Induction and Restriction

Arun Ram
Department of Mathematics and Statistics
University of Melbourne
Parkville, VIC 3010 Australia
aram@unimelb.edu.au

Last update: 6 November 2012

Induction and Restriction

Let A be a subalgebra of an algebra B.

Let V be a representation of B. The restriction V ↓AB of V to A to be the representation of A given by the action of A on V. Let W be a representation of A. Define B⊗AW to be all formal linear combinations of elements b⊗w, where b∈B, w∈W with the relations

(b1+b2)⊗w =(b1⊗w)+ (b2+w), b⊗(w1+w2) =(b⊗w1)+ (b⊗w2), (αb)⊗w=b⊗ (αw)=α (b⊗w), ba⊗w=b⊗aw, (5.1)

for all a∈A, b,b1,b2∈B, w,w1,w2∈W and α∈ℂ. The induced representation W ↑AB is the representation of B on B⊗AW given by the action

b(b′⊗w)= (bb′)⊗w, (5.2)

for all b,b′∈B and w∈W.

(5.3) Proposition. Let A⊂B⊂C be such that A is a subalgebra of B and B is a subalgebra of C. Let V,V1,V2 be representations of C and let W,W1,W2 be representations of C.

1) (V1⊕V2) ↓AC≅V1 ↓AC ⊕V2 ↓AC. 2) (V↓BC) ↓AB≅V ↓AC. 3) (V1⊕V2) ↑AB=V1 ↑AB⊕V2 ↑AB. 4) (V↑AB) ↑BC≅V ↑AC.

Proof.

1) and 2) are trivial consequences of the definition. he fact that the map

ϕ: B⊗A (V1⊕V2) ⟶ (B⊗AV1)⊕ (B⊗AV2) b⊗(v1,v2) ⟼ ( b⊗v1,b ⊗v2 ) .

is a B-module isomorphism gives 3). The map

ϕ1: C⊗B (B⊗AV) ⟶ (C⊗BB) ⊗AV c⊗(b⊗v) ⟼ (c⊗b)⊗v

and the map

ϕ2: C⊗BB ⟶ C c⊗b ⟼ cb

are both C-module isomorphisms. So

C⊗B (B⊗AV)≅ (C⊗BB) ⊗AV≅C⊗A V,

giving 4).

□

Note: Proving that these maps are isomorphisms is not a complete triviality. One must show that they are well defined (by showing that they preserve the bilinearity relations (5.1)) and that the inverse maps are also well defined. It is helpful to use the fact that the tensor product is a universal object as given in Ex. 1.

(5.4) Theorem. (Frobenius reciprocity) Let A⊂B be algebras and Vλ and Wμ be irreducible representations of A and B respectively. Then

HomB ( Vλ ↑AB,, Wμ ) ≅ HomA ( Vλ,Wμ ↓AB ) .

Proof.

The map

Ψ: HomB ( B⊗AVλ, Wμ ) ⟶ HomA ( Vλ,Wμ ↓AB ) ϕ ⟼ ϕ′ ,

where

ϕ′(v)= ϕ(1⊗v),

is an isomorphism. The inverse map is given by Ψ-1 (ϕ′)=ϕ where ϕ is given by

ϕ(b⊗v)=bϕ (1⊗v)=bϕ′ (v),

so that ϕ is a B-module homomorphism.

□

Branching rules

Now suppose that A is a subalgebra of B and that both A and B are semisimple. Let A^ and B^ be index sets for the irreducible representations of A and B respectively. Let Vλ and Wμ be the irreducible representations of A and B labelled by λ∈A^ and μ∈B^ respectively. Let gλμ∈ℤ be such that

Vλ ↑AB≅ ⊕μ∈B^ GλμWμ (5.5)

for each pair (λ,μ), λ∈A^, μ∈B^. Frobenius reciprocity implies that

Wμ↓AB≅ ⊕λ∈A^ gλμVλ (5.5')

for each μ∈B^. An equation of the form (5.5) or (5.5') is called a branching rule between A and B.

One can produce a visual representation of branching rules in the form of a graph. Construct a graph with two rows of vertices, the vertices in the first row labelled by the elements of A^ and the vertices of the second row labelled by the elements of B^ such that the vertex labelled by λ∈A^ and the vertex labelled by μ∈B^ are connected by gλμ edges. This graph is the Bratteli diagram of A⊂B.

As an example, the following diagram is the Bratteli diagram of ℂS2⊂ℂS3, where Sn denotes the symmetric group. Recall that the irreducible representations of S2 and S3 are indexed by partitions of 2 and of 3 respectively.

S2 S3 (111) (21) (3) (2) (11)

Note that in this example each gλμ is either 0 or 1; there are no multiple edges.

