4.20. Lecture 20

4.20.1. The branching rule

We may view Snโˆ’1S_{n-1} as a subgroup of SnS_{n} by letting it consist of all permutations that fix nn, i.e.

Snโˆ’1={ฯƒโˆˆSn|ฯƒโข(n)=n}=StabSnโข(n).S_{n-1}=\{\sigma\in S_{n}\,|\,\sigma(n)=n\}=\mathrm{Stab}_{S_{n}}(n).

The goal of this section is to decompose ResSnโˆ’1Snโก(ฯ€,๐’ฎฮป)\operatorname{Res}_{S_{n-1}}^{S_{n}}(\pi,{\mathcal{S}}^{\lambda}) into irreducible representations of Snโˆ’1S_{n-1}, i.e.ย Specht modules.

We say that a box in the Young diagram of ฮป\lambda is removable if it has no boxes below or to the right of it.

Theorem 4.20.1.

For any ฮปโŠขn\lambda\vdash n,

ResSnโˆ’1Snโก(ฯ€,๐’ฎฮป)=โจฮปโˆ’โŠข(nโˆ’1)(ฯ€,๐’ฎฮปโˆ’),\operatorname{Res}_{S_{n-1}}^{S_{n}}(\pi,{\mathcal{S}}^{\lambda})=\bigoplus_{% \lambda^{-}\vdash(n-1)}(\pi,{\mathcal{S}}^{\lambda^{-}}),

where the direct sum runs over all ฮปโˆ’โŠข(nโˆ’1)\lambda^{-}\vdash(n-1) whose Young diagrams are obtained by removing a removable box from the Young diagram for ฮป\lambda.

Example 4.20.2.
\ytableausetup

notabloids,smalltableaux

ResS4S5โก(ฯ€,๐’ฎ\ydiagramโข3,2)=(ฯ€,๐’ฎ\ydiagramโข3,1)โŠ•(ฯ€,๐’ฎ\ydiagramโข2,2).\operatorname{Res}_{S_{4}}^{S_{5}}\left(\pi,{\mathcal{S}}^{\ydiagram{3,2}}% \right)=\left(\pi,{\mathcal{S}}^{\ydiagram{3,1}}\right)\oplus\left(\pi,{% \mathcal{S}}^{\ydiagram{2,2}}\right).
Proof.

We label the removable boxes of ฮป\lambda by ๐–ผ1,โ€ฆโข๐–ผr\mathsf{c}_{1},\ldots\mathsf{c}_{r}, with the numbering going from bottom to top. For each i=1,โ€ฆ,ri=1,\ldots,r, we define a map Tiโ€ฒ:โ„ณฮปโ†’โ„ณฮปโˆ–๐–ผiT_{i}^{\prime}:{\mathcal{M}}^{\lambda}\rightarrow{\mathcal{M}}^{\lambda% \setminus\mathsf{c}_{i}} by defining it on the basis given by tabloids (and then extending linearly):

Tiโ€ฒโข([t]):={0ifโขnโขandโข๐–ผiโขareโขnotโขinโขtheโขsameโขrow[t]โˆ–notherwise,T^{\prime}_{i}([t]):=\begin{cases}0\quad&\mathrm{if\;}n\;\mathrm{and}\;\mathsf% {c}_{i}\;\mathrm{are\;not\;in\;the\;same\;row}\\ [t]\setminus n\quad&\mathrm{otherwise,}\end{cases}

i.e.ย if not zero, then ย Tiโ€ฒโข([t])T^{\prime}_{i}([t]) is the Young tabloid of shape ฮปโˆ–๐–ผi\lambda\setminus\mathsf{c}_{i}, with the nn removed from [t][t]. We claim that Tiโ€ฒT_{i}^{\prime} is an Snโˆ’1S_{n-1}-homomorphism from (ResSnโˆ’1Snโกฯ€,โ„ณฮป)(\operatorname{Res}_{S_{n-1}}^{S_{n}}\pi,{\mathcal{M}}^{\lambda}) to (ฯ€,โ„ณฮปโˆ–๐–ผi)(\pi,{\mathcal{M}}^{\lambda\setminus\mathsf{c}_{i}}) (see Problem 4.20.3).

