=\pi(h^{-1})f(g)\;\forall\;g\in G,\;h% \in H\}.
  1. (a)

    Compute dimCπ(G,V)\operatorname{dim}C_{\pi}(G,V).

  2. (b)

    Show that (λ,Cπ(G,V))\big{(}\lambda,C_{\pi}(G,V)\big{)} is a GG-representation, where as usual [λ(g)f](g0)=f(g1g0)[\lambda(g)f](g_{0})=f(g^{-1}g_{0}) for all g,g0Gg,g_{0}\in G.

  3. (c)

    Given a set of representatives 𝐫=r1,,rn{\bf r}=r_{1},\ldots,r_{n}, let T𝐫:Cπ(G,V)V𝐫T_{{\bf r}}:C_{\pi}(G,V)\rightarrow V_{{\bf r}} be the map

    T𝐫(f)=r1f(r1)+r2f(r2)++rnf(rn).T_{{\bf r}}(f)=r_{1}f(r_{1})+r_{2}f(r_{2})+\ldots+r_{n}f(r_{n}).

    Show that T𝐫T_{{\bf r}} is a GG-isomorphism from (λ,Cπ(G,V))\big{(}\lambda,C_{\pi}(G,V)\big{)} to IndHG(π,V)\mathrm{Ind}_{H}^{G}(\pi,V).

Remark. This shows that the induced representations of (π,V)(\pi,V) for different choices of representatives for G/HG/H give rise to isomorphic representations!

Solution:

  1. (a)

    Let {𝐞1,,𝐞m}\{{\bf e}_{1},\ldots,{\bf e}_{m}\} be a basis of VV and let {r1,,rn}\{r_{1},\ldots,r_{n}\} be a set of representatives of left cosets of HGH\leq G. Then define the function fi,jCπ(G,V)f_{i,j}\in C_{\pi}(G,V) to be fi,j(ri)=𝐞jf_{i,j}(r_{i})={\bf e}_{j} and zero on all other coset representatives. This is well defined as, for any gGg\in G, we can write g=rkhg=r_{k}h, for some unique rkr_{k} a representative and some unique hHh\in H, and

    fi,j(g)=fi,j(rkh)=π(h1)fi,j(rk)={π(h1)𝐞j if i=k0 otherwise..f_{i,j}(g)=f_{i,j}(r_{k}h)=\pi(h^{-1})f_{i,j}(r_{k})=\begin{cases}\pi(h^{-1}){% \bf e}_{j}&\text{ if }i=k\\ 0&\text{ otherwise.}\end{cases}.

    It is easy to see that if iii\neq i^{\prime} or jjj\neq j^{\prime} then fi,jf_{i,j} and fi,jf_{i^{\prime},j^{\prime}} are linearly independent. Additionally if γCπ(G,V)\gamma\in C_{\pi}(G,V), then for any g=rihGg=r_{i}h\in G we have

    γ(g)=π(h1)γ(ri)=π(h1)1jmμj𝐞j=π(h1)1jmμjfi,j(ri)=1jmμjfi,j(g),\gamma(g)=\pi(h^{-1})\gamma(r_{i})=\pi(h^{-1})\sum_{1\leq j\leq m}\mu_{j}{\bf e% }_{j}=\pi(h^{-1})\sum_{1\leq j\leq m}\mu_{j}f_{i,j}(r_{i})=\sum_{1\leq j\leq m% }\mu_{j}f_{i,j}(g),

    for some μj\mu_{j}s in {\mathbb{C}}. Thus if we range through all gGg\in G we get

    γ(g)=1in1jmμi,jfi,j(g),\gamma(g)=\sum_{\begin{subarray}{c}1\leq i\leq n\\ 1\leq j\leq m\end{subarray}}\mu_{i,j}f_{i,j}(g),

    and we have a basis of Cπ(G,V)C_{\pi}(G,V). The dimension of Cπ(G,V)C_{\pi}(G,V) is therefore nm=[H:G]dimV{nm=[H:G]\operatorname{dim}V}.

  2. (b)

    Let gGg\in G, then

    [λ(g)(f+μf)](g0)\displaystyle[\lambda(g)(f+\mu f^{\prime})](g_{0}) =(f+μf)(g1g0)\displaystyle=(f+\mu f^{\prime})(g^{-1}g_{0})
    =f(g1g0)+μf(g1g0)=[λ(g)(f)](g0)+μ[λ(g)(f)](g0)\displaystyle=f(g^{-1}g_{0})+\mu f^{\prime}(g^{-1}g_{0})=[\lambda(g)(f)](g_{0}% )+\mu[\lambda(g)(f^{\prime})](g_{0})

    for all g0Gg_{0}\in G, all f,fCπ(G,V)f,f^{\prime}\in C_{\pi}(G,V) and all μ\mu\in{\mathbb{C}}. Thus the image of λ\lambda is in GL(Cπ(G,V))\operatorname{GL}(C_{\pi}(G,V)). Let g,g~Gg,\tilde{g}\in G, then

    [λ(gg~)(f)](g0)=f(g~1g1g0)=[λ(g~)f](g1g0)=[λ(g)[λ(g~)f]](g0),[\lambda(g\tilde{g})(f)](g_{0})=f(\tilde{g}^{-1}g^{-1}g_{0})=[\lambda(\tilde{g% })f](g^{-1}g_{0})=[\lambda(g)[\lambda(\tilde{g})f]](g_{0}),

    for all g0Gg_{0}\in G and all fCπ(G,V)f\in C_{\pi}(G,V). Thus λ\lambda is a group homomorphism and hence (λ,Cπ(G,V))\big{(}\lambda,C_{\pi}(G,V)\big{)} is a GG-representation.