Let p∈A and consider the representation of A given by left multiplication on the space Aa. Then

(Ap) ↑AB≅ Bp. (5.6)

To see this, informally, one notes that since Ap⊂A we can move Ap across the tensor product to give,

(Ap)↑AB= B⊗AAp=BAp ⊗A1=Bp⊗A 1≅Bp.

BAp=Bp since 1∈A. More formally we should show that the map

B⊗AAp ⟶ Bp b⊗ap ⟼ bap

is well defined and has well defined inverse given by

bop⟵bp.

Now let pλ be a minimal idempotent of A such that the action of A by left multiplication on Apλ is a representation of A isomorphic to the irreducible representation Vλ of A (3.6). Suppose that

pλ=∑qi

is a decomposition (§1 Ex. 7) of the minimal idempotent pλ of A into minimal orthogonal idempotents of B. Then Bpλ=B∑qi=∑Bqi gives a decomposition of Bpλ into irreducible representations. So, by (5.6) and the branching rule (5.5), for exactly gλμ of the qi we will have that Bqi is isomorphic to the irreducible representation Wμ of B. We can write the decomposition of pλ as

pλ= ∑μ∈B^ ∑i=1gλμ qμi (5.7)

where each qμi is such that Bqμi is isomorphic to the irreducible representation Wμ of B.

Characters of induced representations

Let V be a representation of A where A is a subalgebra of an algebra B and both A and B are semisimple. Let χV be the character of V and let χV↑BA be the character of V↑AB. For each A∈𝒜 let a* denote the element of the dual basis to 𝒜 with respect to the trace, tr, of the regular representation of A such that tr(aa*)=1.

Let ℬ be a basis of B and let t→B=(tμB) be a nondegenerate trace on B. For each b∈ℬ let b* denote the element of the dual basis to ℬ with respect to the trace t→B such that t→(bb*)=1. For any element x∈B we set (as in §3 EX.7)

[x]=∑b∈ℬ bxb*.

(5.8) Theorem.

χV↑AB (b)=munde ∑a χV(a) ⟨ [b],a* ⟩ ,

where ⟨b1,b2⟩ =t→B (b1b2).

Proof.

In keeping with the notations of earlier sections, let A^ and B^ be index sets for the irreducible representations of A and B respectively and let χAλ,λ∈A^ and χBμ,μ∈B^ denote the irreducible characters of A and B respectively. Let zλA,λ∈A^ and zμB,μ∈B^ denote the minimal central idempotents of A and B respectively. Let dλA=χAλ(1) so that dλ is the dimension of the irreducible representation of A corresponding to λ∈A^.

We have the following facts:

  1. (Theorem (3.10)) For each λ∈A^, μ∈B^, zλA = ∑a∈𝒜 tλA χAλ (a) a* ,and zμB = ∑b∈ℬ tμB χBμ (b) b* , respectively.
  2. (§3 Ex. 5) The trace vector )(tλA) of the trace of the regular representation of A is given by tλA=dλA for all λ∈A^.
  3. Suppose that V≅⊕λ∈A^Vλ⊕mλ gives the decomposition of V into irreducible representations of A. Then χV(a)= ∑λ∈A^ mλχAλ (a),

    for all a∈A.

  4. The branching rule (5.5) for A⊂B gives that χV↑AB (b)= ∑λ∈A^mλ ∑μ∈B^ gλμχBμ (b), for all b∈B.
  5. For each λ∈A^ let zλA= ∑i=1dλA pλiA be a decomposition of zλA into minimal orthogonal idempotents of A. For each λ∈A^ and 1≤i≤dλA let pλiA= ∑μ∈B^ ∑j=1gλμ qμjB be a decomposition (5.7) of pλiA into minimal orthogonal idempotents of B. qμj denotes a minimal idempotent in the minimal ideal of B corresponding to μ∈b^, i.e., a minimal idempotent such that the representation Bqμj of B is isomorphic to the irreducible representation of B corresponding to μ∈B^. Then, by (3.12), [qμjB]= (1/tμB) zμB, for each minimal idempotent qμjB, since for each ν∈B^, χν(qμjB) =δμν.
  6. Let b1,b2∈B. Using the trace property, ⟨ [b1], b2 ⟩ = t→B ( ∑b∈ℬ bb1b*b2 ) = t→B ( ∑b∈ℬ b1b*b2 b ) = ⟨ b1,[b2] ⟩ . Now, define z=∑λ∈A^ (mλ/dλA) zλA. Then, using 1), 2) and 3), z = ∑λ∈A^ mλ∑a (tλA/dλA) χAλ(a)a* = ∑aχV (a)a*, and, by 5), 1) and 4), [z] = ∑λ (mλ/dλA) [zλA] = ∑λ (mλ/dλA) ∑i=1dλA [pλiA] = ∑λ (mλ/dλA) ∑i=1dλA ∑μ ∑j=1gλμ [qμjB] = ∑λ (mλ/dλA) ∑i=1dλA ∑μ ∑j=1gλμ (1/tμB) zμB = ∑λ∑μ (mλ/dλA) dλAgλμ (1/tuB) ∑btμB χBμ(b)b* = ∑bχV↑AB (b)b*. Combining these and using 6) we get χV↑AB (b) = ⟨ [z],b ⟩ = ⟨ [ ∑aχV (a)a* ] ,b ⟩ = ∑aχV(a) ⟨ [a*],b ⟩ = ∑aχV(a) ⟨ a*,[b] ⟩ , as desired.