Observe that for elements of ๐’๐˜๐“ฮป\mathbf{SYT}^{\lambda}, the number nn must be in a removable box, since there are no greater numbers to place either to its right or below. Supposing that tโˆˆ๐’๐˜๐“ฮปt\in\mathbf{SYT}^{\lambda} is such that nn is in ๐–ผj\mathsf{c}_{j}, for any element ฯƒโˆˆCโข(t)\sigma\in C(t) the element nn is then in in the same row of [ฯƒโ‹…t][\sigma\cdot t] as ๐–ผj\mathsf{c}_{j} or above. This gives Tiโ€ฒโข([ฯƒโ‹…t])=0T_{i}^{\prime}([\sigma\cdot t])=0 for all j>ij>i, since ๐–ผj\mathsf{c}_{j} is in a higher row than ๐–ผi\mathsf{c}_{i}. In particular, we obtain ๐žtโˆˆkerโกTiโ€ฒ{\bf e}_{t}\in\ker T^{\prime}_{i} for all j>ij>i.

On the other hand, if tโˆˆ๐’๐˜๐“ฮปt\in\mathbf{SYT}^{\lambda} has the number nn in the same row as ๐–ผi\mathsf{c}_{i}, then

Tiโ€ฒโข(๐žt)=โˆ‘ฯƒโˆˆCโข(t)sgnโก(ฯƒ)โขTiโ€ฒโข([ฯƒโ‹…t])=(โˆ‘ฯƒโˆˆCโข(t)โˆฉSnโˆ’1sgnโก(ฯƒ)โขTiโ€ฒโข([ฯƒโ‹…t]))+(โˆ‘ฯƒโˆˆCโข(t)โˆ–Snโˆ’1sgnโก(ฯƒ)โขTiโ€ฒโข([ฯƒโ‹…t])).\displaystyle T_{i}^{\prime}({\bf e}_{t})=\sum_{\sigma\in C(t)}\operatorname{% sgn}(\sigma)T^{\prime}_{i}([\sigma\cdot t])=\left(\sum_{\sigma\in C(t)\cap S_{% n-1}}\operatorname{sgn}(\sigma)T^{\prime}_{i}([\sigma\cdot t])\right)+\left(% \sum_{\sigma\in C(t)\setminus S_{n-1}}\operatorname{sgn}(\sigma)T^{\prime}_{i}% ([\sigma\cdot t])\right).

For ฯƒโˆˆCโข(t)โˆ–Snโˆ’1\sigma\in C(t)\setminus S_{n-1}, Tiโ€ฒโข([ฯƒโ‹…t])=0T_{i}^{\prime}([\sigma\cdot t])=0, since (as described above) ฯƒ\sigma must move nn to a box in a higher row. For ฯƒโˆˆSnโˆ’1\sigma\in S_{n-1}, Tiโ€ฒโข[ฯƒโ‹…t]=[(ฯƒโ‹…t)โˆ–n]=[ฯƒโ‹…(tโˆ–n)]T_{i}^{\prime}[\sigma\cdot t]=[(\sigma\cdot t)\setminus n]=[\sigma\cdot(t% \setminus n)], hence

Tiโ€ฒ(๐žt)=โˆ‘ฯƒโˆˆCโข(t)โˆฉSnโˆ’1sgn(ฯƒ)[ฯƒโ‹…(tโˆ–n))]=โˆ‘ฯƒโˆˆCโข(tโˆ–n)sgn(ฯƒ)[ฯƒโ‹…(tโˆ–n))]=๐žtโˆ–n.T_{i}^{\prime}({\bf e}_{t})=\sum_{\sigma\in C(t)\cap S_{n-1}}\operatorname{sgn% }(\sigma)[\sigma\cdot(t\setminus n))]=\sum_{\sigma\in C(t\setminus n)}% \operatorname{sgn}(\sigma)[\sigma\cdot(t\setminus n))]={\bf e}_{t\setminus n}.