  3. (c)

    Let f,fCπ(G,V)f,f^{\prime}\in C_{\pi}(G,V) and μ\mu\in{\mathbb{C}} then

    T𝐫(f+μf)\displaystyle T_{{\bf r}}(f+\mu f^{\prime}) =r1(f+μf)(r1)++rn(f+μf)(rn)\displaystyle=r_{1}(f+\mu f^{\prime})(r_{1})+\ldots+r_{n}(f+\mu f^{\prime})(r_% {n})
    =r1(f)(r1)++rn(f)(rn)+μ(r1(f)(r1)++rn(f)(rn))\displaystyle=r_{1}(f)(r_{1})+\ldots+r_{n}(f)(r_{n})+\mu\left(r_{1}(f^{\prime}% )(r_{1})+\ldots+r_{n}(f^{\prime})(r_{n})\right)
    =T𝐫(f)+μT𝐫(f),\displaystyle=T_{{\bf r}}(f)+\mu T_{{\bf r}}(f^{\prime}),

    and so T𝐫T_{{\bf r}} is a linear map. Let fkerT𝐫f\in\ker T_{\bf r} then

    T𝐫(f)=r1f(r1)++rnf(rn)=0,T_{{\bf r}}(f)=r_{1}f(r_{1})+\ldots+r_{n}f(r_{n})=0,

    so each f(ri)=0f(r_{i})=0. As we can write any gGg\in G as g=rihg=r_{i}h for some hHh\in H and f(rih)=π(h1)f(ri)f(r_{i}h)=\pi(h^{-1})f(r_{i}) we have that f(g)=0f(g)=0 for all gGg\in G. Thus kerT𝐫=0\ker T_{\bf r}=0. As dimV𝐫=[G:H]dimV=dimCπ(G,V)\operatorname{dim}V_{\bf r}=[G:H]\operatorname{dim}V=\operatorname{dim}C_{\pi}% (G,V) it must be an isomorphism of vector spaces. Lastly let gGg\in G, then for all fCπ(G,V)f\in C_{\pi}(G,V)

    T𝐫(λ(g)f)\displaystyle T_{\bf r}(\lambda(g)f) =r1[λ(g)f](r1)++rn[λ(g)f](rn)\displaystyle=r_{1}[\lambda(g)f](r_{1})+\ldots+r_{n}[\lambda(g)f](r_{n})
    =r1f(g1r1)++rnf(g1rn)\displaystyle=r_{1}f(g^{-1}r_{1})+\ldots+r_{n}f(g^{-1}r_{n})
    =rj(g,1)f(g1rj(g,1))++rj(g,n)f(g1rj(g,n))\displaystyle=r_{j(g,1)}f(g^{-1}r_{j(g,1)})+\ldots+r_{j(g,n)}f(g^{-1}r_{j(g,n)})
    =rj(g,1)f(rj(g1,j(g,1))h(g1,j(g,1)))++rj(g,n)f(rj(g1,j(g,n))h(g1,j(g,n)))\displaystyle=r_{j(g,1)}f(r_{j(g^{-1},j(g,1))}h_{(g^{-1},j(g,1))})+\ldots+r_{j% (g,n)}f(r_{j(g^{-1},j(g,n))}h_{(g^{-1},j(g,n))})
    =rj(g,1)f(rj(e,1)h(e,1)h(g,1)1)++rj(g,n)f(rj(e,n)h(e,n)h(g,n)1)\displaystyle=r_{j(g,1)}f(r_{j(e,1)}h_{(e,1)}h_{(g,1)}^{-1})+\ldots+r_{j(g,n)}% f(r_{j(e,n)}h_{(e,n)}h_{(g,n)}^{-1})
    =rj(g,1)f(r1h(g,1)1)++rj(g,n)f(rnh(g,n)1)\displaystyle=r_{j(g,1)}f(r_{1}h_{(g,1)}^{-1})+\ldots+r_{j(g,n)}f(r_{n}h_{(g,n% )}^{-1})
    =rj(g,1)π(h(g,1))f(r1)++rj(g,n)π(h(g,n))f(rn)\displaystyle=r_{j(g,1)}\pi(h_{(g,1)})f(r_{1})+\ldots+r_{j(g,n)}\pi(h_{(g,n)})% f(r_{n})
    =IndHGπ(g)(r1f(r1)++rnf(rn))\displaystyle=\mathrm{Ind}_{H}^{G}\pi(g)\left(r_{1}f(r_{1})+\ldots+r_{n}f(r_{n% })\right)
    =IndHGπ(g)T𝐫(f),\displaystyle=\mathrm{Ind}_{H}^{G}\pi(g)T_{\bf r}(f),

    so T𝐫T_{\bf r} is a GG-homomorphism. Thus it is a GG-isomorphism and the two representations are isomorphic.