□

Centralizers

Let A be a subalgebra of an algebra B, and let B be a representation of B. Let A‾ and B‾ be the centralizers of V(A) and V(B) respectively. Then B‾ is a subalgebra of A‾; A⊂B and A‾⊃B‾.

(5.9) Theorem. Suppose that

Wμ ↓AB ≅ ∑λgμλ Vλ, V‾λ ↓B‾A‾ ≅ ∑μgλμ′ W‾μ

are the branching rules for A⊂B and B‾⊂A‾ respectively. Then for all λ,μ

gμλ= gλμ′.

Proof.

We know, Theorem (4.11), that, as A⊗A‾ representations,

V≅⊕λVλ⊗ V‾λ,

and as B⊗B‾ representations,

V≅⊕μWμ ⊗W‾μ,

where Vλ, V‾λ, Wμ, and W‾μ are irreducible representations of A,A‾,B, and B‾ respectively.

A⊗B‾ is a subalgebra of both A⊗A‾ and B⊗B‾. We have that as A⊗B‾ representations

V≅V ↓ A⊗B‾ A⊗A‾ ≅ ⊕λVλ⊗ ( ⊕μgλμ′ W‾μ ) ≅ ⊕λ,μ gλμ′ Vλ⊗W‾μ.

On the other hand as A⊗B‾ representations

V≅V ↓ A⊗B‾ B⊗B‾ ≅ ⊕μ ( ⊕λ gμλ Vλ ) ⊗W‾μ ≅ ⊕λ,μ gμλVλ ⊗W‾μ.

□

Examples

  1. Let A,B and C be vector spaces. A map f:A×B→C is bilinear if f a 1 + a 2 b =f a 1 b +f a 2 b , f a b 1 + b 2 =f a b 1 +f a b 2 , f αa b =f a αb =αf ab , for all a, a 1 , a 2 ∈A,b, b 1 , b 2 ∈B,α∈ℂ.
  2. The tensor product is given by a vector space A⊗B and a map i:A×B→A⊗B such that for every bilinear map f:A×B→C there exists a linear map f - :A⊗B→C such that the following diagram commutes:

    A×B C A⊗B i f f

    One constructs the tensor product A⊗B as the vector space of elements a⊗b,a∈A,b∈B, with relations a 1 + a 2 ⊗b= a 1 ⊗b+ a 2 ⊗b, a⊗ b 1 + b 2 =a⊗ b 1 +a⊗ b 2 , αa b=a⊗ αb =α a⊗b , for all a, a 1 , a 2 ∈A,b, b 1, b 2 ∈B and α∈ℂ. The map i:A×B→A⊗B is given by i ab =a⊗b. Using the above universal mapping property one gets easily that the tensor product is unique in the sense that any two tensor products of A and B are isomorphic.

    If R is an algebra and A is a right R -module (a vector space that affords an antirepresentation of R ) and B is a left R -module them one forms the vector space A ⊗ R B as above except that we require a bilinear map f:A×B→C to satisfy the additional condition f ar b =f a rb for all r∈R. Then the tensor product A ⊗ R B once again is constructed by using the vector space of elements a⊗b,a∈A,b∈B, with the relations above and the additional relation ar⊗b=a⊗rb, for all r∈R.

  3. Let A⊆B be semisimple algebras such that A is a subalgebra of B Let A ~ and B ~ be index sets of the irreducible representations of A and B respectively, and suppose that f ij μ ,μ∈ A ~ , is a complete set of matrix units of A

    [Bra1972] There exists a complete set of matrix units e rs λ ,λ∈ B ~ , of B that is a refinement of the f ij μ in the sense that for each μ∈ A ~ and each i , f ii μ =∑ e rr λ , for some set of e rr λ .

    Proof.
    Suppose that B≅ ⊕ λ∈ B ~ M d λ ℂ . Let z λ B be the minimal central idempotent of B such that I λ = B z λ is the minimal ideal corresponding to the λ block of matrices in ⊕ λ M d λ ℂ .