We note also that since every standard Young tableau of shape ฮปโˆ–๐–ผi\lambda\setminus\mathsf{c}_{i} may be obtained by deleting the nn from a standard Young tableau of shape ฮป\lambda with nn in ๐–ผi\mathsf{c}_{i},

Tiโ€ฒโข(spanโข{๐žt|tโˆˆ๐’๐˜๐“ฮปโขandโขtโขhasโขnโขinโขboxโข๐–ผi})\displaystyle T_{i}^{\prime}\big{(}\mathrm{span}\{{\bf e}_{t}\,|\,t\in\mathbf{% SYT}^{\lambda}\;\mathrm{and\;}t\;\mathrm{has\;}n\;\mathrm{in\;box\;}\mathsf{c}% _{i}\}\big{)} =spanโข{๐žt|tโˆˆ๐’๐˜๐“ฮปโˆ–๐–ผi}\displaystyle=\mathrm{span}\big{\{}{\bf e}_{t}\,|\,t\in\mathbf{SYT}^{\lambda% \setminus\mathsf{c}_{i}}\big{\}} (4.20.3)
=๐’ฎฮปโˆ–๐–ผi.\displaystyle={\mathcal{S}}^{\lambda\setminus\mathsf{c}_{i}}.

We denote by T1T_{1} the restriction of T1โ€ฒT_{1}^{\prime} to ๐’ฎฮป{\mathcal{S}}^{\lambda}, and then for all other ii,

Ti+1:=Ti+1โ€ฒ|kerโก(Ti).T_{i+1}:=T_{i+1}^{\prime}|_{\ker(T_{i})}.

This gives rise to a sequence of Snโˆ’1S_{n-1}-invariant subspaces of ๐’ฎฮป{\mathcal{S}}^{\lambda}:

๐’ฎฮปโŠƒkerโก(T1)โŠƒkerโก(T2)โŠƒโ€ฆโŠƒkerโก(Tr).{\mathcal{S}}^{\lambda}\supset\ker(T_{1})\supset\ker(T_{2})\supset\ldots% \supset\ker(T_{r}).

By Lemma 1.2.9 and Proposition 3.11.1 (together with the isomorphism theorem for vector spaces),

(ฯ€,๐’ฎฮป)โ‰…(ฯ€,imโข(T1))โŠ•(ฯ€,kerโก(T1)),(\pi,{\mathcal{S}}^{\lambda})\cong\big{(}\pi,{\mathrm{im}}(T_{1})\big{)}\oplus% \big{(}\pi,\ker(T_{1})\big{)},

and for iโ‰ฅ1i\geq 1,

(ฯ€,kerโก(Ti))โ‰…(ฯ€,imโข(Ti+1))โŠ•(ฯ€,kerโก(Ti+1)).\big{(}\pi,\ker(T_{i})\big{)}\cong\big{(}\pi,{\mathrm{im}}(T_{i+1})\big{)}% \oplus\big{(}\pi,\ker(T_{i+1})\big{)}.

Noting that kerโก(Tr)=0\ker(T_{r})=0, we then have

(ฯ€,๐’ฎฮป)โ‰…(ฯ€,imโข(T1))โŠ•(ฯ€,imโข(T2))โŠ•โ€ฆโŠ•(ฯ€,imโข(Tr)).(\pi,{\mathcal{S}}^{\lambda})\cong\big{(}\pi,{\mathrm{im}}(T_{1})\big{)}\oplus% \big{(}\pi,{\mathrm{im}}(T_{2})\big{)}\oplus\ldots\oplus\big{(}\pi,{\mathrm{im% }}(T_{r})\big{)}.

By (4.20.3), imโข(Ti)=๐’ฎฮปโˆ–๐–ผi{\mathrm{im}}(T_{i})={\mathcal{S}}^{\lambda\setminus\mathsf{c}_{i}}, hence

(ฯ€,๐’ฎฮป)โ‰…(ฯ€,๐’ฎฮปโˆ–๐–ผ1)โŠ•(ฯ€,๐’ฎฮปโˆ–๐–ผ2)โŠ•โ€ฆโŠ•(ฯ€,๐’ฎฮปโˆ–๐–ผr),(\pi,{\mathcal{S}}^{\lambda})\cong\big{(}\pi,{\mathcal{S}}^{\lambda\setminus% \mathsf{c}_{1}}\big{)}\oplus\big{(}\pi,{\mathcal{S}}^{\lambda\setminus\mathsf{% c}_{2}}\big{)}\oplus\ldots\oplus\big{(}\pi,{\mathcal{S}}^{\lambda\setminus% \mathsf{c}_{r}}\big{)},

as claimed. โˆŽ

4.20.2. Induction

Using Theorem 4.20.1 and Frobenius reciprocity, we can now decompose the induction from Snโˆ’1S_{n-1} to SnS_{n} of Specht modules into irreducible representations of SnS_{n}:

Theorem 4.20.4.

For any ฮปโŠขnโˆ’1\lambda\vdash n-1,

IndSnโˆ’1Snโข(ฯ€,๐’ฎฮป)=โจฮป+โŠขn(ฯ€,๐’ฎฮป+),\mathrm{Ind}_{S_{n-1}}^{S_{n}}(\pi,{\mathcal{S}}^{\lambda})=\bigoplus_{\lambda% ^{+}\vdash n}(\pi,{\mathcal{S}}^{\lambda^{+}}),

where the direct sum runs over all ฮป+โŠขn\lambda^{+}\vdash n whose Young diagrams are obtained by adding a single box to the Young diagram for ฮป\lambda.

Proof.

Let ฯ‡ฮป\chi_{\lambda} be the character of (ฯ€,๐’ฎฮป)(\pi,{\mathcal{S}}^{\lambda}). For any ฮผโŠขn\mu\vdash n, by Corollary 3.15.2, we have

โŸจIndSnโˆ’1Snโขฯ‡ฮป,ฯ‡ฮผโŸฉSn=โŸจฯ‡ฮป,ResSnโˆ’1Snโกฯ‡ฮผโŸฉSnโˆ’1.\langle\mathrm{Ind}_{S_{n-1}}^{S_{n}}\chi_{\lambda},\chi_{\mu}\rangle_{S_{n}}=% \langle\chi_{\lambda},\operatorname{Res}_{S_{n-1}}^{S_{n}}\chi_{\mu}\rangle_{S% _{n-1}}.

By Theorem 4.20.1, โŸจฯ‡ฮป,ResSnโˆ’1Snโกฯ‡ฮผโŸฉSnโˆ’1=1\langle\chi_{\lambda},\operatorname{Res}_{S_{n-1}}^{S_{n}}\chi_{\mu}\rangle_{S% _{n-1}}=1 if ฮป\lambda is obtained by removing a box from ฮผ\mu (which is equivalent to ฮผ\mu being obtained by adding a box to ฮป)\lambda), and zero otherwise. โˆŽ

Remark 4.20.5.

The relation between restriction and induction of Specht modules is displayed pictorially by Youngโ€™s lattice.

4.20.3. Exercises

.

Problemย 99. Decompose IndS4S6โข(ฯ€,๐’ฎ(3,1))\mathrm{Ind}_{S_{4}}^{S_{6}}\left(\pi,{\mathcal{S}}^{(3,1)}\right) into irreducible representations of S6S_{6}.

Problemย 100. Let T:Vโ†’WT:V\rightarrow W be a GG-homomorphism between two representations (ฯ€,V)(\pi,V) and (ฯ,W)(\rho,W). Show that

(ฯ€,V)โ‰…(ฯ€,kerโก(T))โŠ•(ฯ,imโข(T)).(\pi,V)\cong\big{(}\pi,\ker(T)\big{)}\oplus\big{(}\rho,{\mathrm{im}}(T)\big{)}.

Problemย 101. Show that the maps Tiโ€ฒT_{i}^{\prime} and TiT_{i} defined during the proof of Theorem 4.20.1 are Snโˆ’1S_{n-1}-homomorphisms.