    For each μ∈ A ~ and each i decompose f ii μ into minimal orthogonal idempotents of B (Section 1, Ex 7), f ii μ =∑ p j . Label each p j appearing in this sum by the element λ∈ B ~ which indexes the minimal ideal I λ =B p j B of B . Then 1= ∑ μ,i f ii μ = ∑ λ∈ B ~ ∑ j=1 d λ p j λ . Now B=1.B.1= ∑ λ,μ∈ B ~ ∑ 1≤i≤ d λ ,1≤j≤ d μ p i λ B p j μ . If λ≠μ then the space p i λ B p j μ = p i λ B z μ B p j μ = p i λ z μ B B p j μ =0 for all i,j. Since p i λ = p i λ .1. p i λ ∈ p i λ I p i λ and p i λ B p j λ p j λ B p i λ = p i λ I λ p i λ ≠0, we know that p i λ B p j λ is not zero for any 1≤i,j≤ d λ . Furthermore, since the dimension of B is ∑ λ d λ 2 each of the spaces p i λ B p j λ is one dimensional.

    For each p i λ define e ii λ = p i λ . For each λ and each 1≤i<j≤ d λ let e ii λ be some element of p i λ B p j λ . Then choose e ii λ ∈ p j λ B p i λ such that e ij λ e ji λ = e ii λ . This defines a complete set of matrix units of B. □

  4. Let G be a finite group and let H be a subgroup of G. Let R= g i be a set of representatives for the left cosets gH of H in G. The action of G on the cosets of H in G by left multiplication defines a representation π H of H in G. This representation is a permutation representation of G. Let g∈G. The entries π H g i'i of the matrix π H g are given by π H g i'i = δ i'k where k is such that g g i ∈ g k H.

    Let V be a representation of H. Let B= v j be a basis of V. Then the elements g⊗ v j where g∈G, v j ∈B span ℂG ⊗ ℂH V. The fourth relation in 5.1 gives that the set g i ⊗ v j , g i ∈R, v j ∈B forms a basis of ℂG ⊗ ℂH V.

    Let g∈G and suppose that g g i = g k h, where h∈H and g k ∈R. Then g g i ⊗ v j = g k h⊗ v j = g k ⊗h v j = ∑ j g k ⊗ v j' V h j'j = ∑ i',j' g i' ⊗ v j' V h j'j δ i'k = ∑ i',j' g i' ⊗ v j' V h j'j π H g i'i . Then χ V ↑ H G g = ∑ g i ∈R, v j ∈B g g i ⊗ v j | g i ⊗ v j = ∑ g i , v j ,g g i ∈ g i H V g i -1 g g i jj .

    Since characters are constant on conjugacy classes we have that χ V ↑ H G g = 1 H ∑ h∈H ∑ g i ; h -1 g i -1 g g i h∈H χ V h -1 g i -1 g g i h = 1 H ∑ a∈H,a∈ C g χ V a , where C g denotes the conjugacy class of g. This is an alternate proof of Theorem 5.8 for the special case of inducing from a subgroup H of a group G to the group G.

  5. Define ℂG ⊗ d ℂG to be the subalgebra of the algebra ℂG⊗ℂG consisting of the span of the elements g⊗g , g∈G. Then ℂG≅ℂG ⊗d ℂG as algebras.

    Let V 1 and V 2 be representations of G. Then the restriction of the ℂG⊗ℂG representation V= V 1 ⊗ V 2 to the algebra ℂG ⊗ d ℂG is the Kronecker product (Section 4, Ex 1) V 1 ⊗ d V 2 = V 1 ⊗ V 2 ↓ ℂG⊗ℂG ℂG ⊗ d ℂG of V 1 and V 2 . Since ℂG≅ℂG ⊗ d ℂG we can view V 1 ⊗ d V 2 as a representation of G.

    Let V λ and V μ be irreducible representations of G such that V λ ⊗ v μ appears as an irreducible component of the ℂG⊗ℂG representation V 1 ⊗ V 2 . The decomposition of the Kronecker product V λ ⊗ d V μ = V 1 ⊗ V 2 ↓ ℂG⊗ℂG ℂG ⊗ d ℂG ≅ ⊕ ν g λμ ν V ν into irreducible representations V ν of G is given by the branching rule for ℂG⊗ℂG⊃ℂG ⊗ d ℂG. Let C 1 and C 2 be the centralisers of the representations V 1 and V 2 respectively. Let C be the centraliser of the ℂG⊗ℂG representation V= v 1 ⊗ V 2 . Applying Theorem 5.9 to V where A=ℂG⊗ℂG and ℂG ⊗ d ℂG=B≅G shows that the g λμ ν are also given by the branching rule for C 1 ⊗ C 2 ⊂C.

Notes and References

The main result, Theorem (5.8), of this section is a generalization of the formula for the induced character for finite groups, see [Ser1977] §7.2. I have been unable to find any similar result in previous literature.

Notes and References

This is an excerpt from the unpublished first chapter of Arun Ram's dissertation entitled Representation Theory, written July 4, 1990.

